MEN'S BASKETBALL Dr. G&W Analysis: What Is the Best MSU Team That Didn't Win a Title?
- Spartans Illustrated Message Board
- 23 Replies
Do to the unprecedented events surrounding the COVID-19 pandemic, we will never know who would have won the 2020 Men's NCAA Basketball Tournament. Granted, MSU was clearly surging, and Joe Lunardi, Nate Silver / 538, and my own simulation all came to the conclusion that MSU would have cut down the nets, but in reality the 2020 Tournament will go down as another example of "what could have been" for MSU fans.
Unfortunately, the history of MSU basketball is littered with these types of "what ifs." For example:
What if...?
While all of those teams referenced above were certainly good, today I would like to tackle the following question:
Which MSU team in the past was actually the best team that didn't win a National Title?
While this is a fun question to debate, it occurs to me that there is a mathematical way to answer that question. While it is always tricky to compare different teams from different years, advanced metrics, such as efficiency margins, provide a possible way to make this comparison.
Late in the season this year, before the world went haywire, I made a plot comparing all Tom Izzo coached MSU teams based on Kenpom's adjusted offensive and defensive efficiency values (with the overall efficiency margin shown on the diagonals). That plot is shown here:
For reference, the shaded green area is the region where all of the NCAA Champions back to 2002 have fallen. While there is no available efficiency data for any of Jud's old MSU teams (sorry, Coach Skiles) all of Coach Izzo's MSU teams are shown. As one might expect, the 2000 Championship team had the highest overall efficiency margin of 33.6, but three other teams were also over 30:
However, there is a slightly different way to look at and define this problem. While the 2019 team was very good last year, MSU finished the regular season ranked "only" 4th overall in Kenpom, behind Virginia, Gonzaga, and Duke. The 2019 field was over all very strong. So, while the 2019 team was, on paper, the best Tom Izzo team of the "Kenpom Era" (back to 2002 where the data is easily available) the odds of that team winning the tournament were not that high, simply due to the level of competition.
Just for fun, last year I ran a series of Monte Carlo simulations of each NCAA Tournament back to 2002 using Kenpom efficiency data. One of the many pieces of data that I extracted from these simulations are the odds for any given team to make the Final Four and to win the tournament. Below, I tabulated the Top 25 teams back to 2002 with respect to National Title odds, as well as all the MSU teams in the Top 300 (out of over 1,200). Teams that won the National Title are shaded in yellow.
There are several very interesting things about this data set. For example, the team with the highest odds to win the tournament in the past 18 years was the 2015 Kentucky team. However, their odds were "only" about 35 percent. While this number may seem low, it's over five percentage points higher than the team in second place (the 2002 Duke squad) and well above the average for No. 1 seeds as a group, which is only about 12.5 percent. Also note that neither the 2015 Kentucky team nor the 2002 Duke team wound up cutting down the nets in early April.
Regarding Michigan State, a quick look down the list shows a bit of a surprise. Of all of Tom Izzo's teams in the past 18 years, the one with the best odds to win the title was the ill-fated 2016 team. That team finished the year ranked No. 2 in Kenpom and had a 16.1% chance to win the title, which just also happens to be the highest odds of any No. 2 seed in the Kenpom Era.
In second place in this group is the 2012 MSU team (which finished the regular season ranked No. 3 in Kenpom) at 11.7%. However, MSU accumulated these stats almost exclusively with a healthy Brandon Dawson, so the odds likely reflect a team where he was healthy in March. In third place was last year's team checking in at 10.0%, and in 4th place is the 2018 team (5.7%) that couldn't solve the Syracuse zone in the second round at Little Caesar's Arena.
So while there is a strong argument to be made that the 1999 MSU team is the best MSU team not to win a National Title, I am going to go with the 2016 team as MSU that was most likely to win a National Title, yet didn't.
Remembering the 2015-2016 MSU Squad
While most MSU fans try to block the memory of the 2016 team out of their mind, I think that it is appropriate to remind everyone of exactly how good that team was and how their season went down. Of course, the team was led by senior point guard Denzel Valentine (who averaged 19.2 point per game, 7.5 rebounds per game, and 7.8 assists per game) was a consensus first team All American and won several National Player of the Year awards. The team also featured follow senior guard Bryn Forbes (14.4 ppg) and senior center Matt Costello (10.7 ppg and 8.2 rpg). The starting line-up also consisted of junior guard Eron Harris (9.3 ppg) and one-and-done freshman Dayonta Davis (7.5 ppg and 5.5 rpg).
Players off the bench that saw some time were freshman Matt McQuaid, sophomore Tum Tum Nairn, sophomore Javon Bess, junior Gavin Shilling, sophomore Marvin Clark, freshman Kenny Goins, junior Alvin Ellis, senior walk-on Colby Wollenmen, and freshman Kyle Ahrens. The team also featured six different players who all shot at least 40 percent from three: Forbes (48.1%), Valentine (44.4%), Harris (43.9%), McQuaid (40.9%), Marvin Clark (42.3%), and Alvin Ellis (40.0%) although Clark and Ellis only took a combined 61 shots all year.
As for the 2016 season, MSU started ranked No. 13 in the AP poll with somewhat limited expectations, following a surprise run to the Final Four the previous year. The season then started with a bang, as MSU upset No. 4 ranked Kansas in the Champions Classic, aided by a triple double from Valentine and a perfect (and surprising) three-for-three shooting from the long line by freshman Matt Mcquaid. MSU soon shot up to No. 3 in the AP and after beating No. 24 ranked Louisville in the ACC-Big Ten challenge, they took over the No. 1 shot in the AP poll, which they held for five weeks.
Prior to the game against Oakland, Denzel was sidelined with a knee injury. MSU was able to avoid an upset to the Grizzlies in overtime, but the Spartan struggled for several games even after Valentine returned. By late January, MSU was sitting as 3-4 in Big Ten play and had sunk to No. 12 in the AP poll. MSU fans, as usual, freaked out.
Then came a home game against Maryland on January 23rd. Maryland was ranked in the Top 10 and MSU was on a three-game losing streak and decided to wear neon green jerseys. It could have been an absolute disaster. But, it wasn't. MSU beat the Terp, 74-65. Matt Costello hugged Izzo. The Spartans were back.
MSU proceeded to win nine of the final ten games in the regular season, losing only at Purdue in over-time by a point. MSU would later avenge this loss in the Big Ten Tournament Final as they cut down the nets on Selection Sunday. While MSU did not claim a regular season title (Indiana somehow finished with only three loses) they still hung a banner and finished the regular season ranked No. 2 in the AP poll.
As for the 2016 Tournament itself, let's just say that what could go wrong, did. First off, in an ironic twist, this was the same year that former MSU athletic director, Mark Hollis, acted as the chairman of the selection committee. Unfortunately, it always felt to me that this did not at all play in MSU's favor.
As stated above, on Selection Sunday, MSU was ranked No. 2 in the AP poll and No. 2 in Kenpom and had just won the Big Ten Tournament. Virtually all of MSU five loses came when Valentine was out with, or recovering from, an injury. A No. 1 seed seemed like a no-brainer. But, when the brackets came out, MSU somehow was given the No. 2 seed in the Midwest. A six-loss Oregon team and a seven-loss Virginia team both somehow got No. 1 seeds over MSU. It seems likely that the committee was afraid that granting MSU a No. 1 seed would give the appearance of favoritism since Hollis was the Chair.
That said, as a No. 2 seed, MSU's first round opponent, No. 15 seed Middle Tennessee State should not have been a problem. MSU's odds to win the game were slightly above 95%, But, things did not go as planned. The Blue Raiders went on a 13-0 run early and MSU never quite gained their footing. MTSU shot a blistering 11-19 (57.9%) from three for the game. MSU cut the lead to just three points late, but could never draw even. The rest is history.
Another ironic part of the story is that is was never clear to me why MSU was paired with MTSU in the first place. As the top ranked No. 2 seed, based on the s-curve, MSU should have naturally been paired with the weakest No.15 seed. In 2016, MTSU was actually the strongest of the No. 15 seeds, based on the committee's published seed list.
The exact placement of the No. 15 seeds is often driven by geography, and in this case the weakest No. 15 seed was Weber State. However, Weber State was matched against Xavier in the first round and both this game and the MSU-MTSU both took place in the same pod in St. Louis. Xavier was actually ranked as the weakest No. 2 seed. MSU and Xavier, by basic seeding principles, should have traded opponents. If I even meet Hollis in person, I will ask him why this decision was made.
Had MSU made it past MTSU there is no guarantee that would have won the Title or would have even made it past No. 10 Syracuse in the second round. MSU would have projected as a nine-point favorite in that game, but Izzo teams have always struggled with Boeheim's zone. Then again, with so many three-point snipers on the team, I tend to think Valentine and crew would have gotten the job done.
In the Sweet Sixteen, MSU would have next faced No. 11 Gonzaga, where MSU would have been a seven-point favorite. Then, MSU would have once again faced Tony Bennett and No. 1 Virginia. I project the MSU would have been slightly favored in this game too. In reality, UVA blew a 15-point lead over Syracuse with 10 minutes to play. Is there any way that the Cavaliers would have not blown it against MSU as well? I don't think so.
Had MSU made the Final Four in 2016, the path would have been a tough one. In the National Semifinals, they would have face No. 1 North Carolina and then No. 2 Villanova in the Finals. Based on the pre-tournament Kenpom numbers, MSU would have been a narrow favorite in both contests. In reality, it is hard to say, as both UNC and Villanova played very well in March.
While we will never know if MSU would have earned Izzo his second title in 2016 had the Middle Tennessee State debacle not happened, based on my math, this was his most likely chance since at least 2001.
Thus ends the lesson for today. As always, stay safe, stay home, wash your hands, and Go Green.
Unfortunately, the history of MSU basketball is littered with these types of "what ifs." For example:
- What if the clock hadn't stopped in Kansas City in the 1986 Sweet Sixteen, allowing top seeded Kansas to force over time against Scott Skiles' scrappy MSU squad?
- What if the officials would have noticed that Kenny Anderson's shot to force over time in the 1990 Sweet Sixteen game against Steve Smith's team actually was launched after the buzzer?
- What if Alan Anderson hadn't hurt his knee during the 2005 NCAA Tournament run?
- What if Kalin Lucas didn't blow out his Achilles' tendon in the 2010 Sweet Sixteen game against Maryland... (Also, was is it with the Sweet Sixteen?)
- What if Brandon Dawson didn't tear his ACL against Ohio State in the Big Ten finale in Draymond Green's Senior season of 2012?
- What if Keith Appling hadn't hurt his wrist in the ACC-Big Ten Challenge early in the 2013-14 season?
- What if Nick Ward and Kyle Arhens hadn't been injured late in the 2019 campaign?
What if...?
While all of those teams referenced above were certainly good, today I would like to tackle the following question:
Which MSU team in the past was actually the best team that didn't win a National Title?
While this is a fun question to debate, it occurs to me that there is a mathematical way to answer that question. While it is always tricky to compare different teams from different years, advanced metrics, such as efficiency margins, provide a possible way to make this comparison.
Late in the season this year, before the world went haywire, I made a plot comparing all Tom Izzo coached MSU teams based on Kenpom's adjusted offensive and defensive efficiency values (with the overall efficiency margin shown on the diagonals). That plot is shown here:
For reference, the shaded green area is the region where all of the NCAA Champions back to 2002 have fallen. While there is no available efficiency data for any of Jud's old MSU teams (sorry, Coach Skiles) all of Coach Izzo's MSU teams are shown. As one might expect, the 2000 Championship team had the highest overall efficiency margin of 33.6, but three other teams were also over 30:
- The 1999 Final Four Team (31.6)
- The 2019 Final Four Team (30.8)
- The 2001 Final Four Team (30.7)
However, there is a slightly different way to look at and define this problem. While the 2019 team was very good last year, MSU finished the regular season ranked "only" 4th overall in Kenpom, behind Virginia, Gonzaga, and Duke. The 2019 field was over all very strong. So, while the 2019 team was, on paper, the best Tom Izzo team of the "Kenpom Era" (back to 2002 where the data is easily available) the odds of that team winning the tournament were not that high, simply due to the level of competition.
Just for fun, last year I ran a series of Monte Carlo simulations of each NCAA Tournament back to 2002 using Kenpom efficiency data. One of the many pieces of data that I extracted from these simulations are the odds for any given team to make the Final Four and to win the tournament. Below, I tabulated the Top 25 teams back to 2002 with respect to National Title odds, as well as all the MSU teams in the Top 300 (out of over 1,200). Teams that won the National Title are shaded in yellow.
There are several very interesting things about this data set. For example, the team with the highest odds to win the tournament in the past 18 years was the 2015 Kentucky team. However, their odds were "only" about 35 percent. While this number may seem low, it's over five percentage points higher than the team in second place (the 2002 Duke squad) and well above the average for No. 1 seeds as a group, which is only about 12.5 percent. Also note that neither the 2015 Kentucky team nor the 2002 Duke team wound up cutting down the nets in early April.
Regarding Michigan State, a quick look down the list shows a bit of a surprise. Of all of Tom Izzo's teams in the past 18 years, the one with the best odds to win the title was the ill-fated 2016 team. That team finished the year ranked No. 2 in Kenpom and had a 16.1% chance to win the title, which just also happens to be the highest odds of any No. 2 seed in the Kenpom Era.
In second place in this group is the 2012 MSU team (which finished the regular season ranked No. 3 in Kenpom) at 11.7%. However, MSU accumulated these stats almost exclusively with a healthy Brandon Dawson, so the odds likely reflect a team where he was healthy in March. In third place was last year's team checking in at 10.0%, and in 4th place is the 2018 team (5.7%) that couldn't solve the Syracuse zone in the second round at Little Caesar's Arena.
So while there is a strong argument to be made that the 1999 MSU team is the best MSU team not to win a National Title, I am going to go with the 2016 team as MSU that was most likely to win a National Title, yet didn't.
Remembering the 2015-2016 MSU Squad
While most MSU fans try to block the memory of the 2016 team out of their mind, I think that it is appropriate to remind everyone of exactly how good that team was and how their season went down. Of course, the team was led by senior point guard Denzel Valentine (who averaged 19.2 point per game, 7.5 rebounds per game, and 7.8 assists per game) was a consensus first team All American and won several National Player of the Year awards. The team also featured follow senior guard Bryn Forbes (14.4 ppg) and senior center Matt Costello (10.7 ppg and 8.2 rpg). The starting line-up also consisted of junior guard Eron Harris (9.3 ppg) and one-and-done freshman Dayonta Davis (7.5 ppg and 5.5 rpg).
Players off the bench that saw some time were freshman Matt McQuaid, sophomore Tum Tum Nairn, sophomore Javon Bess, junior Gavin Shilling, sophomore Marvin Clark, freshman Kenny Goins, junior Alvin Ellis, senior walk-on Colby Wollenmen, and freshman Kyle Ahrens. The team also featured six different players who all shot at least 40 percent from three: Forbes (48.1%), Valentine (44.4%), Harris (43.9%), McQuaid (40.9%), Marvin Clark (42.3%), and Alvin Ellis (40.0%) although Clark and Ellis only took a combined 61 shots all year.
As for the 2016 season, MSU started ranked No. 13 in the AP poll with somewhat limited expectations, following a surprise run to the Final Four the previous year. The season then started with a bang, as MSU upset No. 4 ranked Kansas in the Champions Classic, aided by a triple double from Valentine and a perfect (and surprising) three-for-three shooting from the long line by freshman Matt Mcquaid. MSU soon shot up to No. 3 in the AP and after beating No. 24 ranked Louisville in the ACC-Big Ten challenge, they took over the No. 1 shot in the AP poll, which they held for five weeks.
Prior to the game against Oakland, Denzel was sidelined with a knee injury. MSU was able to avoid an upset to the Grizzlies in overtime, but the Spartan struggled for several games even after Valentine returned. By late January, MSU was sitting as 3-4 in Big Ten play and had sunk to No. 12 in the AP poll. MSU fans, as usual, freaked out.
Then came a home game against Maryland on January 23rd. Maryland was ranked in the Top 10 and MSU was on a three-game losing streak and decided to wear neon green jerseys. It could have been an absolute disaster. But, it wasn't. MSU beat the Terp, 74-65. Matt Costello hugged Izzo. The Spartans were back.
MSU proceeded to win nine of the final ten games in the regular season, losing only at Purdue in over-time by a point. MSU would later avenge this loss in the Big Ten Tournament Final as they cut down the nets on Selection Sunday. While MSU did not claim a regular season title (Indiana somehow finished with only three loses) they still hung a banner and finished the regular season ranked No. 2 in the AP poll.
As for the 2016 Tournament itself, let's just say that what could go wrong, did. First off, in an ironic twist, this was the same year that former MSU athletic director, Mark Hollis, acted as the chairman of the selection committee. Unfortunately, it always felt to me that this did not at all play in MSU's favor.
As stated above, on Selection Sunday, MSU was ranked No. 2 in the AP poll and No. 2 in Kenpom and had just won the Big Ten Tournament. Virtually all of MSU five loses came when Valentine was out with, or recovering from, an injury. A No. 1 seed seemed like a no-brainer. But, when the brackets came out, MSU somehow was given the No. 2 seed in the Midwest. A six-loss Oregon team and a seven-loss Virginia team both somehow got No. 1 seeds over MSU. It seems likely that the committee was afraid that granting MSU a No. 1 seed would give the appearance of favoritism since Hollis was the Chair.
That said, as a No. 2 seed, MSU's first round opponent, No. 15 seed Middle Tennessee State should not have been a problem. MSU's odds to win the game were slightly above 95%, But, things did not go as planned. The Blue Raiders went on a 13-0 run early and MSU never quite gained their footing. MTSU shot a blistering 11-19 (57.9%) from three for the game. MSU cut the lead to just three points late, but could never draw even. The rest is history.
Another ironic part of the story is that is was never clear to me why MSU was paired with MTSU in the first place. As the top ranked No. 2 seed, based on the s-curve, MSU should have naturally been paired with the weakest No.15 seed. In 2016, MTSU was actually the strongest of the No. 15 seeds, based on the committee's published seed list.
The exact placement of the No. 15 seeds is often driven by geography, and in this case the weakest No. 15 seed was Weber State. However, Weber State was matched against Xavier in the first round and both this game and the MSU-MTSU both took place in the same pod in St. Louis. Xavier was actually ranked as the weakest No. 2 seed. MSU and Xavier, by basic seeding principles, should have traded opponents. If I even meet Hollis in person, I will ask him why this decision was made.
Had MSU made it past MTSU there is no guarantee that would have won the Title or would have even made it past No. 10 Syracuse in the second round. MSU would have projected as a nine-point favorite in that game, but Izzo teams have always struggled with Boeheim's zone. Then again, with so many three-point snipers on the team, I tend to think Valentine and crew would have gotten the job done.
In the Sweet Sixteen, MSU would have next faced No. 11 Gonzaga, where MSU would have been a seven-point favorite. Then, MSU would have once again faced Tony Bennett and No. 1 Virginia. I project the MSU would have been slightly favored in this game too. In reality, UVA blew a 15-point lead over Syracuse with 10 minutes to play. Is there any way that the Cavaliers would have not blown it against MSU as well? I don't think so.
Had MSU made the Final Four in 2016, the path would have been a tough one. In the National Semifinals, they would have face No. 1 North Carolina and then No. 2 Villanova in the Finals. Based on the pre-tournament Kenpom numbers, MSU would have been a narrow favorite in both contests. In reality, it is hard to say, as both UNC and Villanova played very well in March.
While we will never know if MSU would have earned Izzo his second title in 2016 had the Middle Tennessee State debacle not happened, based on my math, this was his most likely chance since at least 2001.
Thus ends the lesson for today. As always, stay safe, stay home, wash your hands, and Go Green.
MEN'S BASKETBALL Dr. G&W Analysis: The Odds of Picking a Perfect NCAA Bracket
- Spartans Illustrated Message Board
- 2 Replies
(Note that this is another article from my archive that I wrote last year. But, the data has not changed and the results are still pretty interesting to the math-nerd crowd)
Even though COVID-19 robbed us all of the 2020 version of March Madness, there are still a lot of interesting questions that we can ask about the Tournament. One particular topic that I thought a lot about last year is the question, "What are the odds to correctly pick the results of the entire tournament?" As we will see, this is actually a much more complex question than it might appear on the surface.
As you might expect, I am not the first person to think about this problem. Warren Buffet perhaps stimulated most of this discussion in 2014 when he started to offer multi-million or even $1 billion dollar rewards for different variations on a perfect bracket, either just up to the Sweet 16 or a full perfect bracket. Largely in response to this publicity, math and stats gurus started to take aim at answering the question as to how likely it was that Buffet would need to pay up.
There is one extremely simple way to think about this problem. If you assume a person is simply randomly guessing on the winner of each game, the odds are very easy to calculate. It is the same as the odds of correctly guessing the result of 63 consecutive coin flips, which is 1 in 2 multiplied by itself 63 times (2^63). In other words:
1 in 9,223,372,036,854,780,000
Those are pretty long odds. (Also note that although there are 68 teams in tournament, most mainstream bracket contests ignore the results of the First Four. With the remaining 64 teams, they play in 63 total games, with one team being eliminated in each game until 1 team remains).
Although a lot of people will reference the number above, it isn't correct. That is because games are clearly not just coin flips with each team having equal probability of victory. The obvious example of this is the set of first round games between No. 1 seeds and No. 16 seeds. If one were to enter a contest where one only needs to pick the winner of these four games, the obvious strategy is to take all the No. 1 seeds. In 40 of the past 41 tournament, this would be the winning strategy. However, in 2018, that would not have been the case, as UMBC upset No. 1 Virginia in historical fashion.
This particular example gives valuable insight into how to think about the problem of the odds of a perfect bracket. What would the odds be to "win" the No. 1 vs No. 16 Seed Challenge? In most years, the odds would be close to 100%, as long as one knew that No. 1 seeds (almost) always win in the first round. But, what about in 2018? Let's say that there was one brave UMBC grad that decided to take a flier on his alma mater. What would his odds have been? My math suggests that this type of upset should occur about one percent of the time (about once every 25 years). So, I think that it is reasonable to say that this UMBC grad had about a one percent chance to win this contest with that bracket.
This example tells us several things:
1) The odds to pick a "perfect bracket" seem to be equivalent to the odds of that bracket occurring
2) Therefore the odds of a perfect bracket are not the same from year to year and can vary widely
3) If you can estimate the probability of victory for each potential match-up, that should allow you to make the appropriate calculation
In my various internet searches related to this topic, most people who have thought about this problem have come to the same basic conclusions. Professors at various Universities have weighed in on the topic. One Professor at Depaul gave the odds at 1 in 128 billion (or perhaps "as low as" that). Another professor at Duke suggested it was 1 in 2.4 trillion.
But, perhaps the best analysis I have seen is from Nate Silver at 538.com, who calculated the odds of a "perfect bracket" in 2014 to be 1 in 7.4 billion, yet only 1 in 1.6 billion in 2015. But, these odds seem rather low in comparison to the odds above. As I looked at 538's data in more detail, it became clear that his definition of the "perfect bracket" is different than mine. He is calculating the odds that the tournament goes as close to chalk as possible, as opposed to how the tournament actually played out. So, the numbers that he quotes are not the odds that you will win the Warren Buffet office pool. Instead, they are the odds of the favored team (based on his math and not on seeds) wins every game. This is, technically, the most likely overall single result, and as such it is the lower bound on the odds to win the Warren Buffet office pool for that year.
Although my initial purpose is different, the method used by 538 appears sound. In order to calculate the odds of most likely ("chalk") bracket or the actual final bracket, one only needs a way to calculate the win probabilities for any arbitrary tournament match-up. Fortunately, I know a good way to do that. I used the pre-tournament efficiency data from Kenpom to generate point spreads, and then used my own formula to calculate the probability that either team would win any potential match-up. When I ran these numbers for the 2019 tournament, I got the following numbers
As you can see, there is quite a bit a variation in the odds from year to year, but in two-thirds of the years, the odds are in the one in tens of billions range. There are only two years with better odds: 2019 (with odds of one in 6.4 billion) and 2015 (with odds of one in 4.4 billion). In the four remaining years (2003, 2006, 2010, and 2012) the odds are in the one in hudreds of billions range.
But, what is the source of this variation from year to year? One idea that I had was based on an observation that I made about the 2019 tournament. Based on the Kenpom efficiency margin data, the 2019 tournament seemed a bit top-heavy. The Top 10 teams or so overall seemed to be all above average in comparison to the historical efficiencies of past No. 1, No. 2, and No. 3 seeds. UVA, Gonzaga, Duke, and MSU were all in the Top 21 all-time in Kenpom efficiency margin going into the tournament. So, I plotted the average efficiency margin data as a function the odds of the most likely bracket coming to pass. As I adjusted the parameters, I found that the best fit to the data occurred when I correlated these odds to the average efficiency margin of the No. 1, No. 2, and No. 3 seeds. That data is shown below. With the exception of two mild outliers (2002 and 2005), the fit is really good.
This bring me to one first conclusion:
NCAA Tournaments that are more top heavy, especially in the Top 10 teams or so, are more predictable.
In terms of raw probability the "most predictable" bracket in the last 20 years was in 2015, which was just a little over 100 times more predictable than the least predictable tournament in 2006. While the difference between odds of 1 in 4 billion and 1 in 500 billion is practically meaningless, this does give us an interesting way to measure the parity in a given bracket in any given year.
At a side note, I should mention that the odds that I am calculating are a bit longer than the odds that Nate Silver / 538 got in his analysis of the 2014 and 2015 seasons as referenced above. Of course, I am using a different model and set of data (Kenpom) as my input, so there is no surprise that the numbers don't match. However, based on the analysis above, it seems likely that the 538 model generally must predict less parity than the Kenpom system, especially at the top. If nothing else, while these numbers are interesting to know (at least for me), this is not an exact science, and we are dealing with relative probabilities that we only in reality know within a factor of ten or so.
While knowing how rare it would be for a tournament to happen when the favored teams wins in 63 straight games, I find that the more interesting question is that of the odds of a truly perfect bracket which would win the full tournament version of the Warren Buffet challenge. When I ran those numbers for the last 18 years, here is the result:
While the best possible odds to pick the winning bracket are in the "one in a tens of billion" range, the odds to correctly predict a real tournament are much, much worse. In fact, 2019 was the "most predictable" tournament with odds at one in over three trillion, which is about 500 times less likely than the most likely / chalk bracket. As for just getting up to the Sweet Sixteen fully correct, this year's odds of one in 540 million are by far the best of record. In all previous years back to 2002, the odds were at least one in a billion, and have been as high as one in 1.8 trillion just last year.
Furthermore, the spread in this data set is clearly larger than in the previous set. In the "most likely / chalk" set, the difference between the most likely bracket and the least likely bracket was a factor of 100. When it comes to the real brackets, this difference is a factor of roughly 50,000. As a general rule, your odds of correctly predicting the result of all 63 NCAA Tournament games is about one in 100 trillion, with a very large variance. The odds can be as high as one in over 100 quadrillion for a year like 2011.
What is the source of this variance? It turns out to be simple statistical variation. Basically: luck. In some tournaments, a lot of big upsets are observed, and in others, there are very few. As anyone who has every participated in an NCAA office pool knows, upsets are unpredictable, and some upsets are way less predictable than other.
In order to explore this final avenue, I used the same Kenpom efficiency and probability data to set up a Monte Carlo simulation of the 2019 tournament. I ran a total of 5000 simulations and then calculated the distribution of the odds of a perfect actual bracket. The distribution of the odds is shown below, where in this case I plot the log-10 value of the odds. What I mean by that is "9" would be equal to one in a billion (which has 9 zeros). "10" is one in 10 billion, "11" is one in 100 billion, etc. In addition, I superimposed the data from actual brackets from 2002 to 2019. Those results are shown below.
As you can see, the actual distribution of odds that I calculated from 2002 to 2019 tournaments matches very well with the simulated distribution for 2019. So, I am confident in my claim that the observed variance is simply due to normal statistical variation. Furthermore, I am also confident that the main contributing factor in huge variance is the occurrence of "big" upsets. This is shown in the plot below, which correlates the number of observed "big" upsets to the ratio of the odds of the real bracket to the most likely bracket for that year. This eliminates the variance that is caused by the relative parity of the participating teams.
In this context, I define "big" as an upset where the odds for it to occur are roughly less than 30 percent. Practically speaking, this translate to a first round game where the difference in seed between the two teams in greater than seven (No. 4 / No. 13 upsets count, but No. 5/ No. 12 upsets do not), a second round game where the seed differential is greater than four, and any upset later in the tournament where the seed differential is greater than two. In 2019, there were only three "big" upsets: UC Irvine over Kansas State, and Auburn over both UNC and Kentucky. That feels right to me.
In general, the correlation above is not superb, but I think it does prove my main point.
As a final note, my simulation of 5,000 different potential outcomes for the 2019 tournament did provide some other interesting observations. As for the extreme cases, the least chaotic bracket still had odds of around one in 30 billion., which is only a factor of five less likely than the most likely result. But, the craziest bracket in my particular simulation had odds of:
1 in 5,202,405,322,734,840,000,000 (or one in five sextillion!)
This value is roughly 500 times less likely than even the "random coin flip bracket" mentioned at the very top of this article. I think that might be a fun tournament to see play out, but we might need to wait 5,000 years to see it.
Taking all of this data together, it finally occurred to me that there is perhaps a simpler mathematical way to look at this problem. Whether I am calculating the odds of the most likely / chalk bracket or the actual bracket, the way to calculate it is to multiple together the individual probabilities corresponding to the (real or projected) outcome of each contest. In the case of a big upset (like UMBC over Virgina) that probability is very small (about one percent), in the case of a "normal" No. 1 / No. 16 match-up that probability is close to one (~99 percent), and in an No. 8 / No. 9 seed match-up, that probability is close to 50 percent.
If we consider the entire set of 63 games (or 48, if we are just interested in the results of the 1st weekend), there is a fixed probability that is representative of the tournament as a whole. Mathematically, this works out to be the geometric average of the 63 (or 48 or whatever) individual probabilities. If we know (or calculate) this average probability (let's call it 'P'), the odds of that specific bracket occurring is then simply:
P^63
If this looks familiar, it should because it is essentially the same equation that we started with in the example of the coin flipping experiment. As it turns out, we actually can model the tournament in that way. The only trick is that the "coin" is a weighted one. In this example, a top-loaded tournament with less parity tends to drive this probability up. A tournament with few upsets effectively does the same thing. In contrast, a lot of upsets drives this effective probability down.
In real tournaments, this effective overall probability is in the range of 53-63%. In the most likely / chalk brackets, the effective probability ranges from 65-70%. In both cases, the odds are ridiculously sensitive to small changes. Changing the the average probability by just one percent will change the overall odds by a factor of two to three. This explains why small differences in the probability models can have such a large affect on the overall odds. The graph below shows this power law correlation to the odds.
Well, I think that I finally found the bottom of this particular stats rabbit hole and so that is all for now. For now, stay at home, stay safe and Go Green.
Even though COVID-19 robbed us all of the 2020 version of March Madness, there are still a lot of interesting questions that we can ask about the Tournament. One particular topic that I thought a lot about last year is the question, "What are the odds to correctly pick the results of the entire tournament?" As we will see, this is actually a much more complex question than it might appear on the surface.
As you might expect, I am not the first person to think about this problem. Warren Buffet perhaps stimulated most of this discussion in 2014 when he started to offer multi-million or even $1 billion dollar rewards for different variations on a perfect bracket, either just up to the Sweet 16 or a full perfect bracket. Largely in response to this publicity, math and stats gurus started to take aim at answering the question as to how likely it was that Buffet would need to pay up.
There is one extremely simple way to think about this problem. If you assume a person is simply randomly guessing on the winner of each game, the odds are very easy to calculate. It is the same as the odds of correctly guessing the result of 63 consecutive coin flips, which is 1 in 2 multiplied by itself 63 times (2^63). In other words:
1 in 9,223,372,036,854,780,000
Those are pretty long odds. (Also note that although there are 68 teams in tournament, most mainstream bracket contests ignore the results of the First Four. With the remaining 64 teams, they play in 63 total games, with one team being eliminated in each game until 1 team remains).
Although a lot of people will reference the number above, it isn't correct. That is because games are clearly not just coin flips with each team having equal probability of victory. The obvious example of this is the set of first round games between No. 1 seeds and No. 16 seeds. If one were to enter a contest where one only needs to pick the winner of these four games, the obvious strategy is to take all the No. 1 seeds. In 40 of the past 41 tournament, this would be the winning strategy. However, in 2018, that would not have been the case, as UMBC upset No. 1 Virginia in historical fashion.
This particular example gives valuable insight into how to think about the problem of the odds of a perfect bracket. What would the odds be to "win" the No. 1 vs No. 16 Seed Challenge? In most years, the odds would be close to 100%, as long as one knew that No. 1 seeds (almost) always win in the first round. But, what about in 2018? Let's say that there was one brave UMBC grad that decided to take a flier on his alma mater. What would his odds have been? My math suggests that this type of upset should occur about one percent of the time (about once every 25 years). So, I think that it is reasonable to say that this UMBC grad had about a one percent chance to win this contest with that bracket.
This example tells us several things:
1) The odds to pick a "perfect bracket" seem to be equivalent to the odds of that bracket occurring
2) Therefore the odds of a perfect bracket are not the same from year to year and can vary widely
3) If you can estimate the probability of victory for each potential match-up, that should allow you to make the appropriate calculation
In my various internet searches related to this topic, most people who have thought about this problem have come to the same basic conclusions. Professors at various Universities have weighed in on the topic. One Professor at Depaul gave the odds at 1 in 128 billion (or perhaps "as low as" that). Another professor at Duke suggested it was 1 in 2.4 trillion.
But, perhaps the best analysis I have seen is from Nate Silver at 538.com, who calculated the odds of a "perfect bracket" in 2014 to be 1 in 7.4 billion, yet only 1 in 1.6 billion in 2015. But, these odds seem rather low in comparison to the odds above. As I looked at 538's data in more detail, it became clear that his definition of the "perfect bracket" is different than mine. He is calculating the odds that the tournament goes as close to chalk as possible, as opposed to how the tournament actually played out. So, the numbers that he quotes are not the odds that you will win the Warren Buffet office pool. Instead, they are the odds of the favored team (based on his math and not on seeds) wins every game. This is, technically, the most likely overall single result, and as such it is the lower bound on the odds to win the Warren Buffet office pool for that year.
Although my initial purpose is different, the method used by 538 appears sound. In order to calculate the odds of most likely ("chalk") bracket or the actual final bracket, one only needs a way to calculate the win probabilities for any arbitrary tournament match-up. Fortunately, I know a good way to do that. I used the pre-tournament efficiency data from Kenpom to generate point spreads, and then used my own formula to calculate the probability that either team would win any potential match-up. When I ran these numbers for the 2019 tournament, I got the following numbers
- Correct bracket up to the Sweet 16 (48 games): 1 in 540 million
- Odds of the most likely / chalk bracket: 1 in 6.4 billion
- Odds if the correct full bracket (all 63 games): 1 in 3.2 trillion
As you can see, there is quite a bit a variation in the odds from year to year, but in two-thirds of the years, the odds are in the one in tens of billions range. There are only two years with better odds: 2019 (with odds of one in 6.4 billion) and 2015 (with odds of one in 4.4 billion). In the four remaining years (2003, 2006, 2010, and 2012) the odds are in the one in hudreds of billions range.
But, what is the source of this variation from year to year? One idea that I had was based on an observation that I made about the 2019 tournament. Based on the Kenpom efficiency margin data, the 2019 tournament seemed a bit top-heavy. The Top 10 teams or so overall seemed to be all above average in comparison to the historical efficiencies of past No. 1, No. 2, and No. 3 seeds. UVA, Gonzaga, Duke, and MSU were all in the Top 21 all-time in Kenpom efficiency margin going into the tournament. So, I plotted the average efficiency margin data as a function the odds of the most likely bracket coming to pass. As I adjusted the parameters, I found that the best fit to the data occurred when I correlated these odds to the average efficiency margin of the No. 1, No. 2, and No. 3 seeds. That data is shown below. With the exception of two mild outliers (2002 and 2005), the fit is really good.
This bring me to one first conclusion:
NCAA Tournaments that are more top heavy, especially in the Top 10 teams or so, are more predictable.
In terms of raw probability the "most predictable" bracket in the last 20 years was in 2015, which was just a little over 100 times more predictable than the least predictable tournament in 2006. While the difference between odds of 1 in 4 billion and 1 in 500 billion is practically meaningless, this does give us an interesting way to measure the parity in a given bracket in any given year.
At a side note, I should mention that the odds that I am calculating are a bit longer than the odds that Nate Silver / 538 got in his analysis of the 2014 and 2015 seasons as referenced above. Of course, I am using a different model and set of data (Kenpom) as my input, so there is no surprise that the numbers don't match. However, based on the analysis above, it seems likely that the 538 model generally must predict less parity than the Kenpom system, especially at the top. If nothing else, while these numbers are interesting to know (at least for me), this is not an exact science, and we are dealing with relative probabilities that we only in reality know within a factor of ten or so.
While knowing how rare it would be for a tournament to happen when the favored teams wins in 63 straight games, I find that the more interesting question is that of the odds of a truly perfect bracket which would win the full tournament version of the Warren Buffet challenge. When I ran those numbers for the last 18 years, here is the result:
While the best possible odds to pick the winning bracket are in the "one in a tens of billion" range, the odds to correctly predict a real tournament are much, much worse. In fact, 2019 was the "most predictable" tournament with odds at one in over three trillion, which is about 500 times less likely than the most likely / chalk bracket. As for just getting up to the Sweet Sixteen fully correct, this year's odds of one in 540 million are by far the best of record. In all previous years back to 2002, the odds were at least one in a billion, and have been as high as one in 1.8 trillion just last year.
Furthermore, the spread in this data set is clearly larger than in the previous set. In the "most likely / chalk" set, the difference between the most likely bracket and the least likely bracket was a factor of 100. When it comes to the real brackets, this difference is a factor of roughly 50,000. As a general rule, your odds of correctly predicting the result of all 63 NCAA Tournament games is about one in 100 trillion, with a very large variance. The odds can be as high as one in over 100 quadrillion for a year like 2011.
What is the source of this variance? It turns out to be simple statistical variation. Basically: luck. In some tournaments, a lot of big upsets are observed, and in others, there are very few. As anyone who has every participated in an NCAA office pool knows, upsets are unpredictable, and some upsets are way less predictable than other.
In order to explore this final avenue, I used the same Kenpom efficiency and probability data to set up a Monte Carlo simulation of the 2019 tournament. I ran a total of 5000 simulations and then calculated the distribution of the odds of a perfect actual bracket. The distribution of the odds is shown below, where in this case I plot the log-10 value of the odds. What I mean by that is "9" would be equal to one in a billion (which has 9 zeros). "10" is one in 10 billion, "11" is one in 100 billion, etc. In addition, I superimposed the data from actual brackets from 2002 to 2019. Those results are shown below.
As you can see, the actual distribution of odds that I calculated from 2002 to 2019 tournaments matches very well with the simulated distribution for 2019. So, I am confident in my claim that the observed variance is simply due to normal statistical variation. Furthermore, I am also confident that the main contributing factor in huge variance is the occurrence of "big" upsets. This is shown in the plot below, which correlates the number of observed "big" upsets to the ratio of the odds of the real bracket to the most likely bracket for that year. This eliminates the variance that is caused by the relative parity of the participating teams.
In this context, I define "big" as an upset where the odds for it to occur are roughly less than 30 percent. Practically speaking, this translate to a first round game where the difference in seed between the two teams in greater than seven (No. 4 / No. 13 upsets count, but No. 5/ No. 12 upsets do not), a second round game where the seed differential is greater than four, and any upset later in the tournament where the seed differential is greater than two. In 2019, there were only three "big" upsets: UC Irvine over Kansas State, and Auburn over both UNC and Kentucky. That feels right to me.
In general, the correlation above is not superb, but I think it does prove my main point.
As a final note, my simulation of 5,000 different potential outcomes for the 2019 tournament did provide some other interesting observations. As for the extreme cases, the least chaotic bracket still had odds of around one in 30 billion., which is only a factor of five less likely than the most likely result. But, the craziest bracket in my particular simulation had odds of:
1 in 5,202,405,322,734,840,000,000 (or one in five sextillion!)
This value is roughly 500 times less likely than even the "random coin flip bracket" mentioned at the very top of this article. I think that might be a fun tournament to see play out, but we might need to wait 5,000 years to see it.
Taking all of this data together, it finally occurred to me that there is perhaps a simpler mathematical way to look at this problem. Whether I am calculating the odds of the most likely / chalk bracket or the actual bracket, the way to calculate it is to multiple together the individual probabilities corresponding to the (real or projected) outcome of each contest. In the case of a big upset (like UMBC over Virgina) that probability is very small (about one percent), in the case of a "normal" No. 1 / No. 16 match-up that probability is close to one (~99 percent), and in an No. 8 / No. 9 seed match-up, that probability is close to 50 percent.
If we consider the entire set of 63 games (or 48, if we are just interested in the results of the 1st weekend), there is a fixed probability that is representative of the tournament as a whole. Mathematically, this works out to be the geometric average of the 63 (or 48 or whatever) individual probabilities. If we know (or calculate) this average probability (let's call it 'P'), the odds of that specific bracket occurring is then simply:
P^63
If this looks familiar, it should because it is essentially the same equation that we started with in the example of the coin flipping experiment. As it turns out, we actually can model the tournament in that way. The only trick is that the "coin" is a weighted one. In this example, a top-loaded tournament with less parity tends to drive this probability up. A tournament with few upsets effectively does the same thing. In contrast, a lot of upsets drives this effective probability down.
In real tournaments, this effective overall probability is in the range of 53-63%. In the most likely / chalk brackets, the effective probability ranges from 65-70%. In both cases, the odds are ridiculously sensitive to small changes. Changing the the average probability by just one percent will change the overall odds by a factor of two to three. This explains why small differences in the probability models can have such a large affect on the overall odds. The graph below shows this power law correlation to the odds.
Well, I think that I finally found the bottom of this particular stats rabbit hole and so that is all for now. For now, stay at home, stay safe and Go Green.
RECRUITING Kaiden Salter doing virtual visit
- By Jason Killop
- Spartans Illustrated Message Board
- 90 Replies
RECRUITING Offers at Hough HS, North Carolina
- By Jason Killop
- Spartans Illustrated Message Board
- 25 Replies
Deep Dive on Els & MSU Special Teams
- Spartans Illustrated Message Board
- 0 Replies
Spring Football: Deep dive on Els and special teams
Jim Comparoni • SpartanMag
Publisher
@JimComparoni
Fans are wondering whether there will be a football season in the fall, but Michigan State coaches are doing what football coaches are trained to do - coach like there IS a tomorrow.
SpartanMag’s continuing series on Michigan State football for the spring of 2020:
EAST LANSING - It’s probably easy for a casual Spartan football fans to overlook one guy when counting up the number of coaches on the Michigan State coaching staff with defensive coordinator experience.
We know about Mike Tressel and Harlon Barnett. They’ve been the d-coordinator or co-coordinators at Michigan State in recent years.
Scottie Hazelton is the new defensive coordinator, coming to Michigan State from Kansas State. And of course, head coach Mel Tucker was defensive coordinator at Georgia (2016-18), at three stops in the NFL (Bears, Jaquars, Browns) and was co-defensive coordinator at Ohio State (2004).
Don’t overlook the fact that new Michigan State linebackers coach Ross Els served as defensive coordinator at Purdue for one year in 2016 under Darrell Hazell. Purdue went 3-9 that year, and Hazell was fired after four unsuccessful years. But Els’ experience as a Big Ten defensive coordinator, combined with four years as an assistant at Nebraska, helps beef up the Spartan defensive staff. He’s been a respected assistant in the Big Ten before.
“I just know what the Big Ten delivers - physical, smart football,” Els said. “That’s what we play in this league and I’m glad to be back in it. I really am.”
Els also has head coaching experience, with tremendous success. At the age of 32 in 1997, after assistant stints at Northern Iowa and Nebraska-Omaha, he became head coach of Hastings College in Hastings, Neb.
Els went 32-9 in four years at Hastings, including 10-1 and 11-1 seasons in 1998 and ’99, helping Hastings advance to the NAIA Playoffs.
SpartanMag asked Els if he ever had an itch to get back into head coaching.
“Gawd no,” he said. “Do you know what these guys go through all day long? They’re putting out fires all day long. I want to coach ball, you know?”
But seriously.
“It’ll happen again, I think,” he said. “But I’m really, really happy with where I am right now, to be able to work with a guy like Mel and Scottie Hazelton. Scottie’s been phenomenal. Scottie’s an incredible defensive coordinator and has had a lot of success, and he leaves a lot of things open as to, ‘How do you guys want to do this? How do you want to do that?’
“I’m in such a good position right now that I’m not looking to go back to that head coaching gig right now. I would rather coach football and recruit than deal with administrators and the like. I almost said media, but I didn’t.”
NEW SPECIAL TEAMS DIRECTION
Ross Els has been special teams coordinator at Nebraska and Colorado, and defensive coordinator at Purdue.
Els’ coordinator experience includes being special teams coordinator at Nebraska under Bo Pelini from 2012-14, and last year at Colorado. He now holds the special teams coordinator tag at Michigan State.
Els indicated that special teams coaching will be delegated throughout the staff, as has been the case at Michigan State in recent years, and is commonplace throughout college football.
“I’m not ready to say who’s going to do what unit,” Els said. “We are getting through the introductory phase with our players right now. The coaches have bought in. They have been great in meetings.”
Tressel oversaw special teams throughout the Dantonio era.
“This is a veteran staff and they’re all excited about (special) teams,” Els said. “(There’s) a lot of advice because a lot of guys have done a lot of special teams. As far as divvying up responsibilities, we will be doing that.”
Els will have access to the full roster for special teams roles.
“One of the great things about working for a guy like Mel Tucker is he understands the importance of special teams and working with him for the year I did at Colorado, he never said, ‘Don’t use this guy,’” Els said. “Now we’re going to use common sense on offense, defense and special teams to make sure they’re not over-played but that doesn’t mean that special teams will take the back seat.
“Everybody’s going to be available. We get the best that are available. We made that very clear to the team, both the head coach and myself. It gives them an opportunity to continue to do some of the skills that they do at their position. If they are a wide receiver that blocks, they are going to block on kickoff return. If they are a running back that runs, they are going to run the football. It also helps their resumé because everybody in the NFL if you’re a skill position are going to play on special teams.”
Five starting defensive players played on the punt coverage team in the New Era Pinstripe Bowl: Safety Xavier Henderson, linebacker Noah Harvey, linebacker Antjuan Simmons, linebacker Tyriq Thompson and safety David Dowell.
Thompson and Dowell have graduated.
Henderson, Harvey and Simmons will be back in 2020, as will sophomore long snapper Jude Pedrozo.
“As far as the personnel that we have right now, I’m excited because whenever you have good defense you usually have good special teams,” Els said. “Michigan State is known for their good defense so we should be able to be good in the special team area.”
INSIDE THE ROSTER
Davion Williams was a fixture on special teams in 2019, and is due to make a run at regular playing time on defense in 2020.
Davion Williams and Dominique Long were the gunners for the punt coverage team last year and will be back in 2020, each with a chance to compete for a starting job at cornerback on the defense.
Long, Williams and other reserve defensive players such as Chase Kline, Marcel Lewis and Jack Mandryk played on the kickoff coverage team in the New Era Pinstripe Bowl.
Linebacker Jeslord Boateng and defensive end Drew Beesley played on the punt coverage team last year.
Kline, Mandryk, tight end Trenton Gillison, tight end Parks Gissinger, wide receiver Jahz Watts and back-up punter Tyler Hunt played on the kickoff return team, with Jalen Nailor as the deep return man.
Nailor is likely a leading candidate to be a return specialist in 2020. He played in only four games last year - the first two and last two of the season. Michigan State was 4-0 in games in which Nailor played.
Nailor returned two punts in the season opener against Tulsa for a total of 17 yards. He returned two kickoffs, including a 28-yarder, but then was lost until mid-November.
INSIDE MSU’s RETURN STATS
Michigan State ranked No. 10 in the Big Ten in kickoff return average and ninth in punt return average at 5.4 yards per. Interestingly, Ohio State was 10th, Cinderella story Minnesota was last at 1.3 on only seven punt return attempts on the year).
Iowa was No. 1 in punt return average at 9.1, but the Hawkeyes have been known over the years for selling out to set-up for returns, occasionally becoming susceptible to fakes.
Michigan ranked second in punt return average at 8.9 and easily had the most return attempts (26) and return yardage (232). No other Big Ten team had more than 140 yards on punt returns.
Michigan has been successfully aggressive in its punt return game in recent years. Michigan State has tried to make Michigan pay by faking punts against the Wolverines, often with good results. But Michigan almost always comes out ahead in the punt exchange game against its opponents.
Michigan State was conservative in the return game for most of the Mark Dantonio era. Michigan State didn’t have a kickoff return of more than 31 yards in 2019. After Nailor was lost to injury, Michigan State customarily called for fair catches on kickoff returns. Darrell Stewart led the team in kickoff return attempts with just eight.
Losing Nailor to a lower body injury after the first game of the 2019 season probably had something to do with Michigan State buckling things down in the return game.
In the punt return game, Michigan State often played “punt safe” when the opponent faced four down. Michigan State put high priority on getting the ball back. That meant avoiding a roughing the punter penalty, and not leaving themselves susceptible to a fake.
Even when Wake Forest was in fourth-and-eight midway through the first half, Michigan State played it safe. Michigan State kept 10 defensive regulars on the field. They played at a careful depth and made sure a punt was attempted, rather than sending athletes to go for a block, or retreating to set up for a return. Deep man Brandon Sowards didn’t have much room for a return. But Michigan State got the ball back.
However, Spartan punt returners in recent years have struggled at fielding punts on the fly, resulting in negative hidden yardage after the ball hits the ground and rolls.
With Els, MSU’s special teams philosophy might change from week to week, rather than a season-long mode of operation.
“As far as a general philosophy, (we are) no different than anyone else - we want to run and hit,” Els said. “We want to be aggressive as we possibly can. We are not a passive special teams unit. There aren’t many like that, but there are some. We want to keep attacking.”
But some of that philosophy includes at least a portion of the Tressel/Dantonio mindset.
“The most important thing about a punt return is getting the ball back to the offense and not creating something stupid,” Els said. “When I hear that some program is great at punt returns and all that, or they block all these punts, but they also had three roughing-the-punter penalties. I’m not saying one philosophy is better than the other.”
ANY HAPPY RETURNS?
When healthy, Jalen Nailor was a dangerous return man in 2019.
During a teleconference last week, Els was asked twice about MSU’s inability to generate touchdowns in the return game. R.J. Shelton’s game-opening 90-yard kickoff return against Penn State in 2014 marks the last time a Spartan returned a kickoff for a TD. Shelton’s TD was the first since Keshawn Martin did it against Minnesota in 2009.
Is that drought a big deal? Consider this: There were only five kickoffs returned for touchdowns in conference games in the entire Big Ten last year, two by Maryland.
There were no punt returns for touchdowns in Big Ten conference games last year.
“I’m hearing some negativity toward Michigan State and I have not watched any of their film so I do not know what they have done well or what they haven’t,” Els said. “That’s not my job.
“I can tell you this, though, without telling you exactly how we’re going to do things is: We will be aggressive. We will be aggressive in return game, we will be aggressive in coverage games. What you decide to do, to me, is dictated by the opponent. If they are a bad protection team, we’ll come after them. If they are a bad coverage team, we probably won’t. If they’re very sound, we’ll mix it up.
“Our kids will have the opportunity to do both - set up returns, go after punts. It just depends on what the opponent gives us, and how good are we at holding people up? How good are we at returning? How good are we at bursting off the line of scrimmage? How good are we at blocking punts? So it’s going to be awhile before I know the answers of us, and then we can go to the opponent and find out what they do poorly.”
* Michigan State was No. 12 in the Big Ten in net punting average in 2019 (36.0). Michigan State was last in punt return average allowed at 3.6 per punt.
* Michigan State was No. 10 in kickoff coverage, allowing a net average of 38.3 yards per return. Illinois, Iowa and Wisconsin were the best in the big Ten. Maryland, Rutgers an Nebraska were the worst.
* Kicker Matt Coghlin wasn’t in top form in 2019. He was 22-of-32, for just 68.8 percent, ranking No. 10 in the Big Ten in field goal percentage. However, his 22 successful kicks are tied for second most in single-season school history.
Coghlin bounced back from a mid-season slump to hit four field goals in a 19-16 victory over Maryland, including the game-winning 33-yarder with 2:10 to play. He has three game-winning field goals in his career (including 2017 against Penn State and 2019 against Indiana). However, he missed three field goals in a disturbing 10-7 loss to Arizona State.
“I’m assuming Matt will come back and be ready to do a great job with our kicking,” Els said. “Matt had a great year two years ago and this year, at least on paper, not as good. But we have a lot of confidence in Matt. He’s going to line up there, and he’s going to win some games for us. Games in this league are going to come down to the wire.
“He’s a leader in that room, too. That’s what I like about Matt. When I first met him, he said, ‘Coach, put this room on my shoulders. I’ll take care of these guys and get these guys ready to go.’ So we’re excited with Matt’s future.
“We do have some specialist positions to fill. That’s one of my biggest concerns right now. I don’t know who our punt returner is going to be, I don’t know who our punter is going to be.”
With Jake Hartbarger having graduated, Australian Jack Bouwmeester was in line to take the job as a redshirt freshman in 2020. But he abruptly left school during spring semester. Michigan State expects to activate Match Crawford, a graduated transfer from UTEP. Because that transfer is not yet official, Michigan State coaches aren’t able to comment on him.
“(I) don’t know who’s going to be the guy lining up there (at punter),” Els said. “(I’ve) never seen a punt (here). I don’t know that Jack (Bouwmeester) would have been the guy. I have no idea. I guarantee you this: If we have to punt, we will have somebody back there who can kick the ball. I just don’t know who it is yet.”
Kickoff specialists Evan Morris and Cole Hahn will be back. Morris played in two games and had five touchbacks on 11 kickoffs. However, he didn’t see action after the Arizona State game. Hahn kicked for the rest of the year and had nine touchbacks in 52 kickoffs.
More SpartanMag Spring Football Coverage:
* Els keeping some things quiet at LB
* Gilmore ramping up the tight ends.
* Hawkins bullish on young, hungry wide receiver group
* Coach Kap sees potential in o-line he is inheriting
* Coaching at Michigan State gives Hawkins goosebumps
* Antjuan Simmons moving it forward, keeping it normal
* Barnett officially moves to CB; shake-up taking place at punter
* Spartans add former UTEP punter as grad transfer
* Tucker optimistic during Detroit TV appearance
* Allen getting after it with coaching from big brother
* New staff inheriting a good leader in Antjuan Simmons
* What we learned about o-coordinator Jay Johnson
* Tucker drawing on NFL experience during practice stoppage
* For Hazelton, ’It’s all ball’
* New o-coordinator Jay Johnson ‘looking for a CEO’ at QB
Jim Comparoni • SpartanMag
Publisher
@JimComparoni
Fans are wondering whether there will be a football season in the fall, but Michigan State coaches are doing what football coaches are trained to do - coach like there IS a tomorrow.
SpartanMag’s continuing series on Michigan State football for the spring of 2020:
EAST LANSING - It’s probably easy for a casual Spartan football fans to overlook one guy when counting up the number of coaches on the Michigan State coaching staff with defensive coordinator experience.
We know about Mike Tressel and Harlon Barnett. They’ve been the d-coordinator or co-coordinators at Michigan State in recent years.
Scottie Hazelton is the new defensive coordinator, coming to Michigan State from Kansas State. And of course, head coach Mel Tucker was defensive coordinator at Georgia (2016-18), at three stops in the NFL (Bears, Jaquars, Browns) and was co-defensive coordinator at Ohio State (2004).
Don’t overlook the fact that new Michigan State linebackers coach Ross Els served as defensive coordinator at Purdue for one year in 2016 under Darrell Hazell. Purdue went 3-9 that year, and Hazell was fired after four unsuccessful years. But Els’ experience as a Big Ten defensive coordinator, combined with four years as an assistant at Nebraska, helps beef up the Spartan defensive staff. He’s been a respected assistant in the Big Ten before.
“I just know what the Big Ten delivers - physical, smart football,” Els said. “That’s what we play in this league and I’m glad to be back in it. I really am.”
Els also has head coaching experience, with tremendous success. At the age of 32 in 1997, after assistant stints at Northern Iowa and Nebraska-Omaha, he became head coach of Hastings College in Hastings, Neb.
Els went 32-9 in four years at Hastings, including 10-1 and 11-1 seasons in 1998 and ’99, helping Hastings advance to the NAIA Playoffs.
SpartanMag asked Els if he ever had an itch to get back into head coaching.
“Gawd no,” he said. “Do you know what these guys go through all day long? They’re putting out fires all day long. I want to coach ball, you know?”
But seriously.
“It’ll happen again, I think,” he said. “But I’m really, really happy with where I am right now, to be able to work with a guy like Mel and Scottie Hazelton. Scottie’s been phenomenal. Scottie’s an incredible defensive coordinator and has had a lot of success, and he leaves a lot of things open as to, ‘How do you guys want to do this? How do you want to do that?’
“I’m in such a good position right now that I’m not looking to go back to that head coaching gig right now. I would rather coach football and recruit than deal with administrators and the like. I almost said media, but I didn’t.”
NEW SPECIAL TEAMS DIRECTION
Ross Els has been special teams coordinator at Nebraska and Colorado, and defensive coordinator at Purdue.
Els’ coordinator experience includes being special teams coordinator at Nebraska under Bo Pelini from 2012-14, and last year at Colorado. He now holds the special teams coordinator tag at Michigan State.
Els indicated that special teams coaching will be delegated throughout the staff, as has been the case at Michigan State in recent years, and is commonplace throughout college football.
“I’m not ready to say who’s going to do what unit,” Els said. “We are getting through the introductory phase with our players right now. The coaches have bought in. They have been great in meetings.”
Tressel oversaw special teams throughout the Dantonio era.
“This is a veteran staff and they’re all excited about (special) teams,” Els said. “(There’s) a lot of advice because a lot of guys have done a lot of special teams. As far as divvying up responsibilities, we will be doing that.”
Els will have access to the full roster for special teams roles.
“One of the great things about working for a guy like Mel Tucker is he understands the importance of special teams and working with him for the year I did at Colorado, he never said, ‘Don’t use this guy,’” Els said. “Now we’re going to use common sense on offense, defense and special teams to make sure they’re not over-played but that doesn’t mean that special teams will take the back seat.
“Everybody’s going to be available. We get the best that are available. We made that very clear to the team, both the head coach and myself. It gives them an opportunity to continue to do some of the skills that they do at their position. If they are a wide receiver that blocks, they are going to block on kickoff return. If they are a running back that runs, they are going to run the football. It also helps their resumé because everybody in the NFL if you’re a skill position are going to play on special teams.”
Five starting defensive players played on the punt coverage team in the New Era Pinstripe Bowl: Safety Xavier Henderson, linebacker Noah Harvey, linebacker Antjuan Simmons, linebacker Tyriq Thompson and safety David Dowell.
Thompson and Dowell have graduated.
Henderson, Harvey and Simmons will be back in 2020, as will sophomore long snapper Jude Pedrozo.
“As far as the personnel that we have right now, I’m excited because whenever you have good defense you usually have good special teams,” Els said. “Michigan State is known for their good defense so we should be able to be good in the special team area.”
INSIDE THE ROSTER
Davion Williams was a fixture on special teams in 2019, and is due to make a run at regular playing time on defense in 2020.
Davion Williams and Dominique Long were the gunners for the punt coverage team last year and will be back in 2020, each with a chance to compete for a starting job at cornerback on the defense.
Long, Williams and other reserve defensive players such as Chase Kline, Marcel Lewis and Jack Mandryk played on the kickoff coverage team in the New Era Pinstripe Bowl.
Linebacker Jeslord Boateng and defensive end Drew Beesley played on the punt coverage team last year.
Kline, Mandryk, tight end Trenton Gillison, tight end Parks Gissinger, wide receiver Jahz Watts and back-up punter Tyler Hunt played on the kickoff return team, with Jalen Nailor as the deep return man.
Nailor is likely a leading candidate to be a return specialist in 2020. He played in only four games last year - the first two and last two of the season. Michigan State was 4-0 in games in which Nailor played.
Nailor returned two punts in the season opener against Tulsa for a total of 17 yards. He returned two kickoffs, including a 28-yarder, but then was lost until mid-November.
INSIDE MSU’s RETURN STATS
Michigan State ranked No. 10 in the Big Ten in kickoff return average and ninth in punt return average at 5.4 yards per. Interestingly, Ohio State was 10th, Cinderella story Minnesota was last at 1.3 on only seven punt return attempts on the year).
Iowa was No. 1 in punt return average at 9.1, but the Hawkeyes have been known over the years for selling out to set-up for returns, occasionally becoming susceptible to fakes.
Michigan ranked second in punt return average at 8.9 and easily had the most return attempts (26) and return yardage (232). No other Big Ten team had more than 140 yards on punt returns.
Michigan has been successfully aggressive in its punt return game in recent years. Michigan State has tried to make Michigan pay by faking punts against the Wolverines, often with good results. But Michigan almost always comes out ahead in the punt exchange game against its opponents.
Michigan State was conservative in the return game for most of the Mark Dantonio era. Michigan State didn’t have a kickoff return of more than 31 yards in 2019. After Nailor was lost to injury, Michigan State customarily called for fair catches on kickoff returns. Darrell Stewart led the team in kickoff return attempts with just eight.
Losing Nailor to a lower body injury after the first game of the 2019 season probably had something to do with Michigan State buckling things down in the return game.
In the punt return game, Michigan State often played “punt safe” when the opponent faced four down. Michigan State put high priority on getting the ball back. That meant avoiding a roughing the punter penalty, and not leaving themselves susceptible to a fake.
Even when Wake Forest was in fourth-and-eight midway through the first half, Michigan State played it safe. Michigan State kept 10 defensive regulars on the field. They played at a careful depth and made sure a punt was attempted, rather than sending athletes to go for a block, or retreating to set up for a return. Deep man Brandon Sowards didn’t have much room for a return. But Michigan State got the ball back.
However, Spartan punt returners in recent years have struggled at fielding punts on the fly, resulting in negative hidden yardage after the ball hits the ground and rolls.
With Els, MSU’s special teams philosophy might change from week to week, rather than a season-long mode of operation.
“As far as a general philosophy, (we are) no different than anyone else - we want to run and hit,” Els said. “We want to be aggressive as we possibly can. We are not a passive special teams unit. There aren’t many like that, but there are some. We want to keep attacking.”
But some of that philosophy includes at least a portion of the Tressel/Dantonio mindset.
“The most important thing about a punt return is getting the ball back to the offense and not creating something stupid,” Els said. “When I hear that some program is great at punt returns and all that, or they block all these punts, but they also had three roughing-the-punter penalties. I’m not saying one philosophy is better than the other.”
ANY HAPPY RETURNS?
When healthy, Jalen Nailor was a dangerous return man in 2019.
During a teleconference last week, Els was asked twice about MSU’s inability to generate touchdowns in the return game. R.J. Shelton’s game-opening 90-yard kickoff return against Penn State in 2014 marks the last time a Spartan returned a kickoff for a TD. Shelton’s TD was the first since Keshawn Martin did it against Minnesota in 2009.
Is that drought a big deal? Consider this: There were only five kickoffs returned for touchdowns in conference games in the entire Big Ten last year, two by Maryland.
There were no punt returns for touchdowns in Big Ten conference games last year.
“I’m hearing some negativity toward Michigan State and I have not watched any of their film so I do not know what they have done well or what they haven’t,” Els said. “That’s not my job.
“I can tell you this, though, without telling you exactly how we’re going to do things is: We will be aggressive. We will be aggressive in return game, we will be aggressive in coverage games. What you decide to do, to me, is dictated by the opponent. If they are a bad protection team, we’ll come after them. If they are a bad coverage team, we probably won’t. If they’re very sound, we’ll mix it up.
“Our kids will have the opportunity to do both - set up returns, go after punts. It just depends on what the opponent gives us, and how good are we at holding people up? How good are we at returning? How good are we at bursting off the line of scrimmage? How good are we at blocking punts? So it’s going to be awhile before I know the answers of us, and then we can go to the opponent and find out what they do poorly.”
* Michigan State was No. 12 in the Big Ten in net punting average in 2019 (36.0). Michigan State was last in punt return average allowed at 3.6 per punt.
* Michigan State was No. 10 in kickoff coverage, allowing a net average of 38.3 yards per return. Illinois, Iowa and Wisconsin were the best in the big Ten. Maryland, Rutgers an Nebraska were the worst.
* Kicker Matt Coghlin wasn’t in top form in 2019. He was 22-of-32, for just 68.8 percent, ranking No. 10 in the Big Ten in field goal percentage. However, his 22 successful kicks are tied for second most in single-season school history.
Coghlin bounced back from a mid-season slump to hit four field goals in a 19-16 victory over Maryland, including the game-winning 33-yarder with 2:10 to play. He has three game-winning field goals in his career (including 2017 against Penn State and 2019 against Indiana). However, he missed three field goals in a disturbing 10-7 loss to Arizona State.
“I’m assuming Matt will come back and be ready to do a great job with our kicking,” Els said. “Matt had a great year two years ago and this year, at least on paper, not as good. But we have a lot of confidence in Matt. He’s going to line up there, and he’s going to win some games for us. Games in this league are going to come down to the wire.
“He’s a leader in that room, too. That’s what I like about Matt. When I first met him, he said, ‘Coach, put this room on my shoulders. I’ll take care of these guys and get these guys ready to go.’ So we’re excited with Matt’s future.
“We do have some specialist positions to fill. That’s one of my biggest concerns right now. I don’t know who our punt returner is going to be, I don’t know who our punter is going to be.”
With Jake Hartbarger having graduated, Australian Jack Bouwmeester was in line to take the job as a redshirt freshman in 2020. But he abruptly left school during spring semester. Michigan State expects to activate Match Crawford, a graduated transfer from UTEP. Because that transfer is not yet official, Michigan State coaches aren’t able to comment on him.
“(I) don’t know who’s going to be the guy lining up there (at punter),” Els said. “(I’ve) never seen a punt (here). I don’t know that Jack (Bouwmeester) would have been the guy. I have no idea. I guarantee you this: If we have to punt, we will have somebody back there who can kick the ball. I just don’t know who it is yet.”
Kickoff specialists Evan Morris and Cole Hahn will be back. Morris played in two games and had five touchbacks on 11 kickoffs. However, he didn’t see action after the Arizona State game. Hahn kicked for the rest of the year and had nine touchbacks in 52 kickoffs.
More SpartanMag Spring Football Coverage:
* Els keeping some things quiet at LB
* Gilmore ramping up the tight ends.
* Hawkins bullish on young, hungry wide receiver group
* Coach Kap sees potential in o-line he is inheriting
* Coaching at Michigan State gives Hawkins goosebumps
* Antjuan Simmons moving it forward, keeping it normal
* Barnett officially moves to CB; shake-up taking place at punter
* Spartans add former UTEP punter as grad transfer
* Tucker optimistic during Detroit TV appearance
* Allen getting after it with coaching from big brother
* New staff inheriting a good leader in Antjuan Simmons
* What we learned about o-coordinator Jay Johnson
* Tucker drawing on NFL experience during practice stoppage
* For Hazelton, ’It’s all ball’
* New o-coordinator Jay Johnson ‘looking for a CEO’ at QB
MEN'S BASKETBALL Dr. G&W Analysis: Tournament Performance Metrics
- Spartans Illustrated Message Board
- 5 Replies
(Note that this analysis is a rehash of a post that I made last year following the Tournament, but the data is all still valid and pretty interesting, IMHO. Enjoy!)
Back in 2015, I went down a bit of a math rabbit hole based on a curiosity of mine. I wondered if it was possible to quantify the performance of coaches and teams in the NCAA tournament in ways other than simply wins and losses, Final Fours, and National Titles. In particular, I wanted to quantify under and overachieving in March. Along the way, I developed a few metrics that compared each coaches' and team's performance to the average performance of all other coaches / teams in similar tournament situations.
Somewhere along the way, I discovered that others had also formulated a similar metric called "PASE" (Performance Against Seed Expectation). My metrics were mathematically a bit different, and I settled onto two, one that I call PARIS (Performance Against Round Independent Seed) and PAD (Performance Against exact seed Differential). Three years ago, I gave a pretty detailed mathematical description of each metric and summarized notable coaches performance based on these metrics. That analysis can be found here. While the 2020 Tournament was sadly cancelled, I thought that is was a good time to revisit those metrics again.
For those that are not so interested in mathematical underpinnings of the PASE, PARIS, and PAD metric, the basic idea is as follows. PASE measures the number of games won by a specific coach or team per tournament relative to the average number of games won per tournament by all teams of that seed in tournament history. My PARIS metric essentially does the same thing, only it considers each game independently and not linked to other games played in the same tournament. The PAD metric uses a similar formula, but it instead considers each team's performance relative to the specific seed of the opponent in each game, as opposed to just the performance per round.
The PAD metric essentially corrects for the fact that some teams benefit from easy draws. For example, MSU made the Final Four this year by beating a 15-seed, 10-seed, 3-seed, and 1-seed. But, back in 2001, MSU made the Final Four by beating a 16-, 9-, 12-, and 11-seed. Clearly, the 2019 path was much tougher than the 2001 path. My PAD metric takes this into account, while MSU would get the same "credit" for both paths using both the PASE and PARIS metrics.
Based on data through the 2019 Tournament, we can look at how various coaches stack up with each other. As we will see, in all three metrics Tom Izzo is currently at the top of all three charts. Let's start with the PAD metric. The following chart shows the value of PAD over time for seven of the best tournament coaches in the last 40 years.
Not only in Izzo in first place, his current score in the PAD metric (8.41) is higher than any other coach at any other point in the history of the tournament (since seeding began in 1979). The only time any coach had a higher score was also Tom Izzo following the 2015 tournament. As for other coaches, Roy Williams is currently in second place, followed by John Beilein, John Calipari, Jim Boehiem (not shown), and Coach K. It is interesting to note that even though Coach K won titles in 2010 and 2015, his PAD score has taken a bit of a nosedive since 2001.
For reference, here is the same plot using the PASE metric:
In general, the story is the same here: Tom Izzo is awesome. However, in the case of the PASE metric, Izzo's current score (15.13) is slightly below the PASE score achieved by Coach K is 2001 (15.81) and in 2004 (15.24). But, Izzo's PASE following the 2015 tournament (16.32) remains the highest of any coach at any point in history. While MSU's loss to Middle Tennessee State in 2016 put a dent in that score, another strong run or two in March will likely allow Izzo to soon break his own record.
As for the PARIS metric, it tracks very closely with the PASE metric. So, by itself, those values do not add a lot of additional insight. However, as I mentioned above, the difference between the PAD and PARIS metric has to do with the relative difficulty of a team's NCAA tournament path. This difference in path "luck" can be quantified by subtracting a coach's PAD score from his PARIS score. In this way, it is possible to visualize true tournament performance (based on PAD) in relationship to "luck" (PARIS minus PAD). The chart below compares these two parameters for all 632 coaches to participate in the Tournament since 1979.
This chart really allows us to differentiate the "good" coaches in March from the not-so-good ones and the lucky ones from the not-so-lucky ones. As the chart shows, Izzo is clearly very good, and he has been historically slightly lucky relative to the field, similar to Denny Crum, Rick Pitino and Calipari. But, other "good" coaches have been more lucky, notably Beilein, Calhoun, Boeheim, Coach K, and the king of all tournament luck: Billy Donovan.
A few good coaches are rather unlucky, notably Roy Williams and Rollie Massimino. Lute Olson was very unlucky (and right at zero in the PAD metric). Meanwhile despite some positive luck in their draws, Bill Self, Bob Huggins, and Tony Bennett (despite the 2019 Title) are all still solidly in the "under-achiever" category. But, they can all take solace in the fact that they aren't Rick Barnes, the current champion of under-achievement in March (Gene Keady, Fran Dunphy, and Jamie Dixon are also thankful for Coach Barnes).
As for other data comparisons, there is one more metric that I developed that is interesting to examine and that is each coach's tournament performance not in terms of seed, but instead relative to the Kenpom efficiency margin of each team. In effect, this is similar to a measure of a coach's performance relative to the spread, although it only considers winning and losing, and not the final scores of the games. Another way to think about this is that this metric (which I dubbed "PRAKAEM" or Performance Relative to Average Kenpom Adjusted Efficiency Margin) is that is adjusts for teams actual strength as opposed to just their seed. One downside of this metric is that the data only goes back to 2002, but it is still useful. Below I plot each coach's PAD vs, their PRAKAEM back to 2002.
There are a couple of take-aways from this data. First of all, the PRAKAEM metric is one of the few tournament metrics where Izzo is not in first place. He is actually only in 5th, behind Roy Williams, Boeheim, Beilein, and Calipari. But, also keep in mind that this set of data does not include Izzo's Final Four runs of 1999-2001.
But, what I find interesting here is the coaches that deviate significantly from the best fit line with PAD. As I mentioned above, the difference between the PAD and PRAKAEM metrics is essentially a measure of the accuracy of seeding, at least in reference to Kenpom efficiency. Most coaches are very close to the line. However, some coaches (Roy Williams, Boeheim, Calhoun, and Bill Self) are notably below the line, which suggests that either their teams have on balance been seeded higher than they should or that they have on balance played teams that are seeded lower than they should have been. It seems like the odds of consistently drawing under-seeded opponents are extremely small, so I guess this has more to do with the over-seeding of those coaches.
In contrast, Tom Izzo is notably well above the trend line. Moreover, Izzo is a clear outlier on this graph in general. This suggests that Izzo's teams, by far, have been seeded lower than their Kenpom efficiencies merit relative to literally every other team/coach since 2002. I don't mean to contribute to any conspiracy theories out there, but this data is a pretty strong piece of evidence that the committee has done a poor job of seeding MSU over the past two decades.
So, when it comes to performance against expectation, Tom Izzo is clearly awesome and likely the best coach in the history of the tournament. But, there are other factors to consider as well. So with the remainder of my space here today, let's just review Tom Izzo's current standings in a wide variety of other tournament performance statistics.
Total Win Percentage:
Tom Izzo: 52-21 (71.2%)
Good for 9th place among all coaches with more than 4 appearances since 1979 and 4th place among active coaches (behind Coach K, 76.4%, Roy Williams, 75.2%, and Calipari, 74.7%)
Wins as the Lower Seed:
Tom Izzo: 15
Good for 1st place all-time. Jim Boeheim has 13, and Massimino and Lute Olson retired with 11. The next closest active coaches are Bruce Pearl, Beilein, Gregg Marshall, and Mark Few, all with seven. Also amazing is that Izzo actually has a winning record (15-13) as the lower seed.
Win Percentage as the Higher Seed:
Tom Izzo: 37-8 (82.2%)
Just because Izzo is great as the underdog doesn't mean he struggles as the favorite. Izzo's win percentage as the higher seed is good enough for 8th place all-time among coaches with greater than 10 appearances and 4th place among active coaches. The Top 3 are John Beilein (86.4%), Roy Williams (84.1%), and Calpari (83.3%). Denny Crum and Rich Majerus both retired at 87.5%.
Wins over No. 1 seeds:
Tom Izzo: 5
Good enough for a tie for 4th place all time behind Coach K (8), Lute Olson (7), and Roy Williams (6). Izzo is tied with Boeheim, Rick Pitino, and Dean Smith. Since 1998, Izzo's five wins are more than any other coach. Izzo also has the most wins over No. 1 seeds and No. 2 seeds combined (9), the top three seeds combined (13), the top four seeds combined (17), and even the top five seeds combined (18) since 1998.
Two-Day Prep Record:
Tom Izzo: 23-6 (79.3%)
Izzo is legendary for his 2-day preps and his record on the 2nd game of the weekend bears this out. This winning percentage is 2nd all-time (behind Denny Crum at 82%) and 1st among active coaches with at least 10 tournament appearances. There are a handful of coaches with higher percentages (like Steve Lavin, Tom Crean, and Jim Larranaga) but these records are all based on less than 10 total games. I will also note here that Izzo's record on the 1st day of a weekend is "only" 29-15 (66%) which is one of those rare stats where Izzo is pretty average.
Sweet Sixteens:
Tom Izzo: 14
Good for 6th place behind Coach K (25), Roy Williams (19), Boeheim (18), Dean Smith and Calpari (15). Also, Tom Izzo has made the Sweet 16 in 63.6% of his tournament appearances, which is also tied for 6th place all-time. Since 1998, only Coach K has more Sweet 16s (17).
Elite Eights:
Tom Izzo: 10
Good for a tie for 5th place behind Coach K (16), Roy Williams (13), Calipari and Pitino (12). Izzo is tied with Dean Smith and Bill Self. Izzo has made the Elite 8 in 45.5% of his tournament appearances, which is 7th place all-time. Since 1998, Izzo is tied for first place with Self, Roy Williams, and Calipari.
Final Fours:
Tom Izzo: 8
Since 1979, Izzo has now moved into 3rd place behind only Coach K (12), and Roy Williams (9). Tom Izzo has made the Final Four in 36.4% of his tournament appearances, which is the best rate in history among coaches with more than 8 appearances. Larry Brown (37.5% in 8 years) and Brad Stevens (40% in 5 years) have slightly better rates, but with far fewer attempts. Also, since 1998, Izzo is alone in 1st place in this category.
As we all know, Izzo has struggled a bit once he gets to the Final Four. His record in National Semifinal games (2-6, 25%) is the lowest of any coach with more than 2 appearances in the Final Four. Izzo has only won a Championship in ~5% of his tournament appearances, while the other coaches in his peer group typically have a rate between 10 and 15%.
But, by all indications, Tom is going to be hanging around in East Lansing for a while now, and the future looks very bright. More than one of Izzo's Final Four runs has been submarined by injuries, 2019 included. In 2020, well, we all know what happened. But, Coach Izzo and MSU's luck is bound to turn around at some point. Hopefully Coach Izzo can add to his amazing stats in the years to come.
Back in 2015, I went down a bit of a math rabbit hole based on a curiosity of mine. I wondered if it was possible to quantify the performance of coaches and teams in the NCAA tournament in ways other than simply wins and losses, Final Fours, and National Titles. In particular, I wanted to quantify under and overachieving in March. Along the way, I developed a few metrics that compared each coaches' and team's performance to the average performance of all other coaches / teams in similar tournament situations.
Somewhere along the way, I discovered that others had also formulated a similar metric called "PASE" (Performance Against Seed Expectation). My metrics were mathematically a bit different, and I settled onto two, one that I call PARIS (Performance Against Round Independent Seed) and PAD (Performance Against exact seed Differential). Three years ago, I gave a pretty detailed mathematical description of each metric and summarized notable coaches performance based on these metrics. That analysis can be found here. While the 2020 Tournament was sadly cancelled, I thought that is was a good time to revisit those metrics again.
For those that are not so interested in mathematical underpinnings of the PASE, PARIS, and PAD metric, the basic idea is as follows. PASE measures the number of games won by a specific coach or team per tournament relative to the average number of games won per tournament by all teams of that seed in tournament history. My PARIS metric essentially does the same thing, only it considers each game independently and not linked to other games played in the same tournament. The PAD metric uses a similar formula, but it instead considers each team's performance relative to the specific seed of the opponent in each game, as opposed to just the performance per round.
The PAD metric essentially corrects for the fact that some teams benefit from easy draws. For example, MSU made the Final Four this year by beating a 15-seed, 10-seed, 3-seed, and 1-seed. But, back in 2001, MSU made the Final Four by beating a 16-, 9-, 12-, and 11-seed. Clearly, the 2019 path was much tougher than the 2001 path. My PAD metric takes this into account, while MSU would get the same "credit" for both paths using both the PASE and PARIS metrics.
Based on data through the 2019 Tournament, we can look at how various coaches stack up with each other. As we will see, in all three metrics Tom Izzo is currently at the top of all three charts. Let's start with the PAD metric. The following chart shows the value of PAD over time for seven of the best tournament coaches in the last 40 years.
Not only in Izzo in first place, his current score in the PAD metric (8.41) is higher than any other coach at any other point in the history of the tournament (since seeding began in 1979). The only time any coach had a higher score was also Tom Izzo following the 2015 tournament. As for other coaches, Roy Williams is currently in second place, followed by John Beilein, John Calipari, Jim Boehiem (not shown), and Coach K. It is interesting to note that even though Coach K won titles in 2010 and 2015, his PAD score has taken a bit of a nosedive since 2001.
For reference, here is the same plot using the PASE metric:
In general, the story is the same here: Tom Izzo is awesome. However, in the case of the PASE metric, Izzo's current score (15.13) is slightly below the PASE score achieved by Coach K is 2001 (15.81) and in 2004 (15.24). But, Izzo's PASE following the 2015 tournament (16.32) remains the highest of any coach at any point in history. While MSU's loss to Middle Tennessee State in 2016 put a dent in that score, another strong run or two in March will likely allow Izzo to soon break his own record.
As for the PARIS metric, it tracks very closely with the PASE metric. So, by itself, those values do not add a lot of additional insight. However, as I mentioned above, the difference between the PAD and PARIS metric has to do with the relative difficulty of a team's NCAA tournament path. This difference in path "luck" can be quantified by subtracting a coach's PAD score from his PARIS score. In this way, it is possible to visualize true tournament performance (based on PAD) in relationship to "luck" (PARIS minus PAD). The chart below compares these two parameters for all 632 coaches to participate in the Tournament since 1979.
This chart really allows us to differentiate the "good" coaches in March from the not-so-good ones and the lucky ones from the not-so-lucky ones. As the chart shows, Izzo is clearly very good, and he has been historically slightly lucky relative to the field, similar to Denny Crum, Rick Pitino and Calipari. But, other "good" coaches have been more lucky, notably Beilein, Calhoun, Boeheim, Coach K, and the king of all tournament luck: Billy Donovan.
A few good coaches are rather unlucky, notably Roy Williams and Rollie Massimino. Lute Olson was very unlucky (and right at zero in the PAD metric). Meanwhile despite some positive luck in their draws, Bill Self, Bob Huggins, and Tony Bennett (despite the 2019 Title) are all still solidly in the "under-achiever" category. But, they can all take solace in the fact that they aren't Rick Barnes, the current champion of under-achievement in March (Gene Keady, Fran Dunphy, and Jamie Dixon are also thankful for Coach Barnes).
As for other data comparisons, there is one more metric that I developed that is interesting to examine and that is each coach's tournament performance not in terms of seed, but instead relative to the Kenpom efficiency margin of each team. In effect, this is similar to a measure of a coach's performance relative to the spread, although it only considers winning and losing, and not the final scores of the games. Another way to think about this is that this metric (which I dubbed "PRAKAEM" or Performance Relative to Average Kenpom Adjusted Efficiency Margin) is that is adjusts for teams actual strength as opposed to just their seed. One downside of this metric is that the data only goes back to 2002, but it is still useful. Below I plot each coach's PAD vs, their PRAKAEM back to 2002.
There are a couple of take-aways from this data. First of all, the PRAKAEM metric is one of the few tournament metrics where Izzo is not in first place. He is actually only in 5th, behind Roy Williams, Boeheim, Beilein, and Calipari. But, also keep in mind that this set of data does not include Izzo's Final Four runs of 1999-2001.
But, what I find interesting here is the coaches that deviate significantly from the best fit line with PAD. As I mentioned above, the difference between the PAD and PRAKAEM metrics is essentially a measure of the accuracy of seeding, at least in reference to Kenpom efficiency. Most coaches are very close to the line. However, some coaches (Roy Williams, Boeheim, Calhoun, and Bill Self) are notably below the line, which suggests that either their teams have on balance been seeded higher than they should or that they have on balance played teams that are seeded lower than they should have been. It seems like the odds of consistently drawing under-seeded opponents are extremely small, so I guess this has more to do with the over-seeding of those coaches.
In contrast, Tom Izzo is notably well above the trend line. Moreover, Izzo is a clear outlier on this graph in general. This suggests that Izzo's teams, by far, have been seeded lower than their Kenpom efficiencies merit relative to literally every other team/coach since 2002. I don't mean to contribute to any conspiracy theories out there, but this data is a pretty strong piece of evidence that the committee has done a poor job of seeding MSU over the past two decades.
So, when it comes to performance against expectation, Tom Izzo is clearly awesome and likely the best coach in the history of the tournament. But, there are other factors to consider as well. So with the remainder of my space here today, let's just review Tom Izzo's current standings in a wide variety of other tournament performance statistics.
Total Win Percentage:
Tom Izzo: 52-21 (71.2%)
Good for 9th place among all coaches with more than 4 appearances since 1979 and 4th place among active coaches (behind Coach K, 76.4%, Roy Williams, 75.2%, and Calipari, 74.7%)
Wins as the Lower Seed:
Tom Izzo: 15
Good for 1st place all-time. Jim Boeheim has 13, and Massimino and Lute Olson retired with 11. The next closest active coaches are Bruce Pearl, Beilein, Gregg Marshall, and Mark Few, all with seven. Also amazing is that Izzo actually has a winning record (15-13) as the lower seed.
Win Percentage as the Higher Seed:
Tom Izzo: 37-8 (82.2%)
Just because Izzo is great as the underdog doesn't mean he struggles as the favorite. Izzo's win percentage as the higher seed is good enough for 8th place all-time among coaches with greater than 10 appearances and 4th place among active coaches. The Top 3 are John Beilein (86.4%), Roy Williams (84.1%), and Calpari (83.3%). Denny Crum and Rich Majerus both retired at 87.5%.
Wins over No. 1 seeds:
Tom Izzo: 5
Good enough for a tie for 4th place all time behind Coach K (8), Lute Olson (7), and Roy Williams (6). Izzo is tied with Boeheim, Rick Pitino, and Dean Smith. Since 1998, Izzo's five wins are more than any other coach. Izzo also has the most wins over No. 1 seeds and No. 2 seeds combined (9), the top three seeds combined (13), the top four seeds combined (17), and even the top five seeds combined (18) since 1998.
Two-Day Prep Record:
Tom Izzo: 23-6 (79.3%)
Izzo is legendary for his 2-day preps and his record on the 2nd game of the weekend bears this out. This winning percentage is 2nd all-time (behind Denny Crum at 82%) and 1st among active coaches with at least 10 tournament appearances. There are a handful of coaches with higher percentages (like Steve Lavin, Tom Crean, and Jim Larranaga) but these records are all based on less than 10 total games. I will also note here that Izzo's record on the 1st day of a weekend is "only" 29-15 (66%) which is one of those rare stats where Izzo is pretty average.
Sweet Sixteens:
Tom Izzo: 14
Good for 6th place behind Coach K (25), Roy Williams (19), Boeheim (18), Dean Smith and Calpari (15). Also, Tom Izzo has made the Sweet 16 in 63.6% of his tournament appearances, which is also tied for 6th place all-time. Since 1998, only Coach K has more Sweet 16s (17).
Elite Eights:
Tom Izzo: 10
Good for a tie for 5th place behind Coach K (16), Roy Williams (13), Calipari and Pitino (12). Izzo is tied with Dean Smith and Bill Self. Izzo has made the Elite 8 in 45.5% of his tournament appearances, which is 7th place all-time. Since 1998, Izzo is tied for first place with Self, Roy Williams, and Calipari.
Final Fours:
Tom Izzo: 8
Since 1979, Izzo has now moved into 3rd place behind only Coach K (12), and Roy Williams (9). Tom Izzo has made the Final Four in 36.4% of his tournament appearances, which is the best rate in history among coaches with more than 8 appearances. Larry Brown (37.5% in 8 years) and Brad Stevens (40% in 5 years) have slightly better rates, but with far fewer attempts. Also, since 1998, Izzo is alone in 1st place in this category.
As we all know, Izzo has struggled a bit once he gets to the Final Four. His record in National Semifinal games (2-6, 25%) is the lowest of any coach with more than 2 appearances in the Final Four. Izzo has only won a Championship in ~5% of his tournament appearances, while the other coaches in his peer group typically have a rate between 10 and 15%.
But, by all indications, Tom is going to be hanging around in East Lansing for a while now, and the future looks very bright. More than one of Izzo's Final Four runs has been submarined by injuries, 2019 included. In 2020, well, we all know what happened. But, Coach Izzo and MSU's luck is bound to turn around at some point. Hopefully Coach Izzo can add to his amazing stats in the years to come.
