Characteristics of Final Four Teams/Their Relation to Beilein's Michigan
Link: https://bighouseanalytics.wordpress.com/2017/03/01/madness-explained-a-…
This study examines all of the Final Four teams from 2002-2016 and, based on their characteristics, establishes a criteria for making the Final Four and winning the championship. It's a long article; I've posted the Final Four criteria below for those who aren't interested in combing through the entire piece. Also, the article only briefly mentions Michigan and Beilein - those two are covered with more specificity below.
FINAL FOUR CRITERIA:
– AdjOE rank of at least 50 or greater (54 out of 60 have met)
– AdjDE rank of at least 40 or greater (54 out of 60 have met)
– Luck rank of at least 199 or greater (50 out of 60 have met)
– AdjTempo rank of at least 299 or greater (56 out of 60 have met)
– 3PA/FGA rank of at least 100 or lower (50 out of 60 have met)
– Winning Streak of 10 or less (55 out of 60 have met)
– Ranked in preseason AP poll (50 out of 60 have met)
50 out of 60 Final Four teams in the study met all or all but one of the criteria (2013 Michigan included*). Most Beilein teams decidedly do not, though, and often violate at least three different criteria (too high in 3PA/FGA, too low in AdjTempo, too low in AdjDE). 2 teams out of 60 (2016 Syracuse and 2011 VCU) have made the Final Four while not meeting 3 or more criteria.
Beilein has had a productive run at Michigan and is, by all accounts, a wonderful person and ambassador, but his program construction is fatally flawed in the area of consistent success in March.
*2013 Michigan began the NCAA tournament with a faster tempo than normal (rank: 200s), a lower 3PA/FGA than normal (rank: 100s) and a better AdjDE than normal (rank: 66th).
They are average at pushing the pace after a miss according to the data from hoop-math that I provided in my original post. To wit, if you sort this page by "% of initial FGA -- Rebound, 0-10s", you will find Michigan at #164 of 351, with 12.7%.
Now to be clear, that's 12.7% of all initial field goal attempts; it does not mean that they push the pace after 12.7% of all misses. In fact, as you can see here, they get an initial shot within 10 seconds of an opponent miss about 37.4% of the time. (12.7 / (12.7 + 21.2) -- the first two values in the % of initial attempts column). Now, you may quibble with their definition of transition offense -- you don't necessarily have to be hustling to get off a shot within 10 seconds of a rebound -- but when you're generating the data automatically from the box score, it's as good a place to start as any.
Michigan almost never goes into transition after a make -- they do so on about 2% of all opportunities. They usually go into transition after a steal -- 66% of the time.
Incidentally, you can do the calculation you wanted from the data presented on the team page. Opp Miss: 12.7% * 0.631 eFG = 0.080137; Opp Make: 1.8% * 0.327 eFG = 0.005886; Steal: 5.4% * .753 eFG = 0.040662. Divide the first number by the total, and 63% of Michigan's transition points come off of rebounds. (Note that baskets after dead ball turnovers appear to be excluded from this analysis).
So basically the conclusion is that the best teams go to the final four and only really elite teams win the championship. Beilein's "fatal flaw" is that none of his teams except 2013 and 2014 were good enough to be Final Four teams. Although I can't look it up, I'm guessing Michigan 2014 missed out on 2 criteria. I don't think anyone would argue that we won't win a national championship without a much better defense than we have this year.
I did really enjoy the article though and will go back and look at it when the brackets come out.
Not necessarily. It's more like the most complete teams go to the Final Four. UCLA is one of the best teams in the country but their defense stinks and that may be the cause of an early exit for them. They also rely on 3 point shooting, so one off-shooting night and they're done for. The nature of the tournament favors teams that play a certain style.
If Beilein's program is "fatally flawed" in the tournament, then how do you explain that he's outperformed tournament expectations during his tenure?
https://fivethirtyeight.com/datalab/tom-izzo-is-the-best-coach-in-moder…
Even more simplistically, he has proven it wrong given the 100% plausibility that a team of his could win it all.
To be fair, I actually remember that 538 piece and found it pretty interesting - I realize you were being facetious with your reply, but still you have to admit that Beilein is in some interesting (and not bad by any means) company if you buy into the statistics being presented here.
If we want to frame it in terms of expectations, then I would say that - on this board - tournament expectations are pretty guarded and conservatively stated for the most part, but that's sort of what makes these numbers interesting really.
He had one team that came close to winning it all. Maybe the second was a Mitch McGary short of doing so. The question is, can he replicate it? So far the answer appears to be no.
To be fair, he did well in the tournament with the Pittsnogle teams as well. One year they were an overtime away from the Final Four.
"Maybe the second was a Mitch McGary short of doing so. The question is, can he replicate it?"
That was the team in 2014. That means he's had 2 years to replicate it before this one. Both teams had major injury issues and unexpected early departures. Seems extreme to me. Someone will probably site recruiting and defense as reasons to light everything on fire. He lost 2 assistants this last offseason and made hires to address both of those things.
I thought Beilein cleaned house and let his assistants go or at least took the Carr approach and told them they should look for other jobs.
What are you even talking about? They took HC jobs.
His coaching philosophies and style of play, according to this research, place a ceiling on the program's possibilities of making a Final Four and winning a championship.
Teams pretty much never make the Final Four when they have a high 3PA/FGA rank, low tempo rank, and low AdjDE rank (only 2 teams out of 60 have failed to meet any 3 criteria and made the Final Four in the KenPom era). In every single year under Beilein, except for 2013, Michigan has failed to meet at least 3 of the 7 criteria, including the 2012 team (4 seed) and the 2014 team (2 seed).
Whether this should be fixed with a change of philosophy or the recruiting of better players, I do not know. Whether this is "acceptable" for the masses moving forward, I do not know. The numbers are what they are. Beilein creates an uphill battle for himself with regard to making Final Fours and winning titles because of the way he builds his programs.
Beilein has started to address this by bringing in a defensive minded coach. He's also talked about changes in skills he now recruits.
How's that working out.
It's working out well.
- Defensively, Michigan held Nebraska to .88 points per possessioon.
- Michigan has vastly improved in points per possession against it the second half of the season compared to the first half of the season.
Yep, I'm very happy with how things are working out. Some people, on the other hand, are never happy.
and Nova, who won it all, was 31st in the country with 0.42 3PA/FGA. Basketball is becoming more three heavy at the college level and pro level for a reason. Beilein was ahead of the curve with his offensive philosophy and we've had excellent offenses while he's been here.
This analysis basically says you need to be good at offense and good at defense to make the final four. Duh. We are almost always excellent at offense and fairly suspect at defense. Our defense keeps us from being better. Not our style of offense or pace.
From the article ...
"On 3PA/FGA: There may be something to monitor moving forward with regard to the increase in 3PA/FGA across the entire sport of basketball and its correlation to NCAA tournament success. Historically, having 3 teams make the Final Four who finished the season in the top 60 of 3PA/FGA as we did in 2016 is an incredible outlier. If it happens again in 2017, we may have the start of a trend. TBD."
Also from the article ... on 2016 Villanova:
"However, in an unforeseeable turn of events, Villanova completely altered their offensive profile throughout their 6 game run through the tournament, shooting 3s at a rate that would have ranked them at 159 in 3PA/FGA and making them at a 50% clip (a nearly 16 percentage point jump from their regular season average). This study opines that this transformation was neither predictable (in 2015, with 6 of the same players that would win the title a year later, the Wildcats had a 3PA/FGA figure of 42.9 for the season (rank: 22) and 42.3 for the tournament alone) nor repeatable (2017 Villanova is currently ranked 25th in 3PA/FGA at 43.9 and features only one rotation player who has not shot at least 50 three-pointers this season). Thus, it is treated as a statistical curiosity and nothing more (for now)."
The study lists 5 criteria aside from AdjOE and AdjDE, some of which have very little to do with being good at offense and at defense. It doesn't pretend to make it as simple as "be good at offense and defense and you'll make the Final Four".
or other rule violations factor in?
This is not a good analysis. In fact, I'd argue that this is a particularly poor analysis, as it shows absolutely no understanding of statistics. (If the author is reading this, I apologize for being harsh, but I stand by my comments).
Let's start with tempo. While there is a minor statistical argument in favor of a faster tempo -- if you have a per-possession efficiency advantage, you benefit from there being more possessions in the game -- it's not clear why it would be some kind of meaningful measure of a team's Final Four worthiness. And... spoiler alert.. it's not. 93.3% of teams had an adjusted tempo of a rank of 299 or better, you say? There are 351 teams in college basketball. 299/351 is 85.2%. If there were no correlation whatever between tempo and being a Final Four team, you'd expect 51/60 teams to meet this criterion. 56 did, which seems well within the margin of error (not only is 60 a small sample size, but teams tend not to change their tempo much from year to year, so repeat Final Four visitors may skew this value whether it's causative or not).
The winning streak criterion is absolutely asinine. There is zero reason to believe that regular season / postseason tournament winning streaks have anything to do with success in the tournament, and the particular suggestion made -- look for a shorter winning streak -- is really just a way to eliminate the champions of one-bid leagues. Nobody picked against Kentucky a couple of years ago because of their winning streak entering the tournament.
The 3PA/FGA shooting one? 83.3% of Final Four teams are not in the top 100? OK, fine; 72% of the country isn't in the top 100 either. (252/351).
If this were a valid statistical survey, you should expect to see things like single-variable regression analyses and correlation coefficients. I, personally, would expect to see the analysis done on raw values rather than on ranks, but that's a bias that should be tested.
Regardless, this is no more meaningful than any other set of cherry-picked statistics. As Tjanderson9705 wrote, you could almost certainly do a similar analysis to set Michigan up as a Final Four favorite, and it would be no more or less valid than this one.
This is well-put. I would inform you, though, that this is not so much a statistical analysis as it is an examination of trends over an admittedly brief time period (KenPom's database only dates back to 2002). If you find it irrelevant that there has never been a champion who had a tempo in the 300s in the past 15 years and never* been a champion who ranked in the top 100 of 3PA/FGA in the past 15 years, that is fine; it is still undeniably a trend and does wonders for ruling out possible selections when filling out a bracket. This is not meant to be an end-all-be-all guide for who can make the Final Four or win the championship and who can not but, instead, a guide for establishing who is more likely to make it between teams that might otherwise be indistinguishably close. The piece itself acknowledges the subjectivity and conjectural nature that is inherent within this type of study.
Your defense of the useless nature of examining winning streaks does not really make sense - teams entering the tournament on a 10-game winning streak or higher have won it all exactly once (2013 Louisville). Considering there were 17 1- or 2-seeds who qualified for this criteria, that seems interesting.
It's certainly possible that all of these criteria point to nothing more than outliers that will be blasted to statistical smitheereens by the data provided by next 15 years; it does seem odd to bank on the fact that they prove nothing because they don't include single-variable regression analyses or correlation coefficients.
*2016 Villanova was the first, but really they weren't; they inexplicably shot significantly less 3s in the tournament than they did in the regular season, to the point that they would have met the criteria comfortably if their tournament 3PA/FGA was their regular season 3PA/FGA.
Thank you for not taking my criticism personally. No offense was intended. :-)
Yes, these are trends, but they're being presented as holding predictive value. When someone says "MAAR is shooting 45% from 3 over the last six games," the implication is that he is hot and thus will continue to do so. You're suggesting that people should use this information to fill out their bracket, so it seems fair to look more deeply at the validity of the statistical approach you're using to gather the data. (Furthermore, some people in the thread are/were using this information to argue for the Michigan basketball program to take a different direction, which -- while it may not be what you intended -- clearly means scrutiny should be in order).
The success rate for winning-streak teams isn't meaningful in isolation. In order to tell whether or not there's a correlation between being on a winning streak and winning the tournament; you'd need to know their pre-tournament expected likelihood of winning the championship compared to the actual results. Considering only #1 and #2 seeds, if they were all equally likely to win the tournament, and you had 17 such teams, and you assumed that no seed lower than a 2 had a chance of winning, you'd expect about two of those teams to win. The fact that none have may or may not be meaningful: if your criterion is "failing to win the championship," you can find a lot of statistics that will correlate very well with that. And, by restricting your success criterion to championships, you make it very vulnerable to noise. For example, Butler was on a 20-game winning streak before the 2010 tournament. (They also finished in the top 100 in 3PA/FGA). They failed to win the championship when a beautiful half-court shot didn't quite bank in. If the trajectory of that shot is an inch different, that's a seccond long-winning-streak champion and a 3-point-happy one as well.
Again, I'm not trying to be harsh, but the simple truth is that these statistics do prove nothing without doing the correlation analyses, because the entire point of those analyses is to find out how much they prove -- that is, what percentage of success can be predicted from the value of a given variable.
There are more problems beyond that, however. Take the "300+ tempo" test, for example. First of all, you wouldn't normally use a ranking, because that artificially evens out the data in a way that may not be meaningful. For 2017, the raw difference between the #1 tempo (Savannah State @ 81.5) and the #25 tempo (Rider @ 72.6) is the same as the difference between the #36 tempo (Buffalo @ 71.9) and the number #343 tempo (Michigan -- coincidentally -- @ 63.0). You might use a deviation from the mean number of possessions, for example, but probably not a rank.
Then, you've got the problem of arbitrary boundaries. Why #300? Is it that much better to be #299 than #300? This year, they have identical 65.7 adjusted tempo values. If there is a correlation, one would expect it to be continous, which is why you'd run a linear regression to find the correlation coefficient, which -- roughly speaking -- tells you how much you're likely to change your success value (e.g., Final Fours) by changing your variable (tempo).
Next, you're handicapped by the lack of specificity of the test you're trying to run. In general, the less likely an outcome would be under the null hypothesis, the better your statistical test. In this case, using a top-299 ranking means that 85.2% of the population is in your test group -- so you need an awful lot of data before you can rule out random chance.
That's actually related to another problem -- you're not trying to prove that fast tempo makes you more likely to go to the Final Four, but rather that slow tempo makes you less likely to go. While that may seem like picking nits, it's not -- because 347 teams each year don't go to the Final Four, including 64 tournament teams. That makes it much, much harder to isolate any particular reason.
I don't mean to discourage you -- I know it must have taken a lot of time to do this analysis, and I hope that you're able to win your tournament pool. :-) I hope that you're able to iterate on this and find some new insights into college basketball. I just don't think that these particular ones are those insights. :-)
News flash:
He is making the NCAAs again. No way they replace him until he retires.
Red took hockey to NCAAs only 3 of the last 4 years and he is still there.
No drama program that makes the 64 every year is OK with Michigan.
No drama program that makes the 64 every year is OK with Michigan.
Too bad he doesn't make the 64 every year, he's missed it three times already.
Also is leaders and best just bullshit or are people really OK with a participation ribbon from a tournament that has 64 FREAKING TEAMS.
Not OK with it, but administration at UofM is.
Unless there is a "concussiongate" incident I very much doubt he will be replaced if he makes the NCAAs somewhat regularly.
There is no better way to describe this postseason in advance than interesting. The idea that "Anything can happen in March" is a very common cliché, but a team like this brings more life to that idea.