Thursday, October 9, 2014

Thoughts on the Big East & T-Rank

The most vociferous criticism of the T-Rank results has been about the Big East projections, so I thought I'd take a moment to discuss them here.

Here are the T-Rank projections for the Big East:

T-Rank Team Pyth Proj Hanner
18 Villanova 0.8960 1
36 Butler 0.8047 9
42 Creighton 0.7885 8
50 Georgetown 0.7722 2
74 Xavier 0.7190 3
80 Marquette 0.7087 5
90 Seton Hall 0.6851 7
96 St. John's 0.6740 4
118 Providence 0.5866 6
186 DePaul 0.4320 10

The fourth column is the projected Big East rankings published by Dan Hanner, which tracks the more conventional wisdom.

So there are some clear outliers here, and I would certainly rank the teams differently. The ones that stick out are Butler, Creighton (both too high) and Providence (too low). So what went wrong?

Well, maybe nothing. One of the reasons to do this was to maybe gain some insights where conventional wisdom could be wrong. Is this one of those cases? Probably not! (But I'll be rooting for Creighton and Butler this year, obviously.)

Butler

I've already explained part of what is going on with Butler in the comments to my post at B5Q. Conventional wisdom sees Butler as a program that has lost its mojo because its wunderkind coach has gone to the NBA. T-Rank is blind to that because it doesn't directly account for coaching. Instead, it looks at prior program performance. So coaching changes, particularly ones where a mid-major loses a star coach, can muck things up pretty well.

What T-Rank does like about Butler is the return of a very, very good player (Roosevelt Jones) joining three other decent upperclassmen after a down year for a solid program. So it basically projects Butler to be similar to the team it was the year before last. Most think that won't happen, but it is within the realm of possibility.

One of the hardest things for me to figure out how to model was how to handle this situation where an injured player comes back after sitting out all or most of a year. For one thing, I'm probably not aware of all or even most of these situations, so it might provide a relative boost for the high profile programs that I do get news about. It also mucks up one of the basic inputs to the ranking -- returning minutes. This is something I'll be taking a look at next off-season to see whether the model is better without even accounting for these scenarios.

Creighton

Creighton is a similar story in that it is really hard for the T-Rank model to predict sudden, massive declines. The model is based on the assumption that most programs, most of the time, will progress or decline in a linear fashion from year to year. This is a reasonable and useful assumption, but it will be false in lots of cases. Most people think Creighton is such a case because it loses a superstar, Doug McDermott*, who has propped up the program for the three-year look-back period. They also lose three other seniors. So there's very good reason to think that Creighton will take a huge step back this year.

And, for what it's worth, T-Rank expects Creighton to take a big step back -- last year Creighton was a 3-seed in the tourney, and this year T-Rank projects it as a bubble team at #42. Still, it is common for good teams to lose great players and just reload (think Kentucky, Duke, Kansas, etc.), so that has to be accounted for in the model. Is Creighton likely to reload? Probably not, and that's the issue.

Providence College

On the other end of the spectrum, Providence projected at ninth looks crazy and I agree. But this is actually a scenario where -- as long you take the actual numbers with skepticism -- the T-Rank can maybe provide some helpful insight. There's a pretty good chance people are overrating Providence this year. Dan Hanner's comments were right on, in my opinion:
But despite a very positive outlook in the long-run, 2014-15 looks like a bit of a transition year. First, this team barely snuck into the NCAA tournament last year. They may have given North Carolina a scare, but their margin-of-victory was only 51st in the nation. And losing Bryce Cotton, their most efficient player, their best passer, their best scorer, and a player who never left the floor, is going to hurt.
Second, the team is going to have to give more minutes to freshmen. Last year with Brandon Austin and Rodney Bullock suspended, Providence basically never used any first-year players. This year with Bullock eligible and an outstanding recruiting class coming in, Providence projects to give substantially more minutes to its young players. And while many of them are talented, playing inexperienced players will lead to more mistakes. There will be games where players don’t rotate properly defensively, and games where players simply stand around and don’t run the offensive sets with the same crispness of a veteran team. 
And while many of the names sounds scary, many of the players don’t have great projections for this season. PG Kris Dunn was an elite recruit, but he has struggled massively with injuries, and hasn’t been able to perform at an elite level in his two seasons with the team. That may mean more minutes for freshman PG Kyron Cartwright. Meanwhile Carson Desrosiers is a quality shot-blocking big man, but he is a very passive offensive player. And while transfer Junior Lomomba has received some positive reviews on the team’s European tour, he didn’t have great efficiency numbers at Cleveland St. That doesn’t necessarily bode well for him being an efficient player in the Big East. Rodney Bullock seems like a household name at this point because of the off-court issues, but I also think we need to recognize that he was only a 3-star recruit. He does not necessarily project as a star. Honestly, the true freshmen may be deserving of the most love. Paschal Chukwu and Jalen Lindsey were both consensus Top 100 recruits, and Ben Bentil may be the most polished of the young big men.
T-Rank especially penalizes PC for the loss of superstar Bryce Cotton. This is an example of how it's almost impossible to create an objective model that really accounts for all situations: the Creighton ranking is probably too high because it doesn't fully account for the loss of superstar Doug McDermott*; but in tweaking the model to account for that more, Providence got punished more for the loss of superstar Bryce Cotton. 

Be Humble, Big East Fans

Finally, although I fully agree that no human should rank Butler or Creighton as contenders in the Big East, and no one should relegate Providence to also-ran status, overall I think the craziness of the T-Rank for the Big East (and it's worth noting that the Big East projection is by far the most removed from conventional wisdom) may be telling us something: it could be a crazy year in the Big East. And we don't have to look back very far to see something similar. Why, just last year, Marquette was the unanimous preseason pick to win the conference, and they ended up missing the NIT! Most projections had Villanova (the eventual champion) outside the top 3, and neither they nor Creighton was ranked in the preseason AP poll. (Only Marquette was.)

So I think the idea that the Big East is an easily predictable conference is fanciful. So many teams are in transition, so many teams have new identities - let's not pretend we know how this is going to turn out. T-Rank certainly doesn't!

[message from T-Rank: ****T-RANK DOES NOT PRETEND TO KNOW BECAUSE T-RANK ACTUALLY KNOWS****end transmission]

Okay, T-Rank, okay. Never mind.

*Kind of funny that I consistently referred to Doug McDermott as "Doug Creighton" throughout this post.

4 comments:

  1. It appears that the T-Rank skews the data towards a more conservative projection. As an unbiased outsider, I think T-Rank has the Big East ranked correctly, with the exception of Creighton - ranked to low. The Bluenotes analysis has them 30th (preseason), with room to rise.
    Nice job. Another good tool for analysis.

    ReplyDelete
  2. Would u do the wins and SOS for the BE please.

    ReplyDelete
  3. GtmoBlue - I believe the BE has a true round robin conference schedule, so there is no significant different in SOS. The only difference would be that the good teams will have an "easier" schedule because they don't have to play themselves, and the bad teams will have "harder" schedule because they don't get to play themselves.

    ReplyDelete