Tuesday, March 16, 2010

Beyond Hummel - Other Critical Bracket Adjustments

Robbie Hummel’s injury wasn’t the only injury to consider this year.
What about the mid-season injury to Evan Turner at Ohio St?

Ohio St. with Evan Turner (28 games)
Adj Off Eff 121.0
Adj Def Eff 88.4
Pythag Win Perc = 0.9738

Ohio St. without Evan Turner (6 games)
Adj Off Eff 107.7
Adj Def Eff 96.3
Pythag Win Perc = 0.7841

So obviously we should use the “with Evan Turner” numbers.
But is this really necessary?

Ohio St.’s current numbers on kenpom.com
Adj Off Eff 119.0
Adj Def Eff 89.8
Pythag Win Perc = .9620

Not only did Evan play the majority of the season, but Ken also weights recent games more heavily. So there really is not much of a gain from accounting for this injury on a game-by-game basis.

Odds of beating UC Santa Barbara
With Evan Turner 96.8%
Without Evan Turner 74.7%
Current kenpom.com odds 95.4%

But there is some gain. Consider possible opponents in the later rounds

Odds of beating Georgetown
With Evan Turner 65.8%
Without Evan Turner 15.8%
Current kenpom.com odds 56.7%

Odds of beating Kansas
With Evan Turner 41.5%
Without Evan Turner 6.5%
Current kenpom.com odds 32.5%

And when you multiply this round-by-round, it starts to make a big difference to Ohio St.’s Final Four odds.

(By the way, special thanks to Brian Lerner at Hoya Prospectus. He emailed me injury adjustments, but then I was so exhausted on Saturday that I failed to use them. Then I decided to re-calculate them myself using the final regular season numbers which is why there was an even greater delay in this post.)

What about Notre Dame?
Let’s split them into three parts?

Notre Dame before Luke Harangody injury (25 games)
Adj Off Eff 119.6
Adj Def Eff 104.3
Pythag Win Perc = .8281

Notre Dame without Luke Harangody (5 games)
Adj Off Eff 129.6
Adj Def Eff 96.0
Pythag Win Perc = .9691

Notre Dame after Luke Harangody returns (4 games)
Adj Off Eff 114.1
Adj Def Eff 85.3
Pythag Win Perc = .9660

Notre Dame’s current numbers on kenpom.com
Adj Off Eff 119.8
Adj Def Eff 99.3
Pythag Win Perc = .8967

Now I think sample size is an issue here. There’s no way Notre Dame could maintain that offensive efficiency without Luke Harangody. That would be the greatest offense ever recorded. But the games did happen. They did put up back-to-back offensive ratings over 130 against Pittsburgh and Georgetown without Harangody on the floor.

And the trend on defense is becoming more than just a fluke. Notre Dame’s hasn’t had one of those 120 or 130 defensive rating days since before the Harangody injury.

I’m going to skip the other major injury to Wisconsin’s Jon Leuer because believe it or not, Wisconsin’s Pythag Winning percentage barely changed with him out. What is it about Wisconsin playing equivalently well with anybody in the lineup? I mean it is crazy enough they can lose star players every off-season and not miss a beat, but wouldn’t you think an in-season injury would be costly to them? Instead I want to consider another argument I’ve been making on this blog.

The Stats Don’t Hate Kentucky

I’ve said for a long time that Kentucky would be a national contender by March, but it would take awhile for their defense to come around.

Kentucky early in the year (22 games)
Adj Off Eff 116.5
Adj Def Eff 89.8
Pythag Win Perc = .9521

Kentucky late in the year (last 12 games)
Adj Off Eff 115.4
Adj Def Eff 85.1
Pythag Win Perc = .9707

Kentucky’s current numbers on kenpom.com
Adj Off Eff 115.5
Adj Def Eff 87.7
Pythag Win Perc = .9597

And in fact, Kentucky’s defense has recently reached a level where I can officially call them a national title contender. The .9707 Pythagorean winning percentage trails only Duke and Kansas nationally. (And Ohio St. with Evan Turner.)

One final trend that may be of interest: Everyone always talks about Duke swooning, but it obviously hasn’t happened this year. Duke’s Pythagorean winning percentage was only 0.0014 lower in the final 12 games of the season relative to the first 22 games.

But What About the Second Order Effects?

I enjoy these injury splits, but I’ve always hoped Ken Pomeroy would do this for us. And the reason is simple. When Purdue played poorly without Hummel on Saturday, it not only impacted Purdue’s offensive rating, it impacted Minnesota’s defensive rating. Kevin Pelton emphasized the exact same point in today’s tournament preview.

Ken’s argument against doing this has been a simple one. It is difficult to reasonably account for all the injuries and suspensions we see over the season. And of course Ken is right. Here’s a partial list of events that caused players to miss several games this year

Oklahoma - W. Warren – multiple causes
Arkansas - C. Fortson – suspended
Texas A&M - D. Roland - injured
Minnesota - A. Nolen – academically ineligible
Indiana - M. Creek - injured
Iowa St. - L. Staiger – left for Europe
Oklahoma St. - R. Penn - injured
Rutgers – G. Echenique - transferred
St. John’s – A. Mason Jr – injured
Kentucky – D. Liggins – academic eligibility question
Marquette – J. Maymon - transferred
Georgia Tech – I. Shumpert - injured
West Virginia – Ds Kilicli – eligible at mid-season
Cincinnati – I. Thomas – eligible at mid-season
UConn – A. Majok – eligible at mid-season
Iowa – A. Tucker – suspended and replaced by coach’s son
Kansas – B. Morningstar – suspended

There are many more, but I have to cut it off somewhere. And while you and I can point to many of these and say these aren’t critical, if you watch a team, you know all of these matter to some degree. I can’t tell you how many words Georgetown fans wrote about Nikita Mescheriakov replacing Omar Wattad in the Georgetown rotation last year.

Moreover, this is only a subset of the long-term injuries. What about the short-term events such as Kalin Lucas missing Michigan St.’s loss to Illinois or Chris Allen missing Michigan St.’s loss to Minnesota?

Or the complex situations such as California losing both Theo Robertson and Jorge Gutierrez? What about Villanova missing Reggie Redding and Mouphtaou Yarou? What about North Carolina missing Tyler Zeller, Ed Davis, Travis Wear, and Marcus Ginyard all at different points in the season. Suffice to say, if you start throwing out games based on every little thing that happens, you end up with a piecemeal set with which to rank teams.

And so Ken’s answer to my complaint is that he weights recent games more heavily to account for teams that are changing over time. And when I look at all these injuries, I can’t disagree. Unless a player isn’t coming back, like Robbie Hummel, the season long stats do a pretty good job. Ohio St. with Evan Turner looks pretty similar to what Ken has calculated. But I also think when you think about a deep tournament run for the elite teams, you have to factor these injuries in.

But is Duke Overrated?

Ken Pomeroy is humble. He readily admits that his rankings are not the only piece of information you should consider. I happen to think it is the best piece of information out there, but I’ll agree with him that some of the Basketball Prospectus log5 tournament odds make me scratch my head.

At the top of the list is Wisconsin being favored against Kentucky. I argued above why Kentucky is under-rated. And The Only Colors has recently argued why Wisconsin might be over-rated. The Badgers have been winning big against the weaker Big Ten teams.

And that is probably significant. Wisconsin’s disciplined system is perfect for crushing a young, turnover prone, cold-shooting Indiana team. But I’m not sure that tells us very much about how Wisconsin would fair against Kentucky.

But again, Ken Pomeroy doesn’t disagree that a few ratings seem fishy. Ken wrote today that Duke may be a little overrated in his system. And that’s my final question for today. Is Duke really overrated this year?

Ken Pomeroy’s system is not the only system that loves Duke this year. Sagarin’s Predictor has Duke as the 2nd best team in the nation behind only Kansas. The minimal difference is the control for pace. Ken’s numbers say that on a per-possession basis, Duke is a little better, but Sagarin’s numbers say that since Kansas plays at a faster pace, they get more possessions to prove they are a better team. Either way, the point is the same. Duke has fantastic margin-of-victory numbers and should be a national favorite. But if the margin-of-victory numbers are fantastic, how can we really argue that Duke is overrated?

You can argue about consistency. And there is some evidence that certain styles of play lead a team to be more inconsistent. But Mike Krzyzewski has actually worked very hard to make Duke’s games consistent by minimizing the number of three pointers his opponent’s take and make. (Only five teams allow fewer three point attempts and Duke has the third best three point percentage defense in the country.)

You can argue about players getting worn out, but Duke sure doesn’t seem worn out when they win the ACC tournament almost every year. Moreover, I haven’t seen any large scale study showing that teams with low bench minutes get worn out in the NCAA tournament. If anything, coaches use more timeouts in the NCAA tournament and players seem more well-rested.

And I’ve argued above that Duke is not swooning this year. They’ve completely maintained their early season success this year.

So what is the argument for why Duke is overrated? I don’t have one. All I can say is this. Duke is overrated because the other elite teams have upside, and Duke does not.

Think of personnel management. Earlier this year, a lot of people were complaining that Andre Dawkins wasn’t getting enough playing time to improve his skills. In a sense, they were arguing that Mike Krzyzewski was going all-out for wins today at the cost of wins in the future. But Krzyzewski’s response was pretty clear. That is what practice is for. Mike Krzyzewski’s philosophy is to play the best players on the court in every game. His goal is to maximize the chances of winning every day.

But other coaches have a different philosophy. Tom Izzo and Rick Pitino like to play a lot of players early in the year so that those players will get better. Sure Izzo and Pitino want their players to get better through practice, but they also believe there is something about game conditions that simply cannot be simulated. And so every year, you see Michigan St. and Louisville lose some games early that they should not lose because they are not afraid to let players fail.

The issue with Duke isn’t that they aren’t good. The issue is that this year, like every year, Duke is ready to play in November the way most teams are ready to play in February. Duke has achieved just about the highest Pomeroy ranking it could hope to achieve by playing its best lineup in every game.

But someone out there hasn’t peaked yet. Maybe the team with upside is Louisville. Louisville's ranking is low because of early season losses such as Western Carolina. But Louisville's two wins against Syracuse in the last month have them headed in the right direction. Or maybe the team that hasn’t peaked yet is a surprise. The time to find that surprise is almost here.