Saturday, March 6, 2010

More Surprises and Flops

Today, I share the Pac-10 numbers that I neglected in my last post. And the nice thing about a numeric predictive model is that it allows me to evaluate other leagues that may not have had pre-season predictions in every publication. In addition to the Pac-10, I’m also going to look at the surprises in the A-10 and MWC today.

The following table lists adjusted offensive and defensive efficiency from, and the predicted values from my model for 2010. Conferences are sorted by predicted efficiency margin. Green indicates positive surprises, red indicates negative surprises.

California’s numbers are actually ahead of what we would have expected this year, as the defense has improved a little more than expected. Unfortunately, the rest of the Pac-10 has been so bad that California hasn’t had the opportunities for signature wins.

Arizona St. is almost exactly where expected, but they’ve moved up in the standings thanks to the drop-off in defense for Washington and UCLA.

UCLA’s offense was expected to be worse this year. The only teams that lost more offensive firepower were North Carolina and Pittsburgh. But the drop in UCLA’s adjusted offensive efficiency is even worse than the model predicted. It appears Nikola Dragovic only played well last year because his teammates got him wide-open shots. His previous success as a lethal three point shooter has not carried over to this season.

Oregon St.’s drop-off is even more puzzling because this is basically the same cast of characters as last year. The offense was supposed to improve slightly but Seth Tarver, Roeland Shaftenaar, and Calvin Haynes all have a worse eFG% percentage than last season.

I should come up with a name for what has happened to Oregon St. and USC this year. Call it the desperation factor. When you realize your team is terrible at scoring, sometimes you have to commit to becoming a lock-down defensive team.

The value of a model like this is really the middle and bottom of a less-televised league. This summer, I listened to a program where Andy Katz interviewed coaches from various leagues including the A-10, and they universally said La Salle was going to be a sleeper team this year. But La Salle has been no sleeper - they’ve been asleep in A-10 play. So what went wrong? The defense fell off the map. It stopped forcing turnovers and forcing missed shots.

Dayton has actually been as good offensively as expected and better defensively than was expected. So why aren’t they in second in the A10? There are two answers. First, several other A10 teams have performed better than expected. Second, luck or chance has not been kind to the Flyers.

With Dionte Christmas graduating last year, Temple obviously rededicated themselves to playing defense this season. And while the offense is actually a little worse than expected, the improvement in defense has been enough to keep the Owls at the top of the A-10 standings.

But why did Richmond suddenly become a lock-down defensive team with an elite level of three point defense? What caused that pleasant surprise?

With Duquesne, can I emphasize the Melquan Bolding effect? A prolific freshman last year, Bolding was one of the key reasons for optimism this year, but he fractured his wrist and missed the start of the season, and he has not been able to regain his shooting touch.

Call the Mountain West the conference of offensive surprises. I’d argue the biggest was probably the improvement of Tre’Von Willis at UNLV. He’s improved his eFG% from 43.9% to 53.8% and since he’s always been a high volume shooter, that improvement was critical to UNLV becoming an NCAA contender. But New Mexico’s Dairese Gary has had an equally vital improvement. Gary cut his turnover rate from 25.3% to 15.6% and became the confident point-guard of the Lobos. And if you haven’t heard of BYU’s Jimmer Fredette, whose offensive efficiency also jumped 10 points, you just haven’t been paying attention.

One comment on the results in the last post, I’m even more convinced that I need to include height in the model in the future. For example, my model says that UConn’s drop in defense this year is a surprise, but is it? Shouldn’t we have expected a big drop-off without a 7’3” Hasheem Thabeet manning the middle? And while I can’t account for a lock-down defensive player like Travis Walton, maybe I should also play with steal rate and block rate. I’m a little hesitant to use these numbers, because sometimes bad defenders go for the steal, and good defenders simply play for the missed shot. But there may be some value in the available defensive statistics. That sounds like a summer project.