Wednesday, March 17, 2010

Beating up on the little guys

A few weeks ago John Gasaway looked at whether Wisconsin was over-rated for beating up on Indiana. He concluded that they were pretty efficient even without the Indiana games. Then the Michigan St blog "The Only Colors" looked at whether Wisconsin was over-rated for beating up on ALL the Big Ten’s weaker teams. And indeed, no one beat the Big Ten's lower half quite like the Badgers. But from a predictive perspective, are games against the weaker teams really that informative?

Let’s look at which teams have done the best and worst against NCAA tournament level competition. But since the goal is to focus on elite competition, I’m only going to look at teams seeded 1-13 in the NCAA tournament. Here are the records of the top seeds and the Adj Off Eff and Adj Def Eff against NCAA teams seeded 1-13. These splits recreate what is shown on kenpom.com.



Northern Iowa looks great, but that’s a sample size issue. They only played two games against NCAA tournament teams and they played their best two games of the year against Siena and Old Dominion. Northern Iowa was one of only two teams to break 105 points per 100 possessions against Siena, and they posted a 122.1 rating. Northern Iowa's big offensive day against a better Old Dominion defense was equally impressive.

Siena on the other hand struggled mightily against NCAA tournament competition, failing to win or even post decent numbers in their losses. Perhaps there is hope for Purdue after all.

Note that there’s no reason to expect the adjusted offensive efficiency or adjusted defensive efficiency to go down based on playing NCAA tournament teams. That’s the whole point of the adjustment. In fact Cornell went 0-2 against NCAA competition, but simply by playing Kansas tough, they actually posted some of their best adjusted ratings on the year.

The next table compares the current numbers on kenpom.com to the “vs NCAA tournament team” splits. The teams at the top of this list have fared better against NCAA competition. The teams at the bottom of the list have feasted on weaker opponents.



-First, I want to apologize to Wisconsin. Even though Wisconsin has fared worse against NCAA tournament competition, they are far from the worst offender.

-Teams like Tennessee and Missouri were able to use depth and full-court pressure to punish weaker teams, but saw these tactics were less effective against NCAA tournament teams. (The same would be true for Minnesota if it weren’t for two horrible performances against Michigan.)

-California was without Theo Robertson when they faced their biggest non-conference games which likely explains why they were better against weak teams.

-While Richmond had its best games against good teams, Xavier and Temple were much more likely to feast on the incredibly weak bottom-feeders of the A10.

-And like Wisconsin, Pittsburgh strikes me as a team whose discipline is simply over-whelming to the bad Big East teams. But Pittsburgh did not look nearly as sharp against the Big East elite.

-On the flip side, if the numbers were making you doubt Vanderbilt, maybe you should reconsider. Vanderbilt may not have beaten Kentucky, but they played them tough, and Vandy did quite well against Tennessee and Florida.

-Notre Dame’s late season dominance also corresponded with a lot of games against NCAA teams.

- Clearly Wofford couldn’t beat Southern Conference teams by enough to stand out, but they did better against NCAA competition. Wofford only has two games against tournament teams here, but they also beat Georgia and South Carolina. I don’t think Wofford is quite the prohibitive underdog some people think.

-I expected Georgetown’s numbers to be better against NCAA teams based on the big margins in wins against Duke and Villanova. But they also had some big margins in losses to Syracuse and Notre Dame. The Hoyas turned out to be equally good against NCAA competition and non-NCAA competition. Georgetown has also played the most games against NCAA teams with 17.

And for those of you who want to throw out margin-of-victory and strength-of-schedule and focus only on wins, here are the best and worst winning percentages against NCAA teams seeded 1-13 (minimum 8 games):

87.5% 7-1 Kentucky
85.7% 12-2 Kansas
75.0% 6-2 Vanderbilt
72.7% 8-3 New Mexico
69.2% 9-4 Syracuse

30.8% 4-9 Marquette
30.0% 3-7 Louisville
27.3% 3-8 Florida
25.0% 2-6 California
22.2% 2-7 Xavier

Some Stats Links

Final Four Odds

Even More Final Four Odds

I’m always amazed at the duplicity of good ideas. Crashing the Dance (CtD) posted road/home splits for Duke and Wisconsin. And then John Gasaway posted the road/home splits for all teams. I posted the Hummel odds on Sunday and CtD posted them in that same column on Tuesday . To whatever extent I “scooped” CtD by getting the Purdue numbers out on Sunday, that’s clearly not true. CtD developed net efficiency margin (NEM) and has used it account for injuries in the past.

But today I want to emphasize the value of both forms of analysis. First, I created the new adjusted offense, adjusted defense, and Pythagorean winning percentages so I could literally see how this would change the log5 odds. But the Crashing the Dance NEM methodology allows him to graph the “opponent and venue adjusted” trend which really allows us to see which teams are getting better and worse over time.

Second, there’s a minor technical difference in the approaches that I noticed when I was trying to match the CtD numbers. The difference is that Ken Pomeroy isn’t actually doing a linear comparison. His method divides by opponent's rating. So if in a year where the average team posts a 100 rating, his method would say that a 95 performance against a 90 defense is actually more important than a 110 performance against a 105 defense. 95*100/90>110*100/105. The linear approach is incredibly close in the range of data we are using, but this explains one reason my splits might look slightly different from the splits you see on the CtD website.