Monday, April 9, 2007

Will I Ever Get to the Point?

As I continue to build towards my coach ranking system, today I look at some information on recruiting classes. I have about three days worth of material. Then maybe I'll finally get around to rating the coaches late in the week or early next week.

Recruiting information is available from multiple web sites, but most places charge for this information. CSTV has a free searchable database that goes back to 1999. This information allows me to determine a recruit ranking for each player starting with the 2002-03 season. (I miss a few red-shirt seniors in 2002, but that’s pretty minor.)

Examining the 5 seasons for which I have recruiting data, it appears that recruiting is actually a pretty poor predictor of NCAA tournament success. (Think R-squared values around 0.25. There may be several ways to improve my model and I’ll discuss those in future posts.) To understand why recruiting does not predict the tournament very well, compare this year’s Ohio State team and this year’s Georgia Tech team. Both teams had a freshman that was a top 5 high school player (Greg Oden and Thaddeus Young). Both teams had a freshman McDonald’s All-American point guard (Mike Conley and Javaris Crittenton). But while Ohio State played in the national championship game, Georgia Tech did not win a single NCAA tournament game this year.

For simplicity I use 3 main recruiting classifications: McDonald’s All-Americans (MAA), Top100 Recruits, and Super Recruits (essentially Top 200). Notice that both Ohio St. and Georgia Tech have similar evaluations under this system. This suggests that perhaps I should add another classification for the truly elite college players (think Greg Oden and Carmelo Anthony). But prior to the NBA High School rule there weren’t very many of these truly elite players. (I also don’t remember Carmelo Anthony receiving Oden-level-hype out of college. It seemed more like Tyler Hansbrough-level-hype to me, but whatever.)

Regardless, at least as a first approximation I’m going to stick with these three groupings (MAA, T100, Super). The results suggest that High School talent predicts NCAA tournament success, but clearly other factors (coaching, luck) matter to a large degree. In fact, I could even hear an argument that recruiting is a proxy for coaching ability and that recruiting matters even less, but I’ll save that for another day.

Quick Facts
-McDonald’s All-Americans are the most important in predicting bids, tournament seeding, and tournament success.

-Over the last 5 years, the most important college players have been MAA Sophomores, followed closely by MAA Juniors, then MAA Seniors. It may seem odd that MAA Seniors are not the most important, but this is clearly explained by sample selection. The best McDonald’s All American’s head to the NBA after their Sophomore year, and the next best leave after their Junior year.

-Even as Freshman, MAA do a better job predicting tournament success than even Junior and Senior Top 100 recruits.

-For Top 100 players, the most important players in predicting NCAA tournament wins have been Top 100 Juniors. Again, because of players leaving for the NBA, Top 100 Seniors are less correlated with NCAA tournament success.

-Junior and Senior Super Recruits (Top 200) do a good job predicting NCAA tournament seeding, but do a poor job predicting NCAA tournament wins. This sort of fits the cliché that it takes talent to advance in the NCAA tournament.

2006-07 Talent
Without going into all the details of my recruiting model, let me post one check as to whether I am somewhere in the right ballpark. Based only on the recruiting rank of players on this year’s rosters, here is what the top 25 would have looked like at the start of the year.

1 Kansas
2 North Carolina
3 Duke
4 Ohio St.
5 UCLA
6 Texas
7 Kentucky
8 Louisville
9 Georgia Tech
10 Arizona
11 Connecticut – was in preseason Top 25
12 LSU – was in preseason Top 25
13 Michigan State
14 Alabama – was in preseason Top 25
15 Florida – see below
16 Syracuse
17 Florida State
18 Washington – was in preseason Top 25
19 California
20 Oregon
21 Oklahoma State
22 Notre Dame
23 North Carolina State
24 Villanova
25 Indiana

As has been mentioned previously, Florida’s players were all solid recruits, but not in the truly elite class. On paper, they looked pretty similar to the players Syracuse recruited. The difference is how that talent developed, and how they grew together as a team. Notice that Georgetown’s players were not highly recruited and yet they still made the Final Four. While Roy Hibbert entered college as “the Big Stiff” and not a top 100 recruit, he improved his game and became one of the most important players in the tournament this year. Other highly regarded tournament teams such as Texas A&M and Memphis had tremendous years even without a single McDonald’s All-American.

Recruiting Predictions
Based on talent on-hand, Kansas would have been predicted to win 3.23 tournament games this year. If you recall my post from last Thursday, that was just less than the expected wins for a one seed in the NCAA tournament.

One way to interpret these recruiting rankings is that given the talent on-hand, Kansas essentially started the season with the same odds of tournament success as a one seed. When you consider everything that can go wrong over the course of a season, that’s a pretty amazing high ranking. Because there are many similarly talented teams in the middle of the pack, the predicted wins (based on talent) falls off quickly.

Tommorrow I will look at the teams that out-performed their recruited talent level, and Wednesday I will look at the teams that under-performed based on their recruited talent level.

Email Me
There are a million ways to run these specifications, but if you strongly feel I’m missing something important, email me. I’m re-opening my gmail account for blog traffic. Email me at dlhanner@gmail.com if you have any comments. That second letter is an “L” and not a one.