Recruiting Wins and Losses: How Desirable is Your School?

Even though this week is Hat Week (Go 'Cats! Get the Hat back!), I thought I would share a pet project of mine.  I'm not a big fan of recruiting rankings.  I prefer wins and losses.  So to me, the best way to talk about recruiting is to look at recruiting records, i.e. given a recruit with offers from several schools, where do they go?

Here's how it works: suppose a recruit has offers from Illinois, Indiana, and Iowa (chosen alphabetically).  If that recruit ends up at Iowa, then Illinois and Indiana each receive a "loss" to Iowa.  Appropriately, Iowa also earns one win over each of those teams.  A school can only get one loss on a recruit, but they may get several wins, one for each competing offer.  

Here are the head-to-head records of Big Ten teams (and Notre Dame) over the last 5 years (the players currently on the roster).  

A few observations:

Northwestern and Wisconsin are an interesting pair.  Against "lower tier" teams like Illinois and Purdue, Northwestern has a stellar 21-7 record, compared to Wisconsin's more pedestrian 23-23.  However, against Michigan, OSU, Notre Dame, and Penn State, Northwestern has a dismal 3-41 result.  Wisconsin doesn't perform spectacularly against that group (21-83), but they are at least competitive.  

Ohio State has a winning record against every other team, which is unsurprising given their coaching stability and in-state talent base.  Conversely, Indiana has a losing record against every team.  

The two teams that have gone head-to-head with Nebraska the most are Iowa (not surprising) and Michigan (surprising to me at least).  

I don't know why Iowa has gone up against MSU so many times.  

Just like computer rankings can take win/losses record and determine an overall rating, they can also work on this data as well.  I applied the Elo system, but other systems would work too.  (My beloved transitive rankings don't handle multiple "games" between the same two opponents.  They could be fixed, but that would be another project.)

Brief, skippable description of finding Elo System ratings:

The Elo system picks one team (e.g. Northwestern).  It assumes that all other current ratings are correct, and then tries to find the rating for Northwestern that was most likely to have resulted in reality (i.e. the rating that gives the highest probability to the actual head-to-head win/loss record of Northwestern against each team).  It then assigns this new rating to Northwestern.  Now it assumes that Northwestern's rating is correct and works on another team's ranking.  It repeats that process until every team doesn't change anymore.

End of description. 

The Elo system ratings give the probability of a recruit picking one school over another, and it's logarithmic.  I chose 2 as the base, so if Team A's rating is 1 point higher than Team B's, a recruit with an offer from both schools is twice as likely to pick Team A than Team B.  

For example, Iowa has a rating of 19.04 and Michigan has a rating of 20.05.  If a recruit has an offer from both schools, he is 2^(20.05-19.04)~=2.01 times as likely to pick Michigan than Iowa.  The actual result was 20-8, but the system found the best balance between their direct record and each of their records against the rest of the Big Ten.  If the recruit has an offer from Indiana and Ohio State, he is 2^(21.85-16.35)~=45 times more likely to pick Ohio State.  (Actual result was 12-1; I don't know the story of Shane Wynn but I'm sure it's interesting.)

These ratings don't differ much from expectations.  Ohio State has the best recruiting territory, the most stable coaching situation and the best history of success over this time period, so it's no surprise that they have the highest rating.  I expected Indiana at the bottom, but I was surprised just how much separates them from the others.  Purdue is almost 3 times more likely to get a recruit, and everyone else is even higher.  

Wisconsin seems a bit low, but that's primarily due to their weak record against the bottom half that I talked about earlier.  I expected Nebraska to be higher.  Their low rating comes from two causes: first, they have relatively little in-state talent, so they are always fighting uphill.  Second, this period includes the 2007 and 2008 classes, which were seriously damaged by the conclusion of the Callahan era.  Counting only 07-08, they would be 9th.  In just the last two classes (10-11), they would be 5th.  

There is a large gap between Penn State at 3rd and Michigan State at 5th (1.98~=4 times as likely to get a recruit).  I'm guessing that in "normal times" Michigan and Nebraska would be closer to the top group, but both of them have suffered at least one messy coaching change in this time period.

I also believe that Michigan's loss has been MSU's gain.  

One important point to remember is that these ratings are not comparing the quality of recruiting classes.  They instead are a "get who you want" rating.  These two are related, but not quite the same.  For instance, a school might decide to only give football scholarship offers to children of alumni.  This school would likely have a high rating in my chart (as they would likely have a high rate of success, regardless of other offers), but the quality of the recruiting class might be very low.  

Because I'm a Northwestern fan and private schools face a different set of challenges, here are the same charts for the FBS private schools that graduate at least 75% of players.  

The ratings:

These ratings aren't on the same scale, so they can't be compared directly to the earlier Big Ten rating.  Interestingly, Northwestern does a little bit better relative to Notre Dame on recruits that end up at private schools than recruits that end up at Big Ten schools.  

Rice is surprisingly competitive in this ranking.  

Notre Dame has an overwhelming advantage here.  They're about 2.8 times more likely to get a recruit than Miami and almost 5 times more likely than Boston College.  

Here is the chart for all BCS teams:

I don't know why fans were worried that Texas would gain a recruiting advantage from the LHN.  They already get almost every recruit they offer without it.  Oklahoma was 8-49 against Texas and they got more recruits with Texas offers than anyone else.  Only 52 recruits have gone to another school with a Texas offer.  The next best for fewest rejections is Ohio State with 193.  

If Texas and Washington State both gave an offer to the same recruit, he would be almost 100 times more likely to pick Texas. Put that way, the system seems to overestimate WSU's chances. 

Nebraska's 07-08 troubles appear even more starkly here.  If just those years were considered, Nebraska would be 45th.  

I didn't think that Cal would be so high (15).  They are the flagship university of a very talented state, but all the teams near them in the ranking have either been good or at least hyped to be good.  

The next step in this project will take these ratings and then determine the most desirable recruits (I'm open to any advice on how to do this, in particular how to balance someone with a large number of offers vs someone with just a few exceptional offers vs someone with a Texas offer).  I might then take those recruit ratings to grade each school's recruiting, but I think those would be the same as these ratings.  

(Note: All offer information is from Rivals.  I would like to thank them for IP locking me only twice, at least one of which was deserved.  To see all head-to-head information for all 120 FBS teams in the last 5 years, click here.)