[Husker] A case for recruiting rankings

[Duane Feldman]

SHORT VERSION: there is a correlation between rankings and recruiting, but it doesn't help predict future winners.

LONG VERSION:  OK, I've said it before and you are probably all tired of hearing it, but . . . .  I went thru a similar correlation analysis back when SI posted a study (I think '99).  Same results -- a high correlation between winning teams and high recruiting rankings was the SI conclusion.

I looked more closely at the results and concluded that the correlation was exactly the opposite:  Highly ranked teams got the highest ranked recruiting classes!  The cause and effect were reversed!

My analysis showed that when you did a four or five year lag between rankings and recruiting class rankings, the correlation dropped to near zero -- except in those cases where the team was a perennial top ten contender.  More importantly, the recruiting rankings ALMOST NEVER PREDICTED A CHANGE when a team moved dramatically in the rankings (a negative correlation).  Moreover, when a team climbed, so did the recruiting class rankings and vice versa.

Nebraska's '92 and '93 recruiting classes were ranked in the 30's but those classes went on to produced back to back NC's.  This year, Kansas, Missouri, WVU, Cincinnati, UConn, Hawaii, and others were NOT predictable winners when looking at four year prior recruiting data.  Last the same was true for Rutgers, Louisville, Wake Forest and Boise State last year.

On the other side of the equation, FSU, Miami, and ND all had highly ranked recruiting classes prior to their respective decline.  No way should NU have had a 2007 losing record based on prior recruiting class rankings (for Solich or for Callahan).

I don't care what the level of correlation, if a data stream cannot be used for prediction purposes, any correlation is pretty much useless.  At the same time, I don't know any value of knowing that a rise in rankings will create a rise in recruiting class rankings.  Nebraska is a prime example:  We expected a good year in 2007 and going into the year our class was being ranked top five.  After a dismal season, our recruits jumped ship and the recruiting class rankings dropped correspondingly.  I predict that if we end up ranked next year, our recruiting class rankings will improve over the 2008 class (I know I'm not going out on a limb, but it proves my point).

In short, there is a correlation between rankings and recruiting, but it doesn't help much.

Duane Feldman


>I'm not sure about the overachiever aspect (seems pretty subjective 
>at best), but adding a stagger would sure seem to make the method 
>more appropriate.

>Ted


> So... in order to make it work, you'd stagger the recruiting years 
> and the winning percentages, include more overachievers, and 
> that'd be okay?
>
>
> On Jan 23, 2008, at 8:16 AM, Mike Jaixen wrote:
>
>> It's so flawed it makes the whole article a meaningless exercise. 
>> 2007 recruits made minimal impact on those numbers, and 2002 
>> results on the field had very little to do with these recruits.
>> 
>> In addition, there's a certain chicken vs. the egg argument to be 
>> made here.  If you were to throw out a list of teams that were 
>> good in 2002, chances are that you would have included teams like 
>> Oklahoma, LSU, Florida, Michigan, and USC.  I think if I matched 
>> the 2002 ratings up against the 2007 ratings, I'd see nearly the 
>> same results of this survey.
>> 
>> Especially when the author decides to exclude all of the examples 
>> that break his hypothesis.  (UConn, Hawai'i, and Boise State.)
>> 
>>
>> --- Theodore Heise <theo at heise.nu> wrote:
>>> On Tue, 22 Jan 2008, jon johnston wrote:
>>> 
>>>> http://www.sundaymorningqb.com/story/2008/1/21/1614/43228
>>>> 
>>>> The guy that writes this site, Matt Hinton, is very very good. 
>>>> Here he makes a case for the recruiters' rankings and does a 
>>>> pretty decent job of it.
>>> 
>>> Not bad, but using recruiting rankings and BCS winning 
>>> percentage from the same time period seems like a major flaw in 
>>> the method.

_______________________________________________
husker site list
husker at tssi.com
http://romaine.tssi.com/mailman/listinfo/husker

Note:  Comcast has been bouncing a lot of list traffic recently.  The problem appears to be on their end.