Thursday, December 16, 2004

It was my understanding there would be no math | by Jay

(Timeout now for a message from BGS friend and stat-head Jeff, who sends us this tasty mathematical confection).

Ara Parseghian once said that if he couldn’t win at ND in three years, it wasn’t going to happen at all. Ara was right.

I looked at the ND coaching records going back to Jesse Harper, and I found that after three seasons, about half the coaches improve their record during the next two seasons, and about half slide. However, there is a very high correlation (87%) between their records for the first three years at ND and their next two.

first three next two
Jesse Harper .870 .853
Knute Rockne .917 .881
Elmer Layden .741 .806
Frank Leahy .850 .972
Terry Brennan .633 .650
Joe Kuharich .400 .500
Ara Parseghian .867 .775
Dan Devine .800 .696
Gerry Faust .544 .522
Lou Holtz .714 .840
Bob Davie .568 .609
Ty Willingham .583

Correlation 87%

For you Arts & Letters types like Jay who are scratching your heads right about now, Correlation is a common function used by statisticians, and is a statistical technique to show how numbers are related. Numbers with -100% correlation are inversely correlated (when one goes up, the other goes down). Numbers with 100% correlation are basically functions of each other (they move in tandem). And numbers with 0% correlation have no bearing on each other (they are random).

So in our example, an 87% correlation means that your next two years are going to look fairly similar to your first three.

Looking at the table, the two ND coaches who improved most over years 4 & 5 were Lou Holtz and Frank Leahy, although there were some anomalies with these two: Holtz quickly improved on a very poor first year and Leahy’s fourth year occurred two years after his third year. Yet even these two only added an average of 1.5 victories per year during years 4 & 5.

Perhaps not surprisingly, the 7-5 that Willingham averaged while at ND was nearly identical to the 44-36-1 (6.6 wins and 5.4 losses in a 12 game season) he averaged over 7 years at Stanford. If Willingham had stayed on for years 4 and 5 and matched the performance improvement of Holtz or Leahy -- an absolute best-case scenario -- he would improve from his current 7-5 to one year of 8-4 and one of 9-3. What's more likely would have been more of the same: another couple of years right around 7-5.

So, what does this tell us about Charlie Weis? Well, not much, yet. But his first three years should show us with reasonable certainty (okay, 87% certainty) if he's worth keeping around for another two.

-- Jeff