FINNNNNNALLY it’s game week. Saturday cannot come soon enough. While you spend your week watching FIDM videos, reading press conference transcripts and fidgeting uncontrollably in anticipation of the 2013 season kickoff this weekend, I hope a little statistical analysis of the 2012 campaign will help keep your mind off that agonizingly long wait until 3:30pm ET on Saturday.
Last year I set out to do an analysis of what 5 commonly used statistics were most predictive of overall team success (success is measured by wins, of course). I found these “Final Five Stats” to be offensive passing yards, offensive rushing yards, passing yards allowed, rushing yards allowed and turnover margin. Using multiple linear regression, I measured some other numbers that are frequently discussed including time of possession, 3rd down conversion percentage and penalty yards per game but found that each did very little in improving the predictive power of the Final Five Stats.
For the 2012 season, these Final Five Stats effectively account for 76% of the variation in regular season wins among FBS teams (this number is slightly up from 74% in 2011). In terms of the explanatory power of the model, over half of FBS teams finished within 1 win of the model’s expected value and 104 of the 124 teams finished within 2 wins of the model’s expected value.
Similarly to 2011, offensive passing yards and offensive rushing yards were effectively equal in their impact on wins. On the defensive side of the ball, however, rushing yards given up proved to be much more detrimental than passing yards after the opposite had been true in 2011. Finally, on the whole turnovers also played a more sizeable role in 2012 win totals (1.68 wins per net positive turnover per game) than they did in 2011 win totals (1.40 wins per net positive turnover per game).
The Results
Before we get to the teams that performed the most outside of expectations during the season, let’s kick things off with the Irish. Last year I raised some eyebrows when the model showed an expected 6 wins for ND for the 2011 season (compared to the actual 8 wins) and I suggested the Irish may have actually overachieved despite the fact most of us fans spent the offseason discussing the significant underachievement of a team that had some of the worst luck in modern memory. For some more clarity this year, I want to state that the Final Five Stats in this model, while very important, only explain about 75% of any given team’s win total. There are an uncountable number of factors that affect the outcome of any individual game and the game’s unpredictable nature is one of the reasons we love it so much.
And with that it’s time to take a look at the 2012 Irish regular season. According to the Final Five Stats, we should have expected ND to win 9.2 games last year. Relative to the actual UNDEFEATED regular season…can’t say that enough… this approximately 9 win total suggests the Irish performed outside of the bounds you’d expect from their basic offensive and defensive statistics. When you consider that ND was not particularly dominant against Purdue, Michigan, Stanford, BYU, and Pittsburgh the expected win total starts to make a little more sense. Of course what the Final Five Stats do not account for are perfectly timed passes from Tommy Rees to Robby Toma against Purdue, Rees rushing for a touchdown (!!!!) against Michigan, the goal line stand against Stanford, Theo Riddick bursting from the pile without his knee touching against BYU or the Golson Superman 2-point conversion against Pitt. While the model’s expected wins didn’t quite match up with the actual number, ND still ranked 7th overall in expected wins.
[table “” not found /]For the second year in a row, the Irish also landed in the top 10 notable outperformers. Looking at the quality of the teams on the list below it is apparent that the teams that ended up at the top of the rankings at the end of the year won more games than simple statistics would suggest you should expect. I am not certain I am comfortable enough with the data at this point to claim this is evidence that better teams find ways to win even when the numbers are against them, but it certainly points in that direction. At the same time, maybe it is a good bounce here or there that leads to an extra win that makes us believe some teams are better than they are in reality.
Notable Overachievers
[table “” not found /]Notable Underachievers
[table “” not found /]These statistical measures are far from perfect but they do paint an interesting historical picture. A pertinent next step in this analysis (that I unfortunately didn’t have the time to do at this point) is to adjust for strength of schedule. Looking at last year’s numbers, while Oregon’s 12.0 win expectation is phenomenal, it was aided by some absurd numbers against teams like Tennessee Tech and Arkansas State. Don’t even get me started on Alabama’s games versus Western Kentucky, Florida Atlantic and Western Carolina. Good job, good effort.
To our readers: do you think good teams are more likely to win games they shouldn’t, or is our conception of “good” sometimes based on lucky outcomes? Of the Final Five Stats, where do you expect improvement for the Irish in 2013? Do any of the teams that show up as the top overachievers or underachievers stick out to you?
- Chicks Dig the Long Ball: Irish Receivers ’13 - October 23, 2013
- Breaking Down the Run/Pass Balance - September 26, 2013
- 3 Questions for MSU - September 19, 2013
NDMD62
Tremendous insight and analysis!
NDtex
The Ohio State number is very interesting. I would hope for the other shoe to drop on Meyer and company this season, but with the schedule they have, I doubt it.
trey
A nice little matchup in the MNC game against the Irish would help ease the pain of that, a little.
DenverIrish
I see Weis’s teams are still fitting expectedly into a “regression analysis” model.