College Football Computer Rankings and Strength of Schedule
So here is a conundrum…if college football teams make their schedules years in advance, how can they always ensure they have what the computers will call a “Strong schedule?” Answer: it is impossible. Fresno State beat Illinois this past year. Illinois went to the Rose Bowl very likely at the same time those two teams agreed to play. Turns out Fresno State did not get any boost in the strength of schedule column for that win. Oklahoma lost to Miami (FL) this year. When they made that schedule, Miami (FL) was headed nowhere and not playing very well. Probably looked like an easy game when they made it. So why does schedule strength mean anything to the computer rankings and in the minds of human voters? Good question.
My new computer ranking system will have absolutely nothing to do with the strength of your schedule. This is a totally bogus statistic if you ask me. Bosie State can’t help being in the WAC. Do you think the Pac-10 is at all interested in letting them in? The Big 12? The Big Ten? Of course not, there is no major BCS conference looking to add Boise State. Maybe the MWC would take them, but who really knows. It’s not like schools can just pick their conference and thus pick 2/3 of their schedule. Schools really only have power of 1/3 of their schedule. That’s where you see who is willing to play real teams. You have Ole Miss last year scheduling FCS teams and perennial bottom dwellers of CUSA and the Sun Belt. Does playing in the SEC make up for it?
To me it doesn’t. You should not be rewarded for playing in a tougher conference and you should not be rewarded for scheduling cupcake non-conference games. And so my new statistic completely takes this into account. It is called a Chi Square statistic. It is quite simple yet very effective. Allow me to explain how it will work. In every game played, you can easily keep track of points scored and points allowed. You can also keep track of yards gained and yards allowed. After 6 games, you can start to see some trends among all the teams.
By viewing the trends, you can start to come up with expected outcomes for games. For example, if Boise State is averaging 1.5 times the points their opponents give up on average and allowing 65% of their opponents average points to be allowed, then if they play a team that normally gives up 30 points and normally scores 18 points, then you would expect the score to be 45-12. Same goes for yards per game and all of that.
Let’s suppose two scenarios. First, let’s suppose Boise State wins 63-0. They clearly exceeded expectations by 18 points and held their opponent to 12 fewer points than you would expect. What the Chi Square does is look at those differences and tells you if they are statistically significant. We’ll assume they are. Then this counts as a positive offensive and positive defensive victory.
For the second scenario, let’s say Boise State wins 30-28. They scored 15 points less than you would expect and gave up 16 more points than you would expect. This would count as a negative offensive and negative defensive victory. You can also get a positive offensive and negative victory, or a negative offensive and positive defensive victory.
Worst case scenario is if Boise State were to lose. Then they lost a game they were expected to win, which docks major points. But look at Boise State in the Fiesta Bowl last year. According to my numbers, they were expected to lose 27-20 to TCU in that game. But they won a game they were expected to lose. That adds major points as well. This same test can be run for every team taking their averages and the averages of their opponents. And some are just plain obvious, like Texas and Nebraska. Clearly you would expect Texas to do better than 13-12 in that game.
So now you have strength of schedule completely removed from the equation and no one’s schedule is held against them. If you say someone has an easy schedule, great, then you would expect much more from them in their games. Did they meet expectations? Exceed them? Fall short? Likewise, if you have what others term a strong schedule, then you aren’t expected to clobber the teams you play each week. Either way, your expectations are different. Did you meet them, exceed them or fall short of them? Now it doesn’t matter what your schedule is, it matters how you play the teams on your schedule.
But Sloppy, what about running up the score? Good question, I don’t know how to account for it. If you did run up the score and got 24 more points than expected, you would likely get a huge point gain for that. We obviously have to think of a way to get around that. What are your suggestions? I need some help on how to combat the running up of the score so that sportsmanship is maintained. Please leave your ideas in the comments so I can perfect my computer rankings.
PS- I ran my stats on Boise State and Alabama. Alabama ended with a +4 and Boise State ended with a -14. So Boise on average didn’t quite meet expectations and Alabama on average barely exceeded them. I’ll finish the formula and plug in everyone else later.
Comments
while i understand what it is you’re doing, not sure i agree in principle. “quality of win” argument is kind of hard to justify. the only statistic that should matter is the one on the scoreboard that shows team A has scored more than team B. regardless of what the differential is, that’s what should be focused on. but, i do agree, who those teams are can be hard to justify.
Grindhouse, the quality of win principle is paramount to computer rankings. That’s how an 8-5 team ends up ranked higher than a 10-3 team at the end of the season, high quality wins. I’m saying quality of win shouldn’t be determined by who is on your schedule that year. It should be how your play your schedule. Cupcake schedules should result in frequent lopsided victories. Thus, squeaking by bad teams and crushing good teams have their merits. This way teams with bad schedules still get recognition for playing beyond how they were expected to with such an easy schedule.