September 13, 2004
Fanblogs Interview with Kenneth Massey
In the world of computer polls there are a couple of guys that seem to stand head & shoulders above the rest. Ken Massey has built one of the most accurate poll systems out there. We take a few mintures with Ken to review the Massey College Football Ratings.
Your rating system has certainly proven itself to be exceptional accurate in forecasting games, yet some people say that college sports is always "up in the air" when it comes to predictions. Is college football easier/harder than professional sports to predict? Is the home field advantage more valuable in college than it is in the pros?
The HFA [home field advantage] is about the same (3-3.5) points for college and pro. The NFL has much more parity than college football, so it's harder to predict. About 25% of all college games are 'upsets,' and because of the emotional nature of the game, no single result should shock anyone.
There seems to be a fair degree of variance among college computer polls as items are weighted differently. Other than your ratings, whose numbers do you trust when it comes to trying to rate college teams?
There are a lot of good ones. I grew up following Sagarin, and think he does a great job. Peter Wolfe, who is also part of the BCS, has another system I like.
The BCS is set for yet another new season & new formula. What are your thoughts on accuracy of the human polls? Bias?
You can't get around the bias, even if it is sub-conscious to the human voters. Human polls have short memories, and fairly 'sticky' in the sense that they don't react well to new information. The new BCS formula uses poll votes, not simply ordinal positions, so it will be harder for a few voters to swing the whole outcome.
Margin of victory has gone away as a BCS component. Outside of the debate of whether MOV is "good for college sports" or not, is margin a good and fair factor in ratings?
If you want to predict games, yes, it's essential. For the BCS purposes, it is a noble ideal to remove MOV, but it's hypocritical, since the polls certainly take margin into account. There are simply too few games each year to throw away that much information. Who can argue that 35-17 is no better than 21-20 ?
What sports do you enjoy as a fan? Any rivalries that you enjoy?
College and pro football, college basketball, tennis. I'm a huge Virginia Tech Hokie fan, and I love to beat Miami and UVA.
When a casual fan looks at two teams in the Massey Ratings (your version, not your stripped BCS version), how does he compare two teams in forecasting games?
Simply subtract the 'Power' ratings to come up with a predicted margin of victory on a neutral field. The HFA and offense/defense numbers can help you refine the prediction to come up with an actual score.
Last year you had one of the more accurate polls/ratings in the country. Ever picked the National Champ in the preseason? And, who do you like this year?
My preseason ratings are based on previous years' results, and they don't consider returning starters, etc. Personally, I like Georgia or FSU this year, but their schedules could be tough. USC will have a cake-walk schedule, so they're a safe bet.
