Tuesday, March 01, 2011

Bracket - Math

This year I know almost nothing about college basketabll. I knew the name of that guy on Duke the beginning of the year, and I saw a cover that said Duke and Kansas are good. That's it. So it's a good year to test the idea I had last year, of filling out a bracket based on math instead of knowledge.

I read in 2009 that margin of victory is a telling stat, and more specifically, that blowout wins vs. not-getting-blown-out losses is the way to bet. This fits what I've said before, that the cliche "great teams win close games" is more mythology than truth. Great teams win blowouts. A close game can really be seen as only half a win in terms of its predictive value. So I'm going to start there. I don't know how I'm going to define blowout or close game yet, though.

Actually, I'm going to start by picking all my #1 and #2 seeds over the #16 and #15 seeds. But that should have been obvious anyway.

I have to figure that strength of schedule means something, and I'll work it in somewhere, probably with a significant weighting, but I haven't figured that out either. Last, there is the intriguing notion that winning home games is important, but that games are sometimes not 100%-0% home or away. The thinking there is while Louisville playing at Kentucky's gym is a significant disadvantage - the crowd noise and court appearance - it's still a bit of a home game. It's not far, some of your fans will make the trip, and you've seen the place before. Call it 10-15% a home game. If you are playing at a neutral site that is in your league and you've played there before, it might be 55-60% a home game even it it is far away.

I really don't know how I'm going to calculate that last one. I may just give up. An either-or Home/Away might not make that much difference, because I don't think I'm going to weight that factor as strongly as the other two. I will be doing the calculations by hand, just because it's fun. I'm sure there are sites that do all this for you, far better than I'm going to do, but don't tell me about them, for I would be sorely tempted.

Just running the calculations I will likely learn things about teams that will start to influence my decisions, but I'll try to make it as clean as possible.

Update: I think I had better check whether teams have a major injury or not. Or is that cheating?

Update: First calculations, immediate problem. The first few games of the season are often not against DI teams. I don't know which way I'll jump on that.


Sam L. said...

Let us know how much time you spend on your calculations.

Wyman said...

Also, keep in mind that BYU has suspended one of their best players for the rest of the season.

Wyman said...

I may take your bracket and adjust it via what college basketball knowledge I have. I have a suspicion that the best brackets are X% math and Y% knowledge about team personalities and styles and chemistry and what have you. I just don't know what the percentages are.

Assistant Village Idiot said...

Well, seeing that I now have some actual knowledge about one team, I suppose I should go for the injury piece somehow.

Sam, I don't think the calculations are going to be much, just the setting up. I have decided that both losses and victories under five points will be counted at .5 win, blowout wins or losses will be 15 points or more and will count double.

Strength of schedule may just be a multiplier based on what league they are in, or I may just pull the RPI's off a website and factor that in.

I haven't figured out how to research injuries without learning stuff about the team. I have decided I should start doing the calculations for some teams I know will be in the tournament. As I consider rankings 21-40 fairly arbitrary anyway, those aren't likely to influence me.