Chaos Rankings

Monday, August 07, 2006

The moment you've not been waiting for

The inherent problem with any ranking system in the early stages of the season is that there isn't a lot of data to go on. How often do the preseason favorites end up still at the top of the standings? How often does an early pick lose their first couple of games, then quietly go 10-2 or 9-3 and wind up in a great bowl game? My point is that it takes a couple of weeks for things to settle down and for anyone to get a good feeling about any team.
The same is true of my rankings of course. Perhaps even more exagerrated because of the heavy emphasis (50%) on winning. What this means of course is that after Week 1, the team who will eventually finish the season in the Top 5 could very well be #58. This isn't "wrong" per se, since so far they have yet to prove they're a good team. Let's say that on opening weekend, Miami loses to FSU (I know I know, wide [insert direction], but hey, anything's possible). Let's also say that both teams rack up a lot of points and yards in the process. The SOS component for Miami will be pretty high, but they'll still be in the bottom half of the rankings because of the loss. If then Miami goes undefeated, you will see them climb in the rankings each week until they regain a position that seems more fitting for a team of their talent. Of course, given Coker's loss of momentum in recent years, maybe this was a bad example, but we'll wait for the season to see who's who and what's what.
To summarize, because these rankings are based on a formula, it is very likely that a good team will be ranked poorly early in the season if they drop a decision, but assuming they live up to expectations, not to worry, they will climb right back into things. Also note that it is not impossible for a team with otherwise stellar stats but two loses to climb above a team with 1 loss. VT did it last year over West Virginia.
Other points to note: statistics piled up against non Div. I-A opponents count, but the games are entered as 0s in the SOS calulation. This probably isn't enough of a penalty given that when a team like Texas demolishes a team like Sam Houston State, they might pile up 70 points and 500 yards which may offset the SOS penalty. I haven't done the math yet to determine if this is the case, but it feels about right. Strength of Schedule is based only on games played at the time the rankings come out, and accordingly will change every week based on games played. So the team with the strongest SOS at the end of Week 1 may end up with the weakest schedule at the end of the season.
I think those are all the concerns I have with my ranking system up to this point. I'm sure, just like the BCS, that more will emerge throughout the year and I'll end up adjusting my formula for next year.
Also, right now, my calculations are all done in Excel, which isn't inconvenient per se, but it's just not all that pretty. Ideally, I'd like to have a little program that I can enter in stats from each game and it will do all the backend work for me. I started working on this at the end of last season, but promptly gave up after many many compile errors. My coding skills are substantially rustier than I thought... In any case, my eventual plan is to have a little GUI that allows you to enter data, shows rankings, can import and export data in CSV format, and perhaps has sliders on it that can be adjusted and will automatically re-rank the teams based on the new formula parameters. I'm going to try to write this myself one mroe time while I'm on vacation at the beach next week, but if any CS kids out there want a relatively simple project to work on for free (okay, I'll give you a cookie or something), shoot me an email and I'll pass along some system specs and rough sketchs.

2 Comments:

  • I'll donate a true Southern feast, complete with my famous fried chicken to the poor CS lackey who ends up doing Adam's grunt work :)

    By Anonymous Emily, at 8/08/2006 12:14 PM  

  • My southern feast was delicious! =)

    By Blogger Adam, at 9/07/2006 10:19 AM  

Post a Comment

<< Home