Picking NFL Games

Those who know me, know that I'm a die hard Minnesota Vikings fan. I've been a Vikings fan since I was a little kid. I lost my first football bet rooting against Joe Montana in the NFC Championship game... go figure... I think I was 7 at the time.

With all my computer programming knowledge, you would think I'd be a little bit better at picking football games. I decided to have a go at it this year, and I'm excited about the prospect.

One thing that I have noticed over the last several years is that media hype has a lot to do with how games are handicapped, and that most people focus on a core set of statistics that are flawed in there very basis. When you look at points scored, yards allowed, and yards gained for a given NFL team, you are only looking at a brief summation of what happens on any given week. You don't see things like garbage yards or points in garbage time (when a game is already won or lost).

There are a few teams and players out there that have an old fashioned sense of pride and will play every minute like it is the most important minute of the season. More often though, you see that the level of effort drops off significantly when a game is decided. Preventing injury becomes more important than defending a shutout. And then there are players out there that are statistically great, but you don't necessarily trust them in certain situations - guys like Daunte Culpepper.

Ideally, I'd like to see a community of people pushing in real data - information about grossly ill advised calls by the referrees, meaningless statistics that should be pulled from the analysis bed, etc. For the early on, however, I think it's best to start looking at the standard set of data a little bit differently and seeing how that points to more realistic prognostication. For instance - I think it's a much better assessment of a defense if you analyze how many third down conversions they allow, while overlaying the average statistical success of the offenses they play - rather than just looking at raw yards allowed figures.

Take a look at the upcoming Bears/Vikings game for instance - the raw numbers show a close game - the bears have allowed 2 of 20 3rd down conversions against teams in week 1 and 2 - whereas the vikings have allowed 7 of 27 3rd down conversions. Both point to each team having a powerful defense - with the Vikings allowing more conversions, but forcing more third down situations. Percentage-wise - it's 10% vs. 25% - showing that the Bears are significantly better than the Vikings.

With only 2 weeks down in the season, it's easy to give the bears the edge - considering that they were one of the top defenses last year - but overlaying the data for the teams they are playing against shows a different picture - Washington as a whole is 6 for 27 on 3rd down, and Carolina is 5 for 27 - showing that the Vikings defense is essentially playing on par with the rest of the league. The teams that the bears played - the Packers went 9 for 18 in week 2, and the lions went 5 for 14 in week 1. The bears defensive play is miles above the league competition.

The bears do have an edge, a much more significant edge than you might think - given that the Bears played two downward spiraling teams in the Lions and Packers, while the Vikings have played two of last years playoff contenders.

Prognostication is more than just defensive analysis. How good are the Bears and Vikings offenses, and what caliber of defenses are those offenses have been playing? This game is a division game, for first place in the division - so that is a factor (although IMO, less of one at this point in the season). Then you have to factor in field location, playing surface, and weather - and a host of other variables that traditionally can affect the outcome. All in all, I think that solid data analysis can provide you with enough information to make firm prognostications. During Week 1, my buddy and I went 7 for 8 in picking games with no analysis other than our own personal knowledge. During Week 2, our combined analysis yielded another 7 for 8 prediction. For week 3, I'm putting together a semblance of a data structure to process information - hard to beat what we've done so far on our own, but going 8 for 8 would be really cool anyways.