clock menu more-arrow no yes mobile

Filed under:

Predicting Records

This will be year 3 of my subjective, but systematic NFL record predictions, here's the explanation from last year:

After the schedules come out there's always a flurry of win-loss predictions based on picking the winners of each individual game. I've always thought that was kind of dumb and subjective, even before you take into account how hard it is to now how well teams will play. Some games are obvious and would be a stunning upset others are difficult, could go either way and in my case at least are often decided in the interest making the final records turn out the way I expect them to. So last year I tried to use a method that was less dumb (but at least as if not more subjective).

Each game is scored on a scale from 0 to 6

0-I'm sure the visiting team will win
1-Visiting team probably wins
2-Leaning towards visiting team
3-No Idea
4-Leaning home
5-Probably home
6-Home wins

Last season's predictions were off by 2.9 games per team. Not great, but only a touch off Football Outsiders projections in Pro Football Prospectus (2.8), within a third of a game of FO's final projections (2.6) and a full half a game better than the previous year (3.4). Predicting records right before the season is hard enough, doing it in May/June is asking to fail.

Here's to failing less this season,


I've got a formatted spreadsheet (Update: the total record formulas were referencing the wrong cells. Should be fixed in the new downloadable spreadsheet Update2: found another mistake in the template. It's fixed now and I got all the way through mine without any other issues) with the games to place a 0-6 in marked with a G, for anyone that wants to play along at home. I'm partway through my first run through, and it seems to work fine, but if you find an issues comment and tell me, and if any excel wizards out there want to soup it up, feel free and e-mail it to me.