Friday, August 24, 2012

Holy Cow That's A Bad Way Of Doing It

The new issue of ESPN The Magazine is their NFL preview, which is standard and not notable. In it, they have projected records for each team; again, nothing to see here.

 The way they came up with those records, though, is absolutely hilarious. And even funnier when you consider the project was run by Accuscore, and the magazine is edited by ESPN Insider's own gambling expert(?), Chad Millman.

 Here's how they explain it:

So basically Accuscore took their (super accurate, I'm sure) ratings for each team, figured out who they expected to be favored in each of the 256 NFL regular season games...and then assigned a 100% chance that the favorite would win each game. I'm not sure how exactly that is taking things "down to the last decimal point"; it seems to me that this approach is rounding the win probability for each game to exactly zero decimal points. But hey, details.

This method, as you've probably realized, does not make any sense. If there was a team that had a 55% chance of winning each individual game, any reasonable individual would expect them to go about 9-7; doing it this way would peg them at 16-0. So I'd expect this approach to come up with very extreme projections on both ends of the spectrum. Which is exactly the opposite of what you want for optimal projections, although we're so far from optimal at this point it's not even really worth discussing.

Anyway, let's check out the results!

Well, that would certainly be a season for the ages.

The tails are immediately comical (five teams at 2-14 or worse!), but I think the nearly complete lack of teams with middling (normal?) records is even funnier. One team at 8-8. Nobody at 7-9. Or 10-6. Or 5-11.

It'd be one thing if this mess was the product of some random magazine editor who didn't know what "parlay" meant. But a combined effort between "the clear leader in sports betting systems" and a guy who writes about how to manage your bankroll for the WWL? I don't expect much, but that is remarkable.

(Title of post stolen from a tweet by @SkyKalkman on this very subject.)