Monday, August 11, 2008

The 2008 NFL and the Favorite Toy

Many moons ago, baseball stat guru Bill James created the Favorite Toy, a simple method of computing a player's potential final career stats and his chances of reaching certain milestones, like 3,000 hits, 500 home runs, or 300 wins. To put it simply, it takes the last three years of a player's career, weighing the most recent season more heavily than the second-most recent, which is weighed more heavily than the third-most recent. It's not strictly scientific -- it's called a "Toy," after all -- but it's a fun way to estimate future performance and minimizes the effect of one flukishly good or bad season on those estimates.

It's tougher to use such methodology for football players, given the vast differences in playing time, play calling, and any individual player's reliance on the performance of the rest of his team. But maybe it can give us some insight into overall team performance. Specifically, maybe it can help us predict the number of wins a team will have in the next season.

The formula for figuring out next season's win total is simple:

[(3 * wins from last year) + (2 * wins from two years ago) + (wins from three years ago)] / 6

Thus, last year's win total is weighed most heavily, but previous years' totals are also significant, thus lessening the impact of an unusually good (or bad) season. To test the accuracy of this method, I computed each team's estimated 2007 win total, using their 2006, 2005, and 2004 seasons. Rather than bore you with the whole chart, I'll use this summary chart to show how accurate it was:


"WinDiff" is the difference in wins from 2006 to 2007. "TeamsLY" is the actual number of teams with that difference in wins from 2006 to 2007. "TeamsFT" is the number of teams with that difference in wins between their predicted wins by the Favorite Toy method and their actual 2007 wins.

In other words, only one team -- the Buffalo Bills -- had the same record in 2007 that it did in 2006, and thus was 0 wins off. If you had tried to pick every team's record in 2007 based solely on their performance in 2006 -- which a lot of people tend to do -- you would have gotten one pick correct. You would have been off by one win 4 times, by two wins 9 times, and so on. (That's Baltimore, BTW, which lost 8 wins from its 2006 season.)

On the other hand, if you'd used the Favorite Toy to try and guess teams' wins, you would have been right on target 6 times, off by one win another 6 times, and off by two wins 4 times. I consider anything within two wins to be "close enough," so that means you would have been "close" on 16 or 32 teams.

Granted, that ain't saying much; any method that is right half the time is not much better than a coin flip. However, the FT was, overall, more accurate than the "last year" method, at least for the 2007 season: On average, the FT missed by 2.56 wins, compared to the LY missing by 3.25. I haven't run this for multiple years, so I don't know if it's always more accurate, but it's a nice start.

Also, for many of the teams that badly missed their FT estimate, you can find extenuating circumstances that help explain why they were so far off. Some teams had notable off-seasons, such as the Patriots (off by 4) acquiring Randy Moss or Atlanta's (4) Michael Vick debacle. Others had in-season reasons to explain their inaccuracies, either due to injury or ineffective key players, primarily quarterbacks -- like Chicago (4), St. Louis (4), Baltimore (5), or Miami (6) -- or, in a few cases, unexpectedly great play from the QB position -- as was the case with Green Bay (6), Cleveland (5), Dallas (4), and Arizona (3). In fact, nearly every team that missed their expected number of wins by the FT method had significantly better or worse play at quarterback than expected. While I'm not of the opinion that a great quarterback is absolutely essential to a team's success (and I better not be, being a Vikings fan this year), it seems that a QB performing markedly different from his expectations, for either good or bad, can have a profound impact on his team's successes.

(On a similar note, I thought yesterday, "What if Tarvaris Jackson is really good this year? Say, along the lines of 3,500 passing yards and 25 TDs?" A lot of people think we'll do well with a mediocre QB, but what if we had a really good one?)

Injuries and variances in play are virtually impossible to predict. In terms of notable off-seasons, I think the Packers, Jets -- both teams even before last week's trade -- and the Vikings had the most eventful. Still the Favorite Toy can't take that into account, so we're left with just the raw numbers from the previous three seasons' worth of games. I'll close today with the 2008 NFL standings, as predicted by the Favorite Toy, using the 2005-2007 standings:

New England142
NY Jets610



San Diego124
Kansas City79

NY Giants106

Green Bay106

Tampa Bay88
New Orleans79

San Francisco610
St. Louis511

Comment at will.

No comments: