This can’t be right.
The projections, which twice correctly predicted huge jumps in the standings in 2008 for the Tampa Bay Rays and 2006 for the Chicago White Sox, had a very bad year last year and another bad year might cause people to look to other projection systems or maybe ignore the process.
PECOTA uses an intriguing system. Started by SABR legend Nate Silver, it mapped every career ever had by any player in history and then applied those career arcs to the players with similar production–like I said in an earlier post, baseball is fortunate enough to have such an enormous sample size to take from. From there, it extracted what the likelihood of the player’s production would be–10% (bad), 50% (average for him) and 90% (way above average). Then it takes those player predictions, combines them into a team’s total run production (and prevention on pitching and defense) and voila! You have your pythagorian win-loss record.
One reason why PECOTA had such a bad year in 2009 wasn’t because of bad luck–PECOTA projections do not account for injury, trades or other things that come into play during a season–it’s because it projected inaccurately.* Even by pythagorian record, the A’s and Angels were swapped. The Indians and the Diamondbacks just plain old stunk. Craig Calcaterra pointed out PECOTA predicted one of the most amazing seasons in history by a rookie catcher for Matt Weiters and they were pretty far off.
*And just a minor anecdotal example: Andre Ethier’s projected 2009 home run total in his 10th percentile was something like 18 and his 90th percentile was 24. He hit 31.
Well, look at the rankings for this year (linked at the top and here). There’s some … interesting things in there. A lot of people are pointing out the Yankees missing the playoffs, which PECOTA projects they will be about three games (Rays win 96, Red Sox win 95, Yanks win 93), but more interesting is the NL East.
The Nationals will finish in third place, 82-80. That’s right. An above-.500 season for the Nats even though they’ve made virtually no roster changes after losing 100 games last year. And they’ll finish above the Marlins and Mets.
That first one seems absurd. The Marlins have a roster with some very good young players (and some not so good) and finished with a pythag of 82-80. They had some luck, yes–Coughlan finishing with a very high BABIP will likely mean a hard fall for him–and they’ll likely suffer from regressions with Josh Johnson, but the gang is all still there and they’re another year toward their primes. So it’s that much more curious that the Nats will surpass them after losing their best on-base guy (Nick Johnson) and with only Ryan Zimmerman, Adam Dunn, two or three up-and-coming position players with under 400 major league plate appearances, a number of players who won’t or never did have a prime and very little pitching outside of the upcoming Strasburg and maybe 150 innings of Jordan Zimmermann, that they’ll surpass the Marlins in the standings.
If PECOTA’s right on this one, I’ll eat my hat.
Other amusing things: only three teams will finish with 90 wins and they’ll all be in the AL East; the AL Central winner is the Twins with 82 wins–hmm, wonder what would happen if the Nats moved divisions; Oakland will win the AL East (87 wins) and the Angels will finish last (76), inverse to last year’s order.
That last one at least makes some sense if only because the Angels lost four major contributors to the 2009 season (Abreu, Vlad, Figgins and Lackey) and don’t have much of a rotation for the 2010 season. Hideki Matsui replaces Vlad at DH and Brandon Wood replaces Figgins, but there’s still a big hole in right and in the rotation (and probably a AAA-Majors hurdle Wood has to jump over). Knowing how things ended last year, Mike Scoscia will use his damn voodoo majick to get the Angels another AL West title.
Edit — As PCSimSports pointed out, PECOTA is constantly tweaked and altered before the season and this isn’t the final version. I’d do an “I knew that” retraction, but I’m above that. More commentary will come for the final version.