Covering the Front and Back Pages of the Newspaper
July 14, 2005
BASEBALL: EWSL At The Half
While we're stopped for the All-Star Break, I thought I'd look at the projected standings I did before the season (see the NL standings here and AL here) by comparison to the actual standings at the break. I'm presenting the teams in order from the teams most overachieving compared to their EWSL standings to those falling furthest off the pace (I've prorated the EWSL wins to the number of games each team has played so far):
Overall, the divisions come in as follows: AL Central, +28; NL Central, +7; NL East and AL West, -3 each; AL East and NL West, -16 each; AL overall, +7; NL overall, -12; MLB in total, -5, which means that some of the differences here can be explained just by rounding. On average, AL teams were 5.21 games off their EWSL records, NL teams 4.25 off, MLB as a whole, 4.7.
That's not a great record, but it's too early even now to declare it a big failure. Some teams will always deviate from even the most perfect pre-season estimates, due to trades, unforeseeable injuries, unexpected rookies.
No surprise that the biggest underachievers have been the Giants, since the pre-season EWSL standings were predicated on them having Bonds for half the season, and he hasn't arrived yet.
Also unsurprisingly, several of the teams that are out of whack with their EWSL numbers tend to be the same teams that are way over or under their Pythagorean projections - in other words, their players are playing closer to expectations than their records would suggest. Conspicuous examples include the White Sox (+16 vs. EWSL, +6 vs. Pythagorean record), Nationals (+6/+8), and Mariners (-5/-3). But there are counterexamples as well: the Brewers are +9 vs. EWSL while being -4 vs. their Pythagorean record, the Braves (+5/-4), the Indians (+7/-2), and the Diamondbacks (0/+7). On the whole, the average distance between the EWSL and Pythagorean records is 4.7 games - just exactly the same as with actual records.
I was surprised to discover that EWSL had consistently failed to grasp how bad the really bad teams would be - all the moreso because I thought before the season that the method was being unduly harsh on the Royals, Rockies and Devil Rays. Not hard enough, apparently.
Anyway, I may take a look at some of the big outliers, and I'll have to go back at the end of the year and see how EWSL did, and why, and whether there are further refinements to the method that will improve its predictive value, or whether its shortcomings as a predictor are just the inherent limitations of using past performance to predict the future.