March 14, 2006
Learning from Mistakes
Yesterday I put up one of those columns, my personal version of a 3-8-7-7-5-1 line, complete with a balk, two wild pitches and some really bad pasta from the postgame buffet.
Fortunately, the readers are smarter than the writer, so if you're scoring at home, Takeo Spikes is a Buffalo Bill, not a Cincinnati Bengal; the former Jairo Garcia is now Santiago Castilla, not Sergio Casillas; and the A's bullpen had 8.1 strikeouts per nine innings last season, addressing a problem that had developed in 2004, but apparently embedded itself in my brain.
Sorry, folks. I should be better than that, and will continue working on it. Let me make it up to you.
It's appropriate to open today's column by looking back at some mistakes, because what I'm trying to do today is learn from them. Just shy of a year ago, I published my predictions for 2005, with records and runs-scored/runs-allowed figures for each team. As we look ahead to the '06 season, with prediction season just around the corner, I thought it would be useful to examine last year's calls, makes and misses, for lessons that can be applied to the upcoming season.
First things first: I overestimated league run scoring by about 5% (23573 runs projected, 22325 runs scored). So for this comparison, I adjusted all of last year's projected figures down by 5%; I missed pretty badly on the offensive level last year, and that's meaningful, but I think the comparisons make more sense if I concede that error at the top and then work off adjusted figures.
The coin of the realm here is runs, because that's what teams deal in. My record projections stemmed from projections of runs scored and allowed, with an occasional adjustment for the strength of a team's bullpen. In a single season, records don't always match runs, but that's the safest way to bet, and it's the best way to evaluate my performance. I may have nailed the Blue Jays' 80-82 mark, but it's hard to consider it a success when I had them at -18, and they finished the year at +70. Similarly, the Mets were eight games off my projected W/L of 91-71 but were +74 in runs, just a hair behind my projected +79.
Looking at the data, I'm actually fairly surprised that I did as well as I did. I thought I had a terrible season, but I had three teams--the Red Sox, Pirates and Mets--within five runs of their adjusted spread. I pegged four teams with their actual records (in addition to the Jays, there were the Rockies, Red Sox and Phillies). I have no frame of reference yet to say whether this is good or bad--I'd never projected runs before--but I am encouraged that my predictions weren't all wrong.
Some of them were, though. Some of them were very wrong. The most notable one was the White Sox, who I projected for -101, largely thanks to an estimated 811 runs allowed. They actually allowed 645, and the end gap between their actual spread and my projected one was a whopping 197 runs, the highest on the board. Three NL West teams, the Dodgers, Giants and Diamondbacks, are next in line in the 130 range, followed by two other teams that, like the White Sox, did a great job of preventing runs: the Astros and Indians.
Now, two of those six teams were among the biggest injury cases of 2005. I expected the Giants to get much more from Barry Bonds than they actually did; they scored 118 fewer runs than I projected them to, while allowing just 17 more. The Dodgers were devastated on both sides of the ball, but it was injuries to Odalis Perez and Eric Gagne that showed up most clearly: they allowed 106 more runs than I projected on their way to the biggest gap, 22 games, between projected and expected records.
Three of the other teams were ones that kept runs off the scoreboard like few others. The White Sox allowed 166 runs fewer than I projected, the Astros 137, and the Indians 132.
The White Sox were the biggest baseball story of the year, and rightly so. The combination of a rotation that threw strikes, some unexpected good years from veteran, low-profile relievers, and the game's best defense turned them into a champion. Their offense was a bit better than expected, thanks to power from nine lineup spots, and that was enough.
The lessons I can take here are that a team can improve its efficiency by doing a better job of aligning its talent. The White Sox weren't as good at the plate as they'd been in '04, but they filled a need by getting some OBP to bat in the first two slots ahead of their right-handed power. Scott Podsednik got attention for stealing bases, but it was his OBP, and that of Tadahito Iguchi, that drove the offense. I'm thinking of the Cubs here, who should be better for having Juan Pierre getting on base, and the Twins, who added league-average players in Luis Castillo and Rondell White, but had an offense so bad last year that league-average players represent a considerable upgrade.
The real question raised by the White Sox, though, has to do with the projectability of defense. Should we have known that the addition of Podsednik, essentially a second center fielder, would help the Sox turn singles and doubles into outs at a high enough rate to make a difference? Should we have seen the emergence of a shutdown left side in Joe Crede and Juan Uribe? The Indians' figure is in part a park effect--the Jake has become a very good pitchers' park--but it's also a defense thing, with young players Jhonny Peralta and Grady Sizemore upgrading the range. Can we look at teams right now that will be better than expected in '06 because of their defense?
The easy choice is the Diamondbacks, who added the game's best second baseman and a right-handed, groundballing righty to pitch in front of him, and should be better at third base with Troy Glaus gone. Some of that edge, however, is given back by the signing of Eric Byrnes--a corner outfielder stretched in center field--and the aging G boys on the corners. The sooner Chris Young, and eventually Carlos Gonzales, can get to Phoenix the better. The D'backs may make the leap in '07, not this year.
Other possibilties include the Padres, who added the superior Mike Cameron to play center in cavernous Petco Park. Even at 33, Cameron can play a championship-caliber center field. The Phillies picked up Aaron Rowand, who has quietly become one of the game's best defensive outfielders. He could help statues Bobby Abreu and Pat Burrell survive the season.
Could a team get the kind of frontline pitching that the Astros did in 2006? The answer is most likely no--the 'Stros made all kinds of history last season--but it's not impossible. The A's have a very deep rotation, and their front two look a bit like the Astros' Roy Oswalt and Andy Pettitte. There is no comp, however, for Roger Clemens. The White Sox may have the pitching for another great year, although their defensive performance will likely regress. The Cubs, even without Kerry Wood, have the top two to make a run; it would help if Greg Maddux found his inner 1995 for a summer.
Focusing on runs allowed, and finding the places where teams could dramatically improve, may be the key to more accurate forecasting. Consider that I missed the 30 teams' runs scored by a total of 1312 runs, but missed the runs allowed by 1491 (these numbers don't match because the figures are absolute differences, not signed ones). Eyeballing the spreadsheet, I pegged half the teams in baseball within 30 runs of their total, after making the adjustment for offensive level. I had just a dozen RA figures under that mark. I missed three teams' runs allowed by at least 132 runs; I didn't miss any teams' offensive totals by more than 123. This isn't groundbreaking, actually; offense is simply easier to project than run prevention. Pitchers are wildly unpredictable, and we haven't been able to settle on how to evaluate defense, much less project it.
If you look at RS and RA individually, rather than the gap between the spreads, the picture stays much the same. The White Sox, Astros, Indians, Dodgers, Giants and Diamondbacks all hold spots in the top seven…but the Devil Rays jump to the top, with a net error of 220 runs. I had their gap right (-212, actual -186), but they both scored and allowed a bunch more runs than expected.
I go back and forth as to which of these measures is a more valid way of assessing performance. The results are similar, thought not identical, and the former seems to me to better capture what we're trying to do--evaluate the estimation of the RA/RS relationship--than the latter. On the other hand, missing both by 200 runs shouldn't look good just because you managed to distribute the mistake evenly between bats and arms. Two wrongs most certainly do not make a right. Consider the Red Sox: I projected 860 runs scored and 756 allowed; in actuality, they scored 910 and allowed 756. One scoring method pegs that as a 1, the closest I got to a hit; the other scores it a 99, a below-average figure.
Other things that jump out at me on the spreadsheet:
I hope, that in a week's time when I'm putting together my projections for 2006, I'm able to do a better job. I'm certain, however, that I'll be better-equipped to ask the right questions this year.