Wednesday, January 26, 2005

PECOTA projections released

By the email they send to subscribers, I was just told that Baseball Prospectus has released their 2005 50th-percentile PECOTA projections. This is the fourth set of projections I've looked at so far this off-season - first, the ones mysteriously slapped into the Bill James Handbook; second, the ZiPS forecast from Baseball Primer; third, the forecasts in Ron Shandler's Baseball Forecaster, and now these.

The only surprise that jumps out at me is Brad Eldred. PECOTA loves him, projecting 20 doubles and 20 homers in only 364 at-bats. With 127 strikeouts for a final line of .256 / .326 / .485. Not too shabby.

All the forecasts which fail to work with last-minute playing-time adjustments fail to impress me, overall. Why 364 at-bats? All the systems generate those numbers by comparing the player's profile to that of other players. Or so I gather. Of the 20 guys most like Eldred, that's the rough average of the playing time they received in the year after they put up stats that were so like Eldred.

These aren't forecasts so much as they are ways of summarizing what the player's career has been recently. PECOTA says that Eldred has shown enough power. We should focus on that and not on his inability to make contact. Guys like Eldred have hit enough big-league doubles and homers to be useful. Using PECOTA as evidence in a Brad Manifesto, demanding that management release Daryle Ward and give Brad the first 350 at-bats at first base, would be totally naive.

Here are some other general thoughts I have about forecasts. First of all, you have to read everything at Tango's link on the right. The difference between all the forecasting systems isn't as great as you'd think. But it can be significant.

The Bill James forecast bugs me because there's no explanation of a methodology. They just magically appear. I find Shandler's book appealing and compelling because so much of the methodology is explained. I don't know enough about how the ZiPS forecasst is put together, but it appears to be some kind of juiced-up Marcel the Monkey. The PECOTA system creeps me out because it smacks of eugenics and late-nineteenth-century social Darwinism. I believe it is the only system to put great weight on a player's height and weight. This is why we're always hearing that Khalil Greene is too big for a shortstop. Or is he too slow? Memory fails.

There may some advantage to this system. Only time will tell. The 2004 PECOTA forecasts did not impress me terribly much. There were some hits but there were some misses. And worse, there were a lot of fools who wouldn't believe their lying eyes because players were outperforming their PECOTA forecasts. So that besmirched their reputation a little more for me. The BP guys have struggled to balance their snake-oil hyperbolic marketing against its disciple-producing embarrassments. If they want to be taken as scholarly and serious, they will need to be more scholarly and serious as they market their product.

All that said, maybe PECOTA will start kicking ass right now.

In the meantime, I'm a little creeped out by the decision to only compare guys who are six-four two-hundred to other guys in that ballpark. Perhaps I am too sensitive. Scouts have been looking at size and weight distribution for years and years so there must be something to it. When will PECOTA only compare pear-shaped players, then, to other pear-shaped players? Or barrel-chested guys only to other barrel-chested guys? Are all 200-pounders really that much more alike than any other subset of guys in the 180-to-220 range?

The rigid comparison of players only to other players of the same baseball age creeps me out too. Don't we all know twenty-six year-olds with bodies that look twenty-one? Or forty-four? Just because peak years, on average, tend to clump together, that doesn't mean that everyone's body ages and matures on the same clock. Last I checked, for example, there was no one age when (a) puberty begins or (b) hair falls out. Maybe Brad is an "old" 24-year-old. I don't know. But it does sound like Neil Walker is a young 19 or whatever he is.

There's also the issue of treating height and weight and age as definite facts with the same precision as, say, the number of strikeouts someone accumulated. Players change weight over the course of a season. Folks lie about their heights and weights - pretty much everyone fudges a little bit. One guy listed as "200" may really be 210 for most of the season; another one might be 190. At least half of all the guys who claim to be six feet tall are really five-eleven. And many of the foreign-born players fudged their age to prolong their prospect status. For such reasons, I don't see the justification to give such weight to physical profile when making a projection.

Any pearls of wisdom, dear readers, on the subject of the various competing projection systems? Has Rotowire released theirs yet? My subscription lapsed so I don't know.

No comments:

Post a Comment