In case you hadn't noticed, Fangraphs has a number of different player projection sources shown on each player page. So, which Astros' player is hardest to project? I didn't do anything scientific here---I just eyeballed the 2012 projections for Astros' players. Also this is limited to players who were projected by most of the sources. This exercise is intended to identify players with a a wide range in variation for projected offensive results.
The projection systems are Steamer, Bill James, Marcel, Rotochamps, Fans, and ZIPS. Marcel is an intentionally plain vanilla projection method based on regressing past performance data. The fans' projection is based on the projections submitted by readers of fangraphs. Most people assume that fans of the players' team tend to predominant in the Fan projections.
Based on the fangraphs' listed projections, I nominate Brian Bogusevic as the hardest to project. Rotochamps in the only projection system which didn't provide a projection for Bogusevic. I provide the various projection systems' results for Bogey's OPS, wOBA , and batting average below:
(OPS, wOBA, BA)
Steamer ..671, 300, .239
Bill James .720, .325, .260
Marcel .761, .332, .270
Fans .714, .324, .266
ZIPS .652, .295, .233
Bogusevic is a quite different player, depending on whether Marcel/Bill James or Steamer/ZIPS does the projecting. On the one side, he is close to a league average hitter (though not a league average hitter for the RF position), on the other side, he is lucky to be a bench player. (For those who are unaware, "Marcel" is a plain vanilla projection developed by Tangotier, one of the authors of The Book.) The fans picked offensive results in between those two sides. There are some reasons to expect projection difficulties for Bogusevic. Most obviously, he doesn't have a lot of major league experience. For projection systems which rely greatly on aging curves, the combination of little ML experience and his age (28 years old) is a big negative factor. Bogusevic's minor league results also show some inconsistency.
Looking at the underlying peripherals, Bogusevic's expected BABIP and Isolated Power (ISO) seem like the major areas of disagreement. Most of the sources are pessimistic on Bogusevic's power, and the ISO range is .107 (Fans) to .156 (Marcel). For inexperienced players, determining the expected or "normal" BABIP is not easy. Bogey's BABIP is projected between .299 (Steamer) to .328 (Marcel, Fans). For what it's worth, I think this range for Bogey's BABIP is too low. Bogusevic's x-BABIP, based on the expected BABIP model, for 2011 was .335.
My runner up nominee for the hard to project award is Brett Wallace. This isn't surprising, since most of us are befuddled in our expectations for Wallace's 2012 offense. Look at his projections, below.
(BA, OPS, ISO)
Steamer .247, .693, .133
Bill James .268,.731, .132
Rotochamps .250, .692, .113
Marcel .257, .708, .122
Fans .270,.740, .128
ZIPS .258, .725, .145
None of these projections are exciting. But we can see a noticeable division of opinion on batting average (acceptable vs. bad). And I have to ask, what happened to the guy who was supposed to be a high average hitter (.290 - .300)? The expected OPS ranges from average to below average to bad. The one time slugging prospect has a variety of ISO expectations too, with only ZIPS holding out hope that he has more than slap hitting power. The projection systems have considerable differences in projecting Wallace's BABIP, ranging from .305 (Steamer) to .335 (James).
Next, I'll pick the most difficult to project pitchers.
It seems to me that the pitching projections for various Astros' pitchers are relatively consistent across systems. Of course, all pitchers have differences--no players are that easy to project--but the ERAs and FIPs tend to cluster within a general range of 0.2 - 0.5 runs per game. Players with a lot of experience (and therefore historic data) tend to produce the most consistent projections from source to source.
I nominate two co-candidates for most hard to project: Henry Sosa and Fernando Rodriguez. Sosa and F-Rod have some similarities: older players (age 27 and 28) with little ML experience who were prospects at one time. For each pitcher, the projections are roughly consistent, except for one outlier. The ERA projections are shown below.
Henry Sosa ERA
Bill James 4.54
Fernando Rodriguez ERA
Bill James 5.40
As you can see, ZIPS is a lot more pessimistic on Sosa and Bill James is more pessimistic on Rodriguez. Periipherals appear to be the cause for the differences. In the case of Sosa, ZIPS forsees a much worse K rate (5.99) than the other systems, plus a a bad walk rate (4.22). In the case of Fernando Rodriguez, Bill James doesn't think his K rate will be more than simply average and views him as just more hittable. Most of the other projections project a K rate for F-Rod from elite (ZIPS 10.46) to good (Steamer 8.16).
Do you think these Astros are that hard to project?