WU may not be as excited, but I think being able to do my own projections is pretty sweet. And thanks to Tom Tango's Marcel projection system and this sweet post from Triples Alley, I think I've figured out how. No guarantees, but I think I got it. So what is Marcel? In Tango's words:
It is the most basic forecasting system you can have, that uses as little intelligence as possible. So, that's the allusion to the monkey. It uses 3 years of MLB data, with the most recent data weighted heavier. It regresses towards the mean. And it has an age factor.
That sounds awfully simple, but in fact it does very well in comparison to other systems. Sure, CHONE and ZiPS are better (disturbingly, PECOTA was not...at least not in 2009), but by a pretty small margin. Remember, the point of Marcel is not to outperform other systems. It's to establish a baseline of minimum competence for other systems. Without further ado:
A couple methodology/interpretation notes:
- BR600 = batting runs prorated to 600 PA, fR150 = fielding runs above average at that position for 150 games. I used what I thought were reasonable estimates given UZR, FSR and the SSS consensus. For the sake of simplicity, I'm assuming 600 PA = 150 G, so I can get a total runs per 150 games above average. Posadj150 is the position adjustment for 150 games, as defined here, and replacement level is 20 runs below average per 150 G. The end result is R150, or the total runs above replacement per 150 games.
- From there I used the Fangraphs Fans' projections for the plate appearance projections. I'm pretty sure I can do the Marcel playing time projections, but I think they're replaceable where the wOBA projection is not. The fans' projections looked pretty good to me. If you don't like them, use your own figure. For Morel, I used Bill James' projection. To go from R150 to RAR, I divided by 600 PA and then multiplied by the projected PA.
- In case it isn't obvious, I used what I thought were reasonable playing time estimates for the bench guys. And I totally forgot to do Mark Teahen's projection until about an hour ago. Gross.
- I used .325 wOBA as league average rather than my usual .330 since I think that's somewhat more accurate if somewhat less convenient. It's not going to make a huge difference.
Now for the prospect projections. The Morel, Flowers and Viciedo projections ARE NOT Marcel projections. Marcel does not use any minor league numbers in its projections and all three would be projected to hit league average and receive 200 PA in Tango's system. I thought that would be too boring, so I used minorleaguesplits' Minor League Equivalency calculator to turn their minor league production into a major league line for the past three seasons and then I ran a Marcel on that.
This has two primary drawbacks. First, that means regressing their numbers to MLB average. They have not yet proved themselves to be major leaguers, so this inflates their projections somewhat. This assumption works a lot better for Viciedo, who was very impressive in his cup o' coffee, and Morel, whom the White Sox are likely to give some 400+ PA because they have judged him to be ready. If they hand the job to Teahen instead, we should downgrade Morel's projection.
Two, MLEs tend to be marred to some degree by survivorship bias. The way they are typically calculated is by tracking players who make the major leagues and comparing their major league performance to their minor league performance and then adding some kind of age adjustment. But that means only players good enough to make the majors are in the sample. The best MLEs try to correct for this in various ways, but minorleaguesplits is going through some kind of transitional period and I'm not sure where to find the methodology. I'm guessing that odds are there's some inflation. Again, this is less (though certainly somewhat) relevant to Morel and Viciedo than it is to Tyler Flowers. Seriously, don't look at T-Flo's projection and bemoan AJ's resigning or worse, take up Tdogg's wishcasting. Think of it as more like a 75th (or 90th) percentile projection for him and, let's say, 60th percentiles for Viciedo and Morel.
As for the projections in aggregate, the suggestion is about 21 WAR, which makes me think the playing time projections are kind of inflated. It's also true that I did not park-adjust these projections, so anyone who has been playing in a hitter friendly home yard/league over the last three seasons will have a somewhat higher projection than he ought. Still, I think it's pretty evident that the team is missing the blatant gaping holes at a position or two that have plagued Kenny Williams' recent lineup construction and persuaded SSSers to entertain sadomasochistic notions well before pitchers and catchers report. The result has been an average of 16 fWAR over the past three seasons and far more angst than necessary. 2011 looks different. In qualitative nomen shorthandum, it's a collection of Healthy Average-ish Talent. To put a number on it, let's call that 18 or 19 WAR. Combine that with the Sox' 24 or 25 pitcher fWAR three season average and we'd have a 91 win team. Whether or not they actually surpass the 88 wins of '08 and '10, it would seem this algorithmic divining rod predicts the most watchable White Sox team since game 163.