When Good Writers Get Lazy

The Prospectus Triple Play made its way back to the Mariners today. Go ahead and click; it's a free article.

After making the obligatory mention of new resident stathead Mat Olkin, Jim Baker proceeds into a brief discussion of 2005 PECOTA projections for certain Mariners. Which would ordinarily be perfectly fine, but Baker treats the forecasts almost as indisputable fact, when the reality is that there are a few significant flaws in his reasoning. For example:

(1) Yes, Sexson is pegged for a 23.6 VORP. In about 425 plate appearances. Healthy Sexson collected 2,025 plate appearances between 2001 - 2003 - 675 a year. Projecting that VORP over a full season gives a figure of 37.5, which would put Sexson's production at 2004 Kevin Millar.

But then, I don't really like VORP forecasts anyway - trying to predict couting stats is a fool's errand. PECOTA sees Sexson posting a .254/.347/.480 line in Safeco, which adjusts to a .261/.357/.504 line in a neutral environment. If that doesn't seem like a perfect match to Sexson, circa 2002, then I don't know what does. Still not really worth the contract, but it's a comfortably above-average performance. Baker's selling him short.

(2) All right, so PECOTA doesn't like Adrian Beltre's chance of building on his 2004 season very much. A .279/.337/.486 line in Safeco isn't out of the question. That said, Beltre's Similarity Score has got to be frighteningly low - that is, there isn't an established historical precedent for 25 year olds suddenly exploding with an MVP-type season after underachieving (partially) due to injury. With this in mind, attempting to project Beltre's 2005 season by comparing him to similar players is a futile endeavour. With so few comparisons, the forecast range will be enormous, and averaging them out into a weighted mean is just wasted time. PECOTA's neat, but it can't account for statistical outliers - just like with Jamie Moyer and Randy Johnson, all we can really do with Adrian Beltre is guess and watch.

(3) As if that weren't enough, Baker goes on to mention the Ichiro forecast. You know who PECOTA considered to be the most similar player to Ichiro a year ago? Matty Alou, a career .307/.345/.381 hitter with 156 steals in 1667 games. Others on the list of comparisons include Lance Johnson, Tony Womack, and Luis Polonia. Nate Silver might as well just make a random guess for Ichiro's performance every year; it would probably be more accurate than whatever numbers his computer spits out.

(4) Baker's final point is that Joel Pineiro will lead the pitching staff with a 22.1 VORP, the third-lowest in baseball among team leaders. Looks pretty bad, right? I mean, Ryan Franklin was miserable last year, but even he managed a 22.7 VORP.

So what's the deal? Well, as a conservative forecast system, PECOTA tends to underestimate plate appearances and innings pitched. This year, it projects Pineiro to throw 138.1 innings, Moyer to throw 146, Madritsch to throw 106.2, Franklin to throw 147.2, and Meche to throw 124. The biggest component of VORP is playing time - this is the nature of any counting stat. When you undershoot the probable innings pitched of everyone in a starting rotation by 50-60 IP each (although, again, Franklin could lose his spot), they're obviously going to wind up with unimpressive counting stats.

The thing is, Baker knows all this, but neglected to mention it. Maybe he didn't have enough space in the article. Maybe he assumed that everyone would understand the caveats inherent in his reasoning. Maybe he couldn't think of anything to write about, so he went with something easy. Whatever the case, it came out as a throwaway column. I've come to expect better from the guys over there, and this kind of thing disappoints me.