Bill Phipps and I have been going back and forth on the merits of having a precise system that converts projected stats to dollar values, and I think there’s been some confusion I’d like to clear up.

Here’s where Bill is right: Math is important to evaluating baseball players. Billy Beane’s understanding of how walks and more specifically on-base percentage contribute to runs scored helped him build competitive A’s teams for years on the cheap. In a way, Bill’s model is similar to Beane’s: know what stats are actually worth and build a better team for the money.

Here’s where Bill is wrong: unlike the situation Beane faced in the 1990s, when few GMs appreciated the value of on-base percentage in terms of runs created, Bill is playing against people who do appreciate the value of their players’ various fantasy stat contributions. It’s perhaps true that opposing fantasy GMs don’t have those contributions valued as precisely as Bill does, but as long as they’re in the ballpark, it doesn’t matter very much.

Here’s why. Let’s say Beane has a great system for translating a player’s prior stats into a monetary value. He uses some hybrid of runs created, VORP, WARP, etc. that really drills down into exactly what Jake Fox or Rajai Davis is worth. Some other GM, say Brian Cashman – he knows the importance of OBP, too, but just looks at OPS – on base plus slugging. His read on free agents and minor leaguers is more or less correct – he looks at contact rate, BABIP, etc. but just not quite as precisely as Beane.

Does this confer a significant advantage to Beane? I would argue it does not. Last year’s OPS and other important peripherals are close enough for Cashman to appreciate the player’s overall skill set. Neither GM knows what the player will do in the upcoming year (though the team with the better scouts will have a material leg up on that front), but both have a good enough appreciation for what last year’s stats were worth. That Beane knows a player earned $5.82 million *last year* in terms of value to the team, and Cashman has the number anywhere from 5.5 to 6.8 roughly, doesn’t mean a whole lot. Both appreciate generally what the player can do, and both will need to talk to the scouts to see whether that player is likely to get substantially better – really the key to deciding whether to sign him at a certain cost (along with the team context – another huge key). And anyway, we have to keep in mind that’s what he earned *last year*. Going forward he might earn close to that again, but we can’t say with precision.

I think we’re in a similar situation here. I take it as a given that any expert knows roughly how much any player given his age, prior history, stats and circumstances is worth. Within a few bucks depending on one’s “scouting” of the player – that player’s potential for growth.

For Bill to have a big advantage it would have to be a case where we forgot our league counted stolen bases as a category, or we were drafting strictly by our projections and forgot to include runs scored in our model. Something egregious. But when you’re talking about *last year’s stats* and that’s this year’s projections are always based on the past, a small amount of extra precision in valuing them will not matter much, given the variance and uncertainty we have about this year.

Ask yourself – do you really think it matters for a real life GM to know EXACTLY what a minor leaguer did in terms of runs created in 2009, or if he has a pretty good idea based on his OPS and skill set, is that enough? What advantage does he get from knowing last year’s number’s exactly? Is it important to know whether a player had exactly 36 or 38 home runs last year? Or his exact slugging percentage? Or might it be enough to know that he’s a major power hitter with good contact skills (unlucky BABIP – whatever it was) and improving plate patience. In a better park, and better on-base guys ahead of him. Isn’t that more important than knowing exactly what he earned last year?

The reason for some of the confusion, and why some of my posts could possibly be misread as being anti-math or anti-science isn’t because I don’t pay close attention to things like BABIP, strand rate, homers/fly ball, etc. It’s because I take it as a GIVEN that we all know all of that stuff already. The difference is going to be in how good we are at navigating the unknown, not taking the known and scraping it for whatever tiny untapped morsel is left.

(This is not to say that truly important advances in player evaluation won’t still occur – only that Bill’s model is not even attempting that but merely working at the margins).

Assuming expected growth is priced into the market (the consensus dollar values, the dollar values that others are willing to pay at auction), one will win by beating the market – seeing room for growth where no one expects it – seeing greater likelihood of decline for players who others trust. That’s where you win fantasy leagues.

Put differently, my thesis is that the market for converting stats to dollars (or player skills to dollars as I do it) is far more efficient than the one that prices in player growth or regression. Yes, you can pocket a buck or two here or there if you had a better model for players (and those players projected stats were more or less accurate), but that’s not where the big profits lie. And you’re also dependent on the projections being good in the first place, a dubious proposition given the margin of error and the small edge your model gives you even if the projections are fairly accurate.

Tags: Chris Liss Bill Phipps Billy Beane Player Evaluation

You can follow any responses to this entry through the RSS 2.0 feed. You can skip to the end and leave a response. Pinging is currently not allowed.