Table of Contents

Wednesday, December 10, 2008

Developments in player valuation

Three big developments this week, as pertains to my general interest in valuing players:

1. FanGraphs now has UZR.

It's based on BIS data instead of STATS Inc data, and so as a result is providing different results. We've known for a while that there are big differences between the data provided by each provider, but now we have the exact same methods being used on top of different data sources. ... And we're finding an average difference of ~5 runs between bUZR (the BIS-based UZR) and sUZR (the STATS Inc-based UZR). Which means, again, that I'm most comfortable using a statistic that combines input from both BIS and STATS sources.

From BIS, the best we have is either bUZR or PMR. Dewan's +/- is good as well, but I don't think of it as being on the level of bUZR. RZR from Hardball Times, which I have used a lot the last several years, is now no longer the only option, and so probably won't be something I'll want to use much more.

For STATS, given that MGL can't often release his data, we're pretty much still suck with ZR. And that's ok.

FWIW, at least until mid-next season, I still think our best available option for measuring fielding is CHONE's defensive projections. They average RZR, ZR, and TotalZone, and then regress to the Fans Scouting Report. I expect that future releases will use bUZR instead of RZR...but it probably won't make much difference. Learn to love them.

2. Clay Davenport is revising WARP to make it useful.

WARP sounds like a great stat: a win-based statistic that works throughout baseball history, and incorporates both offense and defense. Unfortunately, it's rendered worthless by two major flaws: it uses much too low of a baseline, and the defensive statistic it uses is essentially range factor--vastly inferior to what we're using today to measure defense (see, for example, FanGraphs' bUZR!).

Well, today they announced that, in fact, Clay is addressing these two issues. They'll use something much closer to the baseline that Tom Tango, Keith Woolner, I, and everyone else uses. And they'll use a "play by play" metric, which I'm guessing is a lot like Dan Fox's SFR, instead of Davenport Translations whenever they have adequate play by play information.

This comes after a lot of harrassment by Tom Tango and others, and it's long overdue. But it takes WARP from irrelevant to something that's actually useful. So we should applaud it.

It still will have some issues. It will use position adjustments based on offensive differences instead of defensive differences, which will mean it'll over-value low-talent positions like 2B. But those differences are far less of a big deal than the issue of baselines and fielding statistics. And it means that a lot more people will be using a quality total value-type statistic moving forward. Thank goodness.

3. Sean Smith's Uberstats

He hasn't released it yet, but Sean Smith is working on an uber database that will be able to calculate WAR for any player season in the retrosheet years. He has indicated that he'll be willing to post the output tables. This would be an unbelievable resource. I'm becoming almost as much of a Sean Smith groupie as I already am a Tom Tango groupie. :)

...I've said several times that I'll be very happy when I don't have to post my own total value stats and can instead just rely on sites that do a better job of it than I can. We're getting there. Might even be there by opening day.