Wednesday, October 01, 2008

Preliminary 2008 Total Value Estimates

I've updated my total value estimates to include all regular season games played, assuming Hardball Times & ESPN have updated all of their sites through last night's game.

You can read about all the methods in this series of posts. I have some additional comments on the methods at the bottom of this post for those who are interested.

Features:
  • Total value estimates for every MLB position player, based on hitting (RAR; Runs Above Replacement level based on linear weights), fielding (Fielding), and their position (PosAdj). Units are runs produced above a replacement player.
  • Fielding measures are based on the average runs saved according to zone rating (ZR) and revised zone rating (RZR).
  • Pitchers are listed according to RAR (base runs saved above replacement) and FIP-Runs (a defense-independent pitching stat estimating runs saved above replacement).
  • League differences are taken into account. So are park differences.
  • Closers get a leverage-based bonus in value (though it's pretty rough).

I'll let you tour through the spreadsheet to see final player rankings. I'll submit a post in a few minutes with my picks for MVP, Cy Young, & Rookie of the Year.

A few additional comments for people interested in methods:

I say that these are preliminary, because there are a few additional adjustments I still have yet to do.

One, these are based on a base runs equation (for pitchers) and linear weights (for hitters) that are all based on 2003-2007 data. I'm going to calculate new equations that are optimized for 2004-2008 data one of these days. It won't make much difference, but the runs environment is decreasing in MLB and so it's worth doing.

Two, there is some discussion again about whether the most appropriate measure of chances for OOZ plays is innings played or balls in zone (BIZ). BIZ has always been a rather shaky estimate, but I've felt it was more likely to reflect the gb/fb tendencies of a given pitching staff. It may, however, confound the results, because it essentially measures chances in regions where OOZ plays were not possible. Innings is less likely to be biased, but is a bit coarser. I may make that change one of these days. This won't make much difference on a gross level, but it could conceivably make a big difference for some individual players with unusually high or low BIZ for their innings played.

One change I have made is that I'm now using the new position adjustment paradigm developed in this thread at Tango's blog. The net effect is to increase the value of skilled infielders (2b, 3b, & ss), and to decrease the value of the center fielders. CF's now have the same defensive value as 2B's & 3B's. C's also get a further boost in this design, whereas 1B's get knocked down a tad. This seems to be more reflective of the current state of thinking of how to rank the different positions than the old scheme. We're talking about a few runs over the course of the season, though, so this doesn't make a big difference.

11 comments:

  1. Justin, where are you getting your Park Factors from. I have tried to replicate your RAR numbers, but I am using the Baseball Reference multi year factors, which may be different then yours.

    ReplyDelete
  2. I use Patriot's, which can be found here.

    They're not updated for 2008 yet, but they're 5-year regressed park factors.
    -j

    ReplyDelete
  3. Justin, here is a link to my 2008 PFs.

    Link

    ReplyDelete
  4. Thanks for that, p. I'll update my own stuff with those new numbers when I get a chance. Interesting to see the Reds' PF continuing to rise...I still wonder if that's a self-fulfilling prophecy.
    -j

    ReplyDelete
  5. Justin---

    Is the Brad Hawpe line correct?

    Thanks.

    ReplyDelete
  6. BD,

    Yes, it is. ZR has him at -27.6 runs this year, while RZR has him at -21 runs, for an average of -24.3 runs. Dreadful.

    The reason that his fielding + position adjustment doesn't completely negate his offense as you'd expect from summing up those numbers is that I'm employing a maximum negative value for fielding numbers that's equivalent to that of a DH. While Hawpe might have been that terrible this year and thus deserving of being labeled a replacement level, I tend to think that a) fielding numbers deserve some substantial regression to the mean because they can vary so much from year to year (last year Hawpe was ~-5 runs, I think), and b) on other teams, if he played as a DH, he'd have a higher value.

    If you don't agree, feel free to do a straight sum across that row and call him a replacement player! :)
    -j

    ReplyDelete
  7. Thanks Justin. I wondered why it didn't some to ~0.

    Looking over the spreadsheet, I think a player the Reds should target is Gutierrez. The Indians have a pretty crowded outfield, and Gut would look good in center for us.

    ReplyDelete
  8. Zone ratings? Are they as poor a measure for outfielders as infielders? Gak.

    ReplyDelete
  9. Anonymous,

    From a practical standpoint, zone rating-type stats actually do a good job of judging fielders at most positions (CF is probably the worst). There have been many, many studies on this, including my own (see the post comparing fielding statistics in my player value series). I agree that the methodology that underlies them isn't ideal, but they generally will tell the same story as any of the better statistics (SAFE, UZR, PMR, Dewan +/-, etc).

    Furthermore (and more importantly), they're the best freely-available fielding statistic we have access to during the season. We work with what we have.
    -j

    ReplyDelete
  10. I might use these numbers, thanks for posting them.

    ReplyDelete