One of the unsettling things for fans who are trying to get a grasp on Wins Above Replacement is that the numbers can and have changed over time. Setting aside the 1000's of historical errors in the baseball record book that are being searched out one-by-one, Miguel Cabrera's batting average is and will always be .330. His WAR (for baseball-reference.com) of 6.9 Wins however may change slightly over time.
Why is this? While we want to think of WAR as a statistic like batting average, it is an estimate rather than a precise measurement. Lots of factors are put used in estimating this value and sometimes they change as better estimates become available either due to more advanced research or new data. Now obviously we think it is a pretty good estimate (at least as good as any other measure of player value). Each step has been rigorously researched and justified and is available for you to review and poke holes in (see the link below).
I think an interesting parallel is stock valuation. The techniques used to value stocks in 1980 or even 2005 are different from the techniques used to estimate the value of stocks in 2012. More data is available now, new computing techniques and even newly discovered mathematics. If we were to go back and apply 2012 techniques on 1980 stocks we would have different valuations using 2012 techniques than what we had in 1980 using 1980 techniques, and we'd probably be a lot more accurate.
Another example would be estimates for the size of the earth. This number has been refined and improved over 1000's of years as new techniques for making this estimate have become available. But even now this size is not a direct measurement (there is no measuring tape or scale big enough), but an estimate.
Now the difference is that probably few people care about that difference when it comes to stocks or the size of the earth, but we continue to be fascinated by past baseball seasons. So when we go to a great deal of study to estimate the effect of not having batter strikeout data on the value of outs in early 1900's baseball or the value of an infield single or an IBB in 2005, that affects our view of how valuable that player was in that season. At the time in 2005, we didn't consider IBB's as different from non-IBB's and we didn't differentiate between infield and outfield singles. Now we know the value of those differences and we apply it to our understanding of 2005, 1955 and 1905 baseball.
Previous to this season, we made several large changes to how we calculate player value. They are all listed below in the link, but the major one is the use of Baseball Info Solutions Defensive Runs Saved. Switching from Total Zone to DRS for 2000-present caused some very large swings in defensive value.
Defense is hard to measure as there are dozens of factors that go into its measurement, but we feel confident that their system is the best. They also are continually trying to improve the system. For instance, this past offseason they added batted ball timer data to refine their estimates of player defense. That means every ball in play for a substantial number of years was reviewed and timed (by hand). This then changed the defensive estimates for nearly every player. The stat got better as newer techniques and more data was applied to the question.
And even then if you think all defensive measure are bunk, use oWAR. It is every part of WAR, but assumes everyone is an average defender.
And if you think replacement level is bunk use WAA or wins above average. For single season measures like MVP races it works just as well as WAR. For careers, you'll probably undervalue average players with long careers.
Now could Cabrera and Trout's numbers for 2012 change next year? Yes, park factors are one factor in how batting is considered and we use 3-year park factors, so ideally the 2012 park factor includes 2011, 2012, and 2013, so if Comerica or Anaheim play much differently next year that could cause a change (albeit small--like half a win at most extreme) in their WAR totals.
Let me say one other thing, because of the fuzziness, I would never look at a WAR of 7.6 and a WAR of 6.5 and say the first player is "clearly better than the second". I would say that the first player is "probably or likely better than the second". However in the AL MVP race we have a 4 win difference which as far as WAR goes is huge, so in my opinion (and yours may be different) Trout was clearly a more valuable player than Cabrera this year. And, of course, if playoff appearances to leaderboard troikas are super important for you and overrides whatever else happened in the regular season then WAR isn't really applicable.
This entry was posted on Thursday, October 4th, 2012 at 9:23 am and is filed under Advanced Stats, Announcement, Baseball-Reference.com. You can follow any responses to this entry through the RSS 2.0 feed. Both comments and pings are currently closed.