Monday, November 21, 2011

Ways to Divvy Up Runs Above Replacement

Baseball-reference's formula for runs above replacement for a position player involves adding together several run values: runs above average from batting, runs above average from baserunning, runs above average from reaching on error, runs above average from not-grounding-into-double-plays, runs above average from fielding, positional-adjustment runs, and replacement-level runs. Add these up and you get Runs Above Replacement; divide by (roughly) ten and you get Wins Above Replacement. They also give a rough division of these into oWAR (offensive WAR) and dWAR (defensive WAR), where dWAR is just runs above average from fielding divided by (roughly) ten and oWAR is just the rest. That's one way to divide up the various categories of Runs Above Replacement, but it doesn't strike me as being in any sense the best one. Here are some alternatives.


The first four categories are what I'd call "runs above average from offense." They measure, in the aggregate, how good an offensive player you are, relative to the league-average offensive player. Now this includes runs from running, roughly-speaking, namely baserunning plus double-play runs, but since there's no particularly categorical reason to combine runs from batting per se and runs from reaching on error, that's not a distinction I think is worth making. The baseball-reference "oWAR" statistic adds to this "Roffense" category the positional adjustment and the replacement-level adjustment. That basically gives us production above a replacement-level player at your position. That's also a potentially interesting thing to look at, but it's not just a pure measure of offense.

Another category to consider is "runs above average from total defense." dWAR includes just fielding relative to league average at your position. But if we add runs from fielding to the positional adjustment, then we get a metric which incorporates the idea that league-average defense at shortstop is more impressive than league-average defense at first base. This is what we refer to when we say that, while Ozzie Smith wasn't a very good hitter, he was very valuable to his team because he played shortstop excellently. If he had played first base as well, relative to league average, as he played short, and been the same type of hitter, he wouldn't have been a good player. dWAR misses that distinction.

There's also the following line to draw: between comparisons to league average and adjustments away from league average. Runs from offense, my first group category, is entirely a versus-the-average statistic, as is runs from fielding. Add them together and you've got Runs Above Average. What's left is the positional and replacement-level adjustments, which I would call Runs from Playing. This is where we say that, for instance, in 2005 Jose Reyes was a below-league-average player, with genuinely bad hitting and mediocre fielding not entirely outweighed by solid peripheral offensive stuff (running, ROE, DP's), but he was still somewhat valuable to the team because he spent 161 games playing shortstop.

So we can have the Runs from Offense, Runs from Defense, and Replacement-Level Adjustment division or the Runs Above Average versus Runs from Playing division. Either one, I think, provides a hell of a lot more insight than the simple oWAR versus dWAR division.

No comments:

Post a Comment