Leave OBP Alone
The On Base Percentage may not be a perfect statistic, but it still has its uses.
Concession number 1: The term “On Base Percentage” is a misnomer. OBP doesn’t truly measure all the times a player is awarded a base. It doesn’t include the times a player reaches base due to fielders choices/errors, dropped third strikes, obstruction, or catcher’s interference.
Concession number 2: OBP isn’t a perfect statistic. Weighing all instances of getting on base the same is functionally flawed, and honestly, wOBA, OPS, or OPS+ do much better jobs of measuring value.
I do a fairly poor job of staying out of fights with coworkers, and got into an argument over OBP with Christian “The Fountain” La Fontaine. Christian thinks that OBP should be modified to include all appearances on base, including the instances listed above, but I disagree — modifying OBP as thus shoots it in the foot and makes it useless.
OBP is among the holy trifecta of the slash line, along with batting average and slugging percentage. BA by itself is widely regarded by Sabermetricians as an inherently flawed statistic, but both OBP and SLG are good statistics by themselves, and better when combined as OPS. OBP may seem archaic when associated with something like BA, especially when a superior statistic (wOBA) exists and is easily accessible, but OBP still has its uses.
The big reason to like OBP is that it does an excellent job of measuring the outcomes that batters can generally control. While all plate outcomes are usually the result of a combination between the pitcher, the batter, and the defense, the ones that OBP measures (hits, walks, and hit by pitches) are the ones that the batter has the most influence over. You can debate that getting hit by a pitch is actually more the pitcher’s fault, but the fact that there are players out there who are able to effectively leverage HBP into base appearances seems to indicate that they have a good amount of control over that.
Now, assume that one instead used “True OBP” — the number of base appearances by a player over their total PA. Suddenly, we’re not just keeping track of what the batter can control — we’re keeping track of what the fielders control as well. Errors, fielders’ choices especially. Everything that OBP counts towards is positive for the batter, and provides value for the team. But with True OBP, we’d be counting negative outcomes for the batter. Say that a batter grounds into a fielder’s choice — a negative outcome. But this increases a player’s true OBP. Suddenly, we have a stat where both good and bad outcomes have the same effect on the stat. This neuters the stat completely in using it to assess a player’s value.
In using true OBP, we’d also be counting errors positively towards a player’s stat. While errors themselves are not negative outcomes for a batter, they only provide positive value because a fielder messed up somewhere along the way. Essentially, the batter provides negative value for his team by, for example, hitting a lazy pop fly, but the fielder provides contrasting negative value for his team by dropping it — so in terms of isolating and then assessing a batter, hitting into an error should provide identical value to hitting into an out. Yes, there are some batters who get on base from errors more frequently than others. But that’s not because of players forcing errors somehow by running hard and forcing fielders to hurry. It’s more the result of putting the ball in play more often in ways that cause fielders to mess up — which still doesn’t necessarily provide value, as ideally, competent fielders would make the play.
The same holds true for dropped third strikes, obstruction, and catcher’s interference. Most of them are instances of batters creating negative value for their teams, only to have them be counteracted by negative defensive value from their opponents. If we truly wish to be able to compare players in a vacuum, then we should look only at the value, positive or negative, that they provide — not that which is gifted to them by their opponents’ ineptitude.
As my friend Matt Martell pointed out, however, traditional OBP counts hits that might have been robbed by a better fielder — in the example given, Manny Machado robs a player of a hit that would have gotten by a lesser defender — for instance, Juan Uribe. But Juan Uribe isn’t charged an error, and the player would have ordinarily provided positive value had Manny Machado not taken it away. There isn’t a good system to keep track of all those would-be-hits that are caught by top tier defenders, unfortunately, so there’s no good way to fix that — we just have to let the web-gems be web-gems.
Personally, I would love to have a “Robbery” stat, the anti-error: a defender makes a play and gets an out that would have been a hit otherwise. Robberies could count towards OBP in this perfect world. But the criteria would be too difficult to define to use Robbery as a statistic, so we’re left with the current system. It may be flawed, but it still does a damn fine job of measuring a player’s value in terms of their ability to get on base, even if it undervalues them slightly.
So again, by counting reaching on errors, both positive and negative outcomes are moving the True OBP needle further and further up, and any ability that it has to assess value accurately goes away. So True OBP doesn’t make sense as a traditional statistic, nor does it fit anywhere in the slash line. Even as a “results based stat,” True OBP doesn’t really fit anywhere, since it’s an amalgamation of positive and negative outcomes — usually a stat should measure good or bad, not combine the two into a single number.
Ultimately, OBP does have its drawbacks, but it still functions well as a stat. Both OBP and True OBP fall victim to the confounding variable of the defense. But True OBP intentionally gives more weight to that confounding variable, the opposite of what a statistician would attempt to do — making it useless.