Why I Won't Use WAR or WARP in Baseball Articles, Ever Again

Someone pointed out to me, a few weeks ago, that I use wins above replacement (WAR) and its cousin, WARP, less often than most writers with whom they typically think of me as a kindred spirit. I was a little bit surprised, and a little bit alarmed. WAR is one of the fundamental concepts of much online baseball writing, and I hadn't noticed that I was neglecting it.

It took me another week of writing to notice and confirm that that was true. The next step I tried to take, going into the following week, was mapping out how I might get back to making greater use of the framework for audiences.

The more I tried to do so, though, the less enthusiasm I felt for it. In fact, at this point, I've come to the opposite conclusion. I don't think WAR is helpful or useful--or at least, I think it misleads as much as it informs, and I'm done using it, even to whatever limited extent I had been in previous months or years.

I don't want to sound retrograde, or atavistic. Early in the era of WAR, some old-school writers hit back against the framework based on things like the existence of competing WARs, with differing fundamental assumptions and widely disparate numbers for the same players. Some rebuked it for the inexactitude of a replacement level, itself. I don't completely disagree with those arguments, but they're not my reason for leaving WAR behind, and I don't think they're especially problematic, really.

Rather, I have a couple of other issues.

1. WAR values have become blunt instruments, not for loosely estimating player value, but for ending conversations by assigning false certainty to player valuations. 
2. Runs, not wins, are the directly measurable contributions to a team made by individual players. Wins are fashioned from the interactions of players, managers, moments, and opponents, and players' contributions translate only very messily into wins. Counting runs produced and prevented is a far superior way to express player value than doing the same thing, then milling those runs into theoretical wins and roughening the estimate without showing the increased error.
3. For my money, replacement levels aren't the right baseline to which to anchor player value, after all. The old heads were right about that, but for the wrong reasons. They were trying to cling to raw numbers, like pitcher wins, home runs, and RBIs; that was analytically untenable. WAR sought to anchor the world to the baseline of a replacement-level player, which is convenient for teams and from the perspective of management vis-a-vis labor, but what we need to anchor ourselves to, instead, is the league average. When you compare a player's batting, baserunning, and fielding to an average player's, you find out how many runs they really contribute to the project of reaching the postseason for a team. Taking the extra step of adding what amounts to padding--credit applied to a player relative to the replacement level, on the unexamined assumption that they were better than the possible replacements actually available to their team--valorizes below-average players and dampens the apparent value of above-average ones.
4. The existing WAR metrics are too rigid; they adapt only too slowly, and thus sometimes retroactively. That doesn't make them useful to me in telling the story of a player or their value for readers. This season, first basemen are hitting at a historically lousy rate, relative to the rest of the league. In theory, according to WAR's underpinnings, that should make the offensive contributions of especially good first basemen especially valuable, and it should make the struggles of some of those players a bit less damaging than they seem. Alas, the first "should" there is only hypothetical, because the positional adjustment that is one key aspect of WAR doesn't flex the way you might think or like, at least in all cases. The second "should", meanwhile, is just an assumption I don't really agree with.
5. There seems to me a logical inconsistency in the places where the various flavors of WAR do and don't elect to depart from observed reality--or, perhaps, in how they adjust it. This is most readily apparent in pitching WAR, which ranges from being rooted in runs allowed and only lightly bumped along the spectrum based on park factors and defensive support to being rooted almost solely in strikeouts, walks, and home run rates. But it's really everywhere within the framework. I don't like the opacity of the process, even though it's not left intentionally opaque by any of the sources who provide it and even though they're not opaque to me, personally. I think these layers of assumptions and adjustments, baked in neatly en route to a single-number estimate which fans then treat as something real and concrete and non-negotiable, end up doing more harm than good--even though most of them are immensely valuable, if kept separate and noticed along the way.

I would rather continue to break down player performance using more telling indices, with either greater predictive or greater descriptive power, than try to speak the language of WAR again. Finding that I'm out of practice in that tongue turned out to be liberating. What WAR is trying to do, and it's admirable enough in a certain way, is to shrink the complexity and the difficulty of player evaluation and roster construction, until those abstract tasks become apparently concrete and easy to grasp. Unfortunately, as most attempts at such radical simplification do, it fails.

I would rather live here, in the discomfort of uncertainty and abstraction, than seize upon the cozy but false sense of surety WAR offers. I'd rather continue to help interested readers learn more about players and how they come together to win games than invite them to continue conforming to a constructed reality I feel does a poor job of capturing the one on the field.