More on Ryan Howard’s WAR

This is an addendum to the recent “Understanding Ryan Howard’s WAR” post. Many people were receptive to the ideas when they were laid out step-by-step and it’s great to see more and more people taking an interest in understanding Sabermetrics. One point, which I failed to address, that seems to be sticking in the craw of a lot of holdouts, is that WAR is a descriptive statistic and not a predictive statistic. What that means is WAR will tell you what happened; it will not tell you what is likely to happen in the future. Just because a player has a higher WAR in one season does not mean that WAR is saying the player is, overall, a better player or that the player will be better going forward.

Dennis Deitch of the Delco Times wrote on Twitter:

Mark Reynolds WAR: 2.4

Ryan Howard WAR: 1.9

It just debunks that stat. It’s not really credible.

The WAR stat needs some serious, serious work. Something is seriously wrong with its computation.

I don’t mind the math or the complexity.But it needs to be indicative of real value. Something in the equation is lethally flawed

As you can see, Deitch looks at Reynolds having a higher WAR than Howard and immediately concludes that WAR is saying that Howard is a worse player. That is not what WAR is saying, and I can’t really fault him for thinking that because it is not blatantly obvious.

Bat Field Rep Pos
Howard 18.6 -9 18.5 -9.7
Reynolds 4.4 -1.5 18.5 1.9

One can avoid those logical pitfalls by taking time to understand the statistic (or asking someone who does) and by performing a bit of research. To the right is a breakdown of each player’s WAR.

A lot of Reynolds’ value comes from playing third base, and playing better defense at his position than Howard does at his. Given what we know about UZR — that it is not entirely reliable until you have about two seasons’ worth of data — we can take those evaluations with larger error bars. There’s a 7.5 run difference between Howard and Reynolds but it could very easily be, say, Howard at -3 and Reynolds at -5.5 which is a shift of about ten runs, or one WAR.

Additionally, there’s nearly a 12 run difference between the two just from the positional adjustment. As we learned on Saturday, third basemen are credited 2.5 runs and first basemen are debited 12.5 runs (the numbers are adjusted slightly to reflect playing time).

What we learn from this is that Reynolds is credited a lot just for where he stands on the diamond, regardless of what he does with the bat. This is likely where the stat-heads and traditionalists diverge. There is nothing wrong with this — we are measuring value, whereas traditionalists tend to what to measure ability. WAR is not saying that Reynolds is as skilled a player as Howard; in fact, Howard is in many ways vastly superior. But Howard spends his time at first base, suppressing his value.

We can compare the players’ abilities. Offensively, Howard dwarfs Reynolds with a .391-to-.355 career wOBA advantage. Howard hits for much more power with a .296-to-.245 career ISO advantage. Reynolds both walks and strikes out more, as he is one of the kings of the “Three True Outcomes“. Defensively, Howard really isn’t as bad as UZR has shown him to be this year. Over his career, Howard is at -2.1 UZR per 150 defensive games whereas Reynolds is at -7.1 at third base.¬†Howard has clearly been the better player. Even WAR agrees with that: since 2007, Howard has a WAR advantage of 13.9 to 8.9, or an average of 1.25 WAR per season.

WAR is telling you that this year, Reynolds has been more valuable. That may be the case going forward (especially if Reynolds sticks around at third base), but WAR has no way of telling you that. It is simply saying that, given the evaluation of each player’s offensive and defensive contributions, adjusting for each player’s position, and considering replacement level, Reynolds has been a slightly more valuable player this year.

When you look up WAR and you find some fishy rankings, dig a little deeper to find out why. Ask questions when you’re not sure. Don’t cherry-pick the outliers and call for immediate disposal of a perfectly useful and reliable statistic. Criticism is absolutely warranted and always welcomed (and entirely necessary!) but it needs to be justified by good and faithful science.