Hitting is less controversial than all the other avenues of player value. I’m not sure whether this is because fans are just more comfortable with the idea that hitting provides value, or because measuring hitting outcomes is pretty straightforward. Before we talk about what does count as “hitting value,” let’s talk about what doesn’t count:
- Just batting average. Why not? Because it doesn’t really do anything special. It’s just a weird ratio, it doesn’t even tell you how frequently the hitter avoids outs.
- Runs and RBI. Why not? The point of value is to credit a hitter for things he did, not for things anyone else did. The hitter scoring a run, and the hitter driving in a run, are the result of a hitter’s teammates, not him himself. Hitters do get credit for outcomes more likely to score runs (those outcomes are worth more), but whether or not they actually led to a run scoring right there and then is not relevant to a hitter’s own value.
- OPS. Why not? OPS is a weird stat that adds together two things (OBP, SLG) with different denominators. It’s actually not too far off from how hitting value is done, but it’s less clean/elegant than hitting value.
- Stolen bases. Why not? These are accounted for, it just takes place in baserunning. This isn’t hitting, it’s stuff that happens after.
Descriptive Hitting Statistics
When we talk about hitting, the two stats that are brought up most often are wOBA and wRC+. wOBA and wRC+ are actually really similar, as are a couple of stats rarely mentioned offhand (wRAA, wRC). All of these stats basically do the same thing; the basic idea is just to take the different offensive outcomes for a hitter and then assign some value to them. When I say “assign value,” I’m glossing over a pretty complicated process called linear weights, but basically, it works like this:
Imagine you were a computer and could comprehend a whole bunch of information at once. So you decided to create a landscape of every plate appearance in baseball. The ones with none on, two out, the ones with the bases loaded, zero out, the ones with a man on second and one out, and everything else. Then, in that landscape, you ran an experiment. First, you started with outs. You started in every plate appearance, had the hitter at the plate make an out, and then recorded how many runs ended up scoring that inning (after the out). Then, you did singles. Singles score more runs than outs -- not only can they sometimes drive runners in, but they also extend innings and can lead to runs after the single. And then doubles. And triples. And homers. And walks. And hit by pitches. And then, after you did your experiment for every type of outcome, you just averaged the amount of “runs contributed” from the out (note: this is not really going to be any runs!) versus the single, the double, etc.
...Aaaand that’s it. That’s how the value of each offensive outcome is calculated (more or less). Then, to see how valuable a guy is at hitting the baseball, you just take all of his outcomes, give him the assigned value for each outcome (multiplied by how many times he had that outcome), and take an average or summarize it some other way. There are many ways to express this number:
- wRC (without the plus sign) is basically just a counting total of the “runs contributed” (actually it’s weighted Runs Created, to use the official wording);
- wRAA is expressing this same total in terms of runs above/below average -- it compares the hitter’s offensive outcomes to those of an average player;
- wOBA expresses this same number but on the OBP scale, giving a familiar number that kind of looks like a guy’s slash line; and finally
- wRC+ is like wOBA, but handily, it’s put on a really nice and easy scale where 100 is league average, each number above 100 is that many percent above average, and each number below 100 is that many percent below average. It’s also park-adjusted (more on that later).
wRC+ is the one people tend to use most often, because it’s really easy to understand. If I tell you a guy has a .340 wOBA, do you inherently know what that means? I, personally, have no idea. An OBP of .340 means a guy reaches based 34 percent of the time. A wOBA of .340 means… I have no idea. A wRC+ of 110, though? That’s easy, it means the batter was 10 percent above average. Now, hitting isn’t only about the sum total of a guy’s average production. But just like WAR has a great advantage in being a single number that communicates a lot of information, wRC+ does the same thing for hitting. Neat.
I’m going to head off one bit of criticism here, which goes like this: “Wait, why did you only talk about outs above? Why didn’t you differentiate strikeouts versus groundouts? A strikeout almost never leads a run, and in some of your situations above, a fly out or groundout will!” With that said, here’s the reason why outs are mostly lumped together: because it mostly doesn’t matter. In 2015, the run value of a single was 0.7, and the run value of a double was 1.00. That means a double was over 40 percent more valuable than a single! Meanwhile, the highest difference between a strikeout and another type of out, according to work by John Choiniere (see here: https://www.beyondtheboxscore.com/2014/5/21/5733754/selling-out-for-power-home-runs-strikeouts-linear-weights) in 2014 was never greater than 0.1 runs, and had trended another order of magnitude down in the modern game as played in the 2010s. As a result, it would only make a marginal difference if linear weights were calculated for different out types, and for most batters, probably not at all. You can see a bunch of different run values for outs here (http://spiff.rit.edu/richmond/baseball/lwts/lwts_intro.html) and here (https://www.beyondtheboxscore.com/2013/4/1/4165664/how-can-strikeouts-be-great-for-pitchers-but-not-that-bad-for-hitters). Again, you could make a case that maybe we should calculate wOBA and wRC+ using all these linear weights, rather than just the weights for “outs” versus all the different kinds of non-outs. That’d probably be fine, it wouldn’t be worse. But it wouldn’t be much better, either. Further, when you consider that this criticism is usually made to malign high-strikeout, power-focused approaches, and the fact that fly balls are better than grounders in this framework, the high-strikeout guys get back via their fly balls whatever they lose from their strikeouts, relative to someone who hits a bunch of grounders. (And if you ever need to point it out to someone without using the phrase “linear weights,” just remind them that a strikeout doesn’t tend to result in a double play, which is the second-worst outcome possible, just behind a triple play.)
Alright, so, wRC+! It’s pretty great. I do have one bone to pick with it, though, but it’s a fairly small one: park factors. Right now, the way that wRC+ works is by carrying half of a player’s home park factor across their outcomes. A park factor is just a measure of how easy/hard it is to score a run in a given park; the easiest way to think about it is in terms of Coors Field and how batting events need to be devalued there because of the thin air and the relative ease of hitting there compared to other parks. By taking only half of the park factor, wRC+ implicitly assumes that each hitter is affected by the park factor of their own park half the time, and plays in a neutral or average park the other half. In general, this isn’t a huge issue. But it is kind of annoying. If you’re Trevor Story and you hit a homer at Coors Field, wRC+ is like, “Bah, big deal, it’s Coors Field!” and devalues it. But then you go and hit a homer at Petco Park, and wRC+ still devalues it, because your home park is still Coors Field. Meanwhile, if you’re Dansby Swanson and you homer at Coors Field, your homer is worth more than Trevor Story’s homer, because your home park isn’t Coors Field. Again, it probably doesn’t make a huge deal, except potentially for players wearing Colorado Rockies uniforms, but it is something I hope gets fixed in the future.
Anyway, there you have it. If you want to look and see how well a hitter is doing, why bother looking at counting stats, trying to figure out the different relative values of a triple-slash line, or anything else? wRC+ is one simple number that’s already scaled to league average, and even factors the hitter’s home park into account to some extent.
Still not convinced? Well, I’ve got one more thing for you. Remember the WAR mantra: if it wasn’t useful, we wouldn’t be using it? This works really well for this sort of linear-weight based offense, too! Pick any stat you want, and see how well it explains run scoring. Batting average is pretty bad; in 2018, differences in team batting average only explained about half (52 percent) of differences in team run scoring. OBP is better (much better), at 77 percent. Slugging matters a ton, at 87 percent. Strikeout rate (24 percent) is much worse than walk rate (49 percent). But wOBA? wOBA stands atop this pile, at around 93 percent. Again, what we have here is a stat that knows nothing about when the players got their offensive results. Yet even without this knowledge, it correlates almost perfectly to run scoring. (Note: the wRC+ correlation with run scoring is actually not great, at 69 percent, because wRC+ is park-adjusted. Since different teams have different run-scoring environments due to their home park, but wRC+ will scale while wOBA doesn’t, it’s not going to correlate that well. The Rockies won’t have a high wRC+ just because they play at Coors, but they will score a lot of runs. But, since games are always played at the same park, this isn’t really an issue -- wRC+ is still a great measure of hitting success, and if you take the park adjustment out of it, you get wOBA.)
Forecasting Hitting
So, we’re down with wRC+. We can easily tell good hitters from bad hitters now, right? We know that the top 10 percent of hitters in a year have wRC+s of around 130 or above, and the bottom 10 percent are around 70 and below. So, if we see three hitters, who have wRC+s of 90, 100, and 110, respectively, we should know which one of those guys is going to produce above-average offense in the future, right? Right?
Unfortunately, it’s not so easy. If you had no other information, then sure, the 110 wRC+ is the best bet to hit the best in the future. But, we do have other information. So much other information. Up until a few years ago, most of the conversation about forecasting hitting had to do with BABIP. BABIP stands for batting average on balls in play, and basically describes the hitter’s propensity to have his batted balls land for hits within the confines of the field (homers don’t count). The general idea was that for the most part, a ball in play resulted in a hit 30 percent of the time or so, and if a hitter had a high BABIP (say, 37 percent) or a low one (say, 21 percent), this figure would regress, changing the hitter’s batting line in the future even if the hitter made no other changes himself. Similar to BABIP, HR/FB was another pretty useful tool in this regard, under the same principle: most hitters had their fly balls land over the fence at some kind of rate commensurate with their strength and power, so a hitter with an abnormally high or low HR/FB should be expected to regress towards their average rate, taking their batting output in the same direction as homers become flyball outs or doubles (or vice versa). I could spend a while talking about sample sizes, BABIP, and the like here, but it’s no longer necessary, due to yet another alphabet soup metric: xwOBA.
xwOBA, which you’ll have to go to the wondrous Baseball Savant to access, is the “what should have happened” hitting stat. Let’s say you absolutely murder a ball. 90 times out of 100, it screams through the infield and goes for a single, five times out of 100 it’s perfectly placed so that it allows for a double, and the other five percent, sadly, it’s hit right at a fielder’s glove and results in an out. Under the wOBA/wRC+ paradigm, how hard you hit the ball doesn’t matter. You get credit for whichever outcome occurred, whether that’s a single, a double, or an out. That’s it. If you murder 100 balls, you might get a perfect 90-5-5 breakdown, but if you only murder 20 balls, the chances that they’ll be 18-1-1 aren’t that high. (Even for a sample of 100, repeating the underlying pattern perfectly isn’t that likely.) xwOBA, though, is more magnanimous. Rather than giving you the result of what actually happened, it gives you the average value of 90 singles, five doubles, and five outs (which is pretty close to a single). So even if the ball was caught, xwOBA does you a solid, because you killed that ball. It doesn’t help you right there and then, but the fact that you hit an at-’em ball isn’t held against you.
To be more specific and less colloquial, xwOBA looks at two things for each ball in play: its exit velocity and its launch angle. For each such pair, it determines the average wOBA for all balls struck that way. Then, the hitter gets credit not for the result of the play, but for the average wOBA for the batted balls that match the velocity/angle profile. In addition, xwOBA also gives the same values as wOBA for walks, hit by pitches, and strikeouts -- that is, things that don’t result in batted balls. There’s a big discussion of xwOBA here: https://www.talkingchop.com/2018/8/27/17783348/charlie-culberson-mockery-xwoba-statcast-atlanta-braves, and there are links in that article to further xwOBA explorations.
The general idea, and again, the reason why xwOBA warrants mention here, is because xwOBA predicts future wOBA better than wOBA itself. Another way of saying this is that the gap between xwOBA and wOBA for a set of players at a point in time tends to be normally distributed around zero, i.e., the most common thing is to have a wOBA in line with one’s xwOBA. Which means that as more time elapses, there’s more of a gravitational pull to bring the wOBA in line with the xwOBA.
Using xwOBA, then, is really easy. Look at a guy’s xwOBA and his wOBA. If his wOBA is higher, he got somewhat lucky. (Or, if you don’t like the word “luck,” he benefited from random variation.) If his wOBA is lower than his xwOBA, he got somewhat unlucky. Of course, in the future, a lot of things could happen. Maybe a hitter actually starts hitting worse (or better) in the future. In that case, his xwOBA will change too. So it’s important to not get caught up trying to hit or describe a moving target. Rather, think of it this way: xwOBA tells you what the hitter should have done, wOBA-wise (and therefore, wRC+-wise). That means that if you assume no other changes, the hitter is more likely to perform to his xwOBA in the future. If other changes do occur, those will need to be accounted for separately and accordingly.
In general, forecasting isn’t only as simple as looking at someone’s wOBA and xwOBA and going from there. Aging, ability to make adjustments, contact profiles, handedness splits, and many other things drive a hitter’s ability to succeed in successive futures. But, comparing these two stats provides a ton of information, and shouldn’t be ignored. To go back to the prior example with our three hitters, if the 110 wRC+ hitter was overperforming his xwOBA by a lot, but the 90 wRC+ hitter was underperforming his xwOBA by a lot, there’s a good chance that the latter will outhit the former in the future. Ignore this at your own peril.
With that said, here are some things xwOBA doesn’t handle perfectly:
- Faster hitters will have higher wOBAs than xwOBAs, because they can reach base and/or extend singles into doubles on some balls in play where slower hitters can’t manage the same. Many xwOBA outperformers are really fast. If you’re looking at a speedster and he’s consistently outperformed xwOBA by .020 or so, that might just be a thing he can do. But I’d probably wait for a decent sample size before concluding that, either way. [Note: In between me writing this sentence and it being published, the xwOBAs on Baseball Savant have now been adjusted to take sprint speed into account. So you can kind of ignore this bullet, for now, at least until we figure out to what extent we really can and can’t ignore this idea going forward.]
- xwOBA doesn’t care where the ball is hit, only how hard and how high. That means that hitters who are easy to defend via shift will underperform xwOBA, and hitters who are particularly difficult to defend via positioning may be less prone to underperforming theirs. It also means that hitting a cheap homer down the line can result in a lot of wOBA with a fairly low xwOBA (because that short distance would result in a fly out in most cases), but this isn’t a concern as much as a curiosity, because hitters don’t tend to have “hit cheap homers down the line” as a repeatable skill.
- xwOBA still requires some kind of sample size to work. That sample size doesn’t need to be as robust as the one for wOBA or wRC+ (because it takes luck out of the equation), but a week of xwOBA doesn’t necessarily mean much by itself. As with all stats, exercise prudence before you extrapolate. If a guy has 10 plate appearances against left-handed pitching and a 1.300 xwOBA against them, that could mean he’s a southpaw-murdering savant, or it could just mean it was all of 10 plate appearances.
Other Hitting Bits
Hitting is a WAR component for position players (duh). An easy shorthand is that over 600 PAs, each point of wRC+ above/below 100 is worth +/- 0.75 runs. This lets you quickly think about player value. For example, if you have a guy that you think can play average major league defense and run the bases perfectly average-y, but can manage a 110 wRC+, then you can start with your perfectly average 2 WAR, and add 10 x 0.75 = 7.5 runs, or about 0.75 wins to hit ledger. That transforms a generic 2 WAR player with a 100 wRC+ into a pretty neat 2.8ish WAR player with a 110 wRC+. Or, say that you had a defensive and baserunning savant who was worth +20 runs above average in that regard, and you wanted to know how well he’d need to hit to be at least average. With a 100 wRC+, he’d be a 4 WAR player, so if you wanted to allow him to bleed up to two wins with his hitting, that would be 20 runs divided by 0.75, or about 27 points of wRC+. In other words, he could hit at a 73 wRC+ and still be average with his great defense and baserunning.
In keeping with the above, the current WAR paradigm does not care about position; hitting value is hitting value, and 100 wRC+ is average for the league, not average for a player’s peers at a given position. League-average wRC+ by position does vary, and it’s not difficult, if one were so inclined, to create some kind of position-specific offensive metric. But, that’s just not done right now, as it’d be an extra step. Doesn’t mean you couldn’t develop it, though.
Baseball Prospectus recently debuted a metric called DRC+. It’s an interesting concept, and attempts to thread the needle between a purely descriptive stat and a purely predictive one. I find it a bit of an unintuitve fit in the landscape of wRC+/wOBA/xwOBA, just because this landscape already differentiates very well between only things that did happen and only things that should have happened, while DRC+ does a mix of both in one stat. In general, it doesn’t tell you things that looking at these other stats won’t, but that doesn’t mean it won’t have interesting insight about some specific player. It’s only been out for a bit, but I’ve definitely integrated peeking at DRC+ into quick glances to assess hitting ability for a given player, which is pretty much all this primer is trying to explain anyway. I wrote some longer thoughts on DRC+ here if you’re curious. Again, these are just my thoughts, not some kind of fiat judgment on DRC+.
tl;dr takeaway for hitting - Look at wRC+ to determine how good a hitter has been in the past. Look at xwOBA, and the difference between a hitter’s xwOBA and his wOBA, to determine whether he’s been lucky or unlucky; his xwOBA tells you what he’s likely to do in the future based on his hitting inputs rather than his results. DRC+ is also worth a look, as kind of a mix between the two.