Context Batting Runs

If you use FanGraphs regularly, you’re probably aware of the WAR framework, and the general ideas behind it. Basically, WAR attempts to sum up the value of a player’s hitting, baserunning, and defense and then compare it to what a replacement level player would have done with the same amount of playing time. While WAR is certainly not perfect, it works pretty well, and it tries to answer the question that baseball fans are frequently asking.

But it doesn’t every question, of course, and sometimes, we ask questions that WAR wasn’t designed to answer. For instance, questions of context are outside the scope of the metric, as it was intentionally created to be context neutral, identifying just the number and value of positive and negative events without including the situation they occurred in. WAR considers every home run to be equally valuable, whether it comes with the bases empty in a blowout or the bases loaded in the bottom of the ninth.

When doing player comparisons, you’re usually better off with a context neutral metric, since history has shown that there’s no real tangible skill being measured in “clutch” or “unclutch” situations. Context specific metrics include a measure of randomness, basically, and we’re more often concerned with how good a player is rather than how good a player is and in what manner his events were randomly distributed. If a team is trying to decide whether to sign a free agent or make a trade for a certain player, they are better off with context neutral measures that more directly measure what the player did without the influence of his teammates.

However, there are times when we’re not as concerned about isolating a player’s individual contributions to run scoring, and are more concerned with telling the story of what actually happened in the past. After all, a three run homer counts for more runs than a solo home run, and a team who hits a bunch of three run homers will score more runs than a team that hits a bunch of solo home runs. While it might not be a skill, the distribution of events matter. Single-single-homer is better than homer-single-single in terms of outscoring your opponents.

This is, essentially, why RBIs are still so popular. While the metric itself has all kinds of problems, it measures context-specific performance, and specifically, quantifies our memory of a specific hitter doing a thing that created runs for his team. We can talk all day about how it fails to measure the disparate number of opportunities that players receive, but we’re never going to take away the human desire to quantify how well a player did when he had a chance to put runs on the board for his team.

Thankfully, David Appelman has coded a bunch of statistics into this site that do take context into account, and there’s one stat in particular that does a fantastic job of measuring the idea that drives the popularity of RBIs; that stat is (unfortunately) called RE24. I say unfortunately because, while the name does tell you that it’s based on the run expectancy of the 24 possible base/out states, it still sounds like an error code that your oven gives you, which results in a very expensive service call from the one remaining appliance repair guy left in town. While people make fun of our affinity for acronyms, even I don’t like saying RE24, even though I really like the stat itself.

So, of late, I’ve taken to calling RE24 by another name: Context Batting Runs. This isn’t any kind of official name change announcement for the site or anything, and I’m pretty sure everyone else will go on calling it RE24, but I like the term context batting runs. It says what it is in plain english. I guess if we were being technically, it would be Context Batting Runs Above Average, and maybe I should start referring to CBRAA that so we can all pronounce it Cobra. That would be a fun conversation to have around people who have no idea what’s going on: “His WAR is +5, but once you adjust for the Cobra, he’s off the charts!”

Anyway, enough of that rabbit trail. This post isn’t about renaming RE24 — though I’m in favor of that, I think — but about promoting the idea of what it measures. Too often, I think the mainstream audience and statistically inclined writers end up talking past each other simply because one side puts a large value on context specific metrics and the other does not. Instead of telling them why their context specific metric is lousy, we can instead just offer them a better one.

For instance, here’s a plot of the 150 qualified MLB hitters from this season, with the batting runs component of WAR and then RE24 (Cobra!) on the same graph. Because both of them go negative for below average performances, I’ve stripped out the axis labels, so this plot is more just to show the relationship between the two measures.


That’s a pretty linear relationship — the r squared is .86, for those who are into such things — and shows that most players offensive performance with context included is nearly identical to their context neutral numbers. It also shows just how far ahead of everyone else Miguel Cabrera is (he’s on the far right), which is kind of fun. But you’ll note that there are some points that are pretty far from the line, which tells us that those players have been either significantly better or worse in specific base/out situations than their raw batting line would suggest. Let’s dig into a few of those outliers.

First, the positive.

Player Bat RE24 Difference
Allen Craig 22.7 45.1 22.4
Adrian Gonzalez 13.8 32.9 19.1
Brandon Phillips -4.6 14.1 18.7
Freddie Freeman 24.3 42.8 18.5
Chris Davis 50.3 65.9 15.6
Carlos Santana 18.5 32.6 14.1
Mike Trout 50.7 64.3 13.6
Yoenis Cespedes -3.5 10.0 13.5
Paul Goldschmidt 35.4 48.4 13.0
Robinson Cano 24.6 36.0 11.4

You’ve probably heard about Allen Craig’s crazy performances with runners in scoring position during his career, and RE24 rewards him for his absurd skew towards hitting in scoring opportunities, as his RE24 is twice as high as his batting runs total. It isn’t predictive, and it’s almost certainly not something Craig can keep doing in the future, but the Cardinals have benefited heavily from Craig’s performance in those situations. If you’re wondering why they keep winning even with a line-up of guys that don’t look to be having amazing seasons from first glance, it’s performances like Craig’s that are driving the Cardinals offense this year.

The rest of the list is fun too. The controversy in Cincinnati about Brandon Phillips or Joey Votto feels contrived, but the people advocating for Phillips aren’t incorrect about the fact that he’s been a fantastic run producer this year. Same thing with Freddie Freeman in Atlanta and Paul Goldschmidt in Arizona. While mainstream writers are going to use RBIs to tell the story of their clutch hitting, the story in and of itself isn’t wrong just because the metric they’re using is poorly designed. Freeman and Goldschmidt have been significantly better in those situations, as has Adrian Gonzalez, whose context neutral numbers make him look like a disappointment.

Now, for the other side of the coin.

Player Bat RE24 Difference
Josh Hamilton -1.7 -10.4 -8.7
Nick Swisher 6.9 -2.6 -9.5
Michael Young -2.5 -12.1 -9.6
Howie Kendrick 8.1 -1.6 -9.7
Mike Napoli 6.9 -2.8 -9.7
David Freese 2.1 -7.8 -9.9
Gerardo Parra 0.0 -10.0 -10.0
Chase Headley -0.5 -11.3 -10.8
Adrian Beltre 29.2 17.5 -11.7
Carlos Gonzalez 24.0 11.8 -12.2

Congratulations Josh Hamilton, you’ve been even worse than people think! As we talked about yesterday, the Angels have been much worse than you’d think based on their raw performances, and Hamilton’s struggles in run producing situations are one of the reasons why.

However, no one has been less effective at creating runs relative to their overall batting line than Carlos Gonzalez. If you look at his numbers, along with the stats being put up by Troy Tulowitzki and Michael Cuddyer, it’s hard to understand why the Rockies aren’t leading the Majors in runs scored. Instead, they are tied for 13th, and the struggles of Gonzalez to replicate his overall line when there are people to drive in is one of the factors in their relative struggle to score runs.

Now, I would not make a leap from this data to say that Carlos Gonzalez is “unclutch”, or that you should intentionally walk Allen Craig every time he comes up because he’s just going to drive in all the runs otherwise, but if we’re trying to describe what has happened in the past, the gap between RE24 and a player’s Batting Runs gives us a pretty good indication of the value added or lost due to situational performance.

So, if you’re talking with a friend who likes to quote RBIs, point him towards RE24 instead. It’s getting at the same idea that he’s trying to measure, only it provides a more effective way of valuing those performances than simply counting up all the players who scored. And if someone wants to tell the story of Allen Craig or Brandon Phillips or Carlos Gonzalez this year, well, you can make a pretty good case that the story is better told with context batting runs than just looking at their raw lines.

Print This Post

Dave is the Managing Editor of FanGraphs.

56 Responses to “Context Batting Runs”

You can follow any responses to this entry through the RSS 2.0 feed.
  1. Jaack says:

    Dave Parker’s career Cobra was 384.28.

    Because it was necessary to tell you this.

    +14 Vote -1 Vote +1

  2. William says:

    I’ve been thinking about this for a while, and this might just be basis but it seems like the big home run hitters are actually more valuable then context neutral stuff makes them out to be, and light hitting batters even less valuable

    Vote -1 Vote +1

    • Detroit Michael says:

      I believe that the opposite is true. Using this site’s “clutch” statistic, I think that sluggers as a group do worse than a context-neutral analysis might apply and those pesky left-handed contact hitters that people call “professional hitters” really do as a group do better.

      If you are interested in exploring it, look for threads on the Inside The Book blog.

      Vote -1 Vote +1

  3. tz says:

    How about “Offense Versus Average”? Maybe not for readers here, but for general consumption by the average fan.

    Vote -1 Vote +1

  4. #6org says:

    was all this really necessary?


    -69 Vote -1 Vote +1

    • Avery says:

      Yeah it was.

      It solves a statistical and a semantics problem. It solves the statistical question; what players play at a level above their context neutral statistics say they should in “high-leverage” situations? It also helps to answer why sometimes the context neutral statistics can be “misleading” or can differ from the reality they are supposed to “predict.”

      Moreover, it gives a way of bridging the traditional baseball world, with the one here at Fangraphs and other similarly slanted sites. Stats like Re24 help bring traditional baseball into the language of sabermetics.

      Vote -1 Vote +1

  5. Dan Farnsworth says:

    I’ve often thought this metric or one like it would be a better number to use in comparisons of offensive contributions for the MVP voting. Just like with All-Star selections, I would like to see the player who has had the best year win the award, not necessarily the most talented player. Yeah, I want the most talented players on my team and recognize when luck or context isn’t working in their favor, but we can recognize a player’s great year even if there is a 0.00001% chance of it ever happening again.

    While context-neutral statistics are much more predictive, end-of-year awards are not meant to predict who will be better next year or the rest of their career. Plus, as you mentioned, it would be a much better bridge to the traditional stat-lovers who sometimes only see context. Baseball is a context-dependent sport in every way; why not embrace it?

    +17 Vote -1 Vote +1

    • Baltar says:

      I don’t really care about converting the majority. I have had a number of typical fans thank me for telling them about stats such as a pitcher’s strike to ball ratio, mostly in the context of watching a game together. If it interests them, I am happy to do it. If it doesn’t, I don’t care.
      I definitely don’t see RE24 as a good starting point for those who are interested. Most people readily grasp more important stats such as wOBA.

      Vote -1 Vote +1

    • Nathan says:

      I like what you’re saying, but I still think it is a combo meal.

      WAR and its components are still measuring things that actually happened (let’s just leave this at non-pitchers for now and save that debate for another time). I think it is important to look at both the context-dependent and context-neutral numbers. To be honest, I still put a lot more stock in the context-neutral numbers, simply because they are more appropriately crediting (or not crediting) the player for what he’s actually doing.

      A comparison this year between Chris Davis and Miguel Cabrera is a pretty good example. The context-neutral numbers show us that Cabrera has been (and is) a markedly better hitter than Davis. So much better, in fact, that when you look at something like the Clutch score or WPA, Davis’s advantage seems pretty irrelevant to me because Cabrera’s baseline context-neutral numbers are that much farther ahead anyhow. The RE24 number sort of shows this, as Cabrera is still ahead here.

      In a tighter race — let’s say instead of a 20+ point gap in wRC+, there was a gap between the two of 5 or less — then I’d be more inclined to look at context-dependent stats and use them as a determinant.

      Of course, I’m only leaving out Mike Trout for the sake of the discussion and the reality that Trout is not going to be in the running for the MVP.

      (Let’s take a second just to say, holy whatthefuckmiguelcabrera. Chris Davis is posting a 2nd place wRC+ of 180, which is only 2 points up on Trout, but an insane 23 points better than Joey Votto. Almost any year, he’d be far and away the best bat in the game. But Cabrera is posting an “OMG HOW”-inducing 205! The delta from Cabrera to Davis is bigger than the delta from Davis to Votto. HOW?)

      Vote -1 Vote +1

  6. Bab says:

    Would be interesting to compare these numbers with respective players’ scouting reports on ‘makeup’ etc.

    Vote -1 Vote +1

  7. Green Mountain Boy says:

    I found this article very interesting, because I’ve had an idea for some time now that starts out with the basic premise that every base is worth .25 runs. So if your game leadoff hitter gets a single, he has acquired 1 of the 4 bases available to him. That results in a score of .250 for that AB. If he then steals second, his score becomes 2 of 4, or .500. If he is thrown out stealing, he costs one out, or 4 bases, making his score 1 of 8, or .125. Let’s say he stays put. Batter #2 now has 7 bases available – his 4 plus the remaining 3 from the leadoff hitter. He strikes out, he’s 0 for 7, and so on and so on, until you build up a database of ABs that add up to a percentage of how many bases he gained vs how many were available. Walks count too, as the leadoff walk is in effect the same as the leadoff single.

    This tells us something, but game context means something too, as the article says. What I’ve struggled with is how to properly factor in the inning, the run differential at the time of the AB, and the number of outs. These count too, as a walk-off tie game HR in the ninth would seem to indicate more “clutchness” than the same HR in the first inning, and should result in a higher AB score for that batter. Likewise, the same HRs have a different context if the batter is on the Astros in September vs being on a team in a tight race. To really get it right, who the HR was hit off would have to be factored in too. It’s one thing to do it off Kershaw in the ninth, but something else entirely if it’s off an AAA pitcher making a spot appearance. And don’t forget about park factors… it’s enough to make your head explode.

    The point is though that each base attained equalling .25 runs would seem to have merit as a starting point, and it seems that the same type of calculation could be made for any pitcher (in reverse), giving you a pitcher/batter context. How to factor in the standings, the time of season, and the inning/score/outs is way beyond me, but I’d love to see someone with more time than me and a statistical background pursue my idea and see what they come up with.

    Vote -1 Vote +1

    • Youthful Enthusiast says:

      Sounds like you’re looking for a 3D Run expectancy matrix with baserunners, outs, and score as the 3 inputs. However, before anyones runs off to create that, I’d have to ask, does the relative score affect the traditional RE matrix? If not, then the existing RE24 and linear weights tell the story well enough. If so, then we’ll have a new RE24, probably should call it RE168, and we can re-calculate the RE base on that to see if there is any additional contextual production.

      Vote -1 Vote +1

    • Bip says:
      FanGraphs Supporting Member

      One of the problems with your idea is it assigns each base the same amount of value.

      With bases empty
      Out: 0/4
      Single: 1/4
      Double: 2/4

      This suggests that the difference in value between a double and a single is the same as the difference between a single and an out. In fact, this is not the case at all. There is a huge increase in value in changing an out to a single, while there is a relatively minor difference between a single and a double.

      RE24 basically already accounts for this. It uses run-expectancy, which tells you how many runs a team is expected to score in each base-out state. So the value added from a bases-empty, no out single is the difference in run expectancy between bases-empty and none out and man on first with none out. This essentially does what you’re saying, but using values found from real games.

      Vote -1 Vote +1

      • Green Mountain Boy says:

        Agreed, but where is the context? You’re using averages. The context of Dave Roberts stealing 2nd against the Yankees in 2004 against Mariano Rivera is WAY different from Joe Blow stealing 2nd off Justin Case in a meaningless April Game bewteen non-contenders. Until someone finds a way to measure context, to measure a player’s competitiveness, his heart… averages mean squat. As much as I love sabermetrics, this is their weak spot.

        It’s like the sacrifice as a strategic play. In the first inning with 1st and 2nd no out does it make sense (see Hobson, Butch) vs in the top of the ninth, same situation? Hell no! One is wasting a chance at a crooked inning, the other is putting the odds in your favor to win a game.


        Vote -1 Vote +1

        • Nathan says:

          I don’t think anyone here (other than trolls) would suggest that a player’s “competitiveness, his heart” mean squat. I think the point is two-fold:

          1) Many numbers inherently measure this by measuring the results of a player’s performance. It is irrelevant if David Eckstein looks in the mirror before a game and truly has this unwavering emotion that drives him to compete his absolute hardest if that “heart” doesn’t combine with talent to actually produce. As a counter-example, guys like Ty Cobb and Rickey Henderson are widely considered to be of the fiercest competitors in the history of baseball. They were also great talents. The results (whether you look at BA, OBP, wRC+, WAR, SB, etc.) are there. These numbers are essentially a measurement of everything that made these guys great ballplayers, and that includes their physical skills, mental makeup, and whatever intangibles you like to consider.

          So, I don’t want to be totally cold here, but I really don’t care what a player’s “heart” for the game is like if that player isn’t producing. And if a player is producing, that is what I will judge the player on — it is entirely irrelevant to me if the player is producing through pure physical superiority, a honed skill/talent, or by trying harder than everyone else.

          2) If a player only produces meaningful results in specific situations, then really, how good is that player?

          Take a hypothetical NL pitcher that sac bunts almost every time he’s at the plate, and has an absurdly high success rate at executing the sac bunt. That pitcher is producing a meaningful result in a very specific context. But the only reason the pitcher is producing that meaningful result is because he’s being put into a context where he can do something marginally better than utterly fail. This pitcher only produces an okay result at the plate under very specific circumstances because he is simply not good with the bat.

          Let’s look at a less pessimistic context. Let’s say we have an average hitter that is well above-average with RISP. This is a context-dependent measurement. (I would argue that this player is probably better with RISP because he’s not good enough to draw intentional walks, and the presence of the runners changes the approach of the pitcher and defense in a way that suits the batter’s tendencies.) That said, let’s grant that this player is much better with RISP in some large part because at some psychological level, he’s able to focus better in these high-leverage situations than he is in normal situations.

          If we could measure that “heart” in this high-leverage situation, what does it matter? I fail to see how we’d value this player any differently than we do by looking at the fact that he’s an average hitter by context-neutral measurement. I see it as objectively true that given no outside constraints (e.g. financial), every baseball team would rather have the player that performs at a higher level across more contexts (or neutral to all context!) than the player that almost requires certain situations to be valuable/productive to his team.

          The player that has the mystical ability to go from a true-talent 100 wRC+ player to a 150 wRC+ player with RISP is inherently a pretty flawed player, because just like the pitchers hitting in the NL, you only get optimal performance out of him under hard-to-control circumstances. Continuing on the assumption we could measure “heart” in some way, let’s say this player has more measurable “heart” than a Miguel Cabrera clone that has off-the-charts context-neutral stats. So what? The Cabrera clone, with his lesser “heart,” still has a much stronger baseline of performance at the plate — his results are still inherently better in every situation. As a manager, you have to hide Mr. RISP, or at the very least, you have to rely on your better hitters to put Mr. RISP in situations where he performs. For Cabrera clone, you just hit him somewhere in the top 3 to maximize his PAs, and count on large levels of success no matter the context. Who is the better player?

          As I said in another post above this one, I appreciate context. As an example from this year, if Cabrera and Davis were something like 5 points apart in wRC+ (instead of a whopping 25 points), then I’d say to myself, alright, I need something to differentiate. That’s a spot where I’d ask myself, who was better in high-leverage situations? Even though the context is out of their control, when splitting hairs, I’d want to know who performed better in those tough spots, and I’d give them due credit for making it happen. But as it is, Cabrera’s context-neutral baseline is so much better than that of Davis that there’s no reason to look there. And if you look at context-dependent numbers like RE24, or Clutch score, or WPA, or WPA/LI, they actual bear out what I’m saying. Cabrera is still having a significantly better season than Davis.

          To be clear — I see TONS of value in context-sensitive stats. Just not that much value in using them to objectively measure which players are the best players, have been the best players, and will be the best players. I think context-dependent numbers are probably hugely valuable to actual ballclubs because they can help address real-world problems that exist when building or managing a team (e.g. I have a budget and need to build a team within it; talent is a scarce resource; so I’ve got to be able to figure out how to piece together a team of players that are good in complementary contexts so I minimize the gaps I have).

          Vote -1 Vote +1

    • Baltar says:

      I had ideas like that 57 years ago when I first became interested in Major League Baseball, but I let them slide when I entered my prime years–teens to thirties–when other things interested me more.
      I am very happy that, now that I’m old and those other interests are dropping out due to necessity, other people have refined and are refining those ideas with much more diligence and precision than I can summon.

      Vote -1 Vote +1

      • VB says:

        I think what Green Mountain Boy is looking for is a hybrid between WPA and RE24. This would be kind of cool.

        Vote -1 Vote +1

  8. Guest says:

    Thanks, Dave! Great stuff. Readers of this site need to pay way more attention to this metric when evaluating past performance for awards voting purposes. Now calculate a WAR number heavily weighted with RE24 and we really have something here. Context matters!

    Vote -1 Vote +1

    • Guest says:

      Also, do you guys calculate this number for pitchers as well? That would be great stuff, too.

      Vote -1 Vote +1

      • cass says:

        Pitchers don’t need it because they generally create their own context, at least in the case of starting pitchers. RA9-WAR is about the closest equivalent I can think of.

        Vote -1 Vote +1

      • tangotiger says:

        Fangraphs shows that already.

        However, except for mid-inning pitching changes, RE24 is exactly proportionate to runs allowed per 9IP.

        Vote -1 Vote +1

  9. TKDC says:

    Regarding Josh Hamilton, you could say the exact same thing about Chase Headley, only his difference is even worse! But Josh Hamilton is a free swinger who doesn’t walk much, so let’s point at him.

    Vote -1 Vote +1

    • Gob Hamilton says:

      Right, the guy with the $125M contract takes a back seat to the guy who’s still arb eligible. Come on!

      +5 Vote -1 Vote +1

      • TKDC says:

        So Hamilton is the whipping boy because he took advantage of free agency to receive a lucrative contract? I’m sorry, but that is an even worse reason.

        Vote -1 Vote +1

  10. ralph says:

    I agree, Allen Craig pretty much has to regress here. He’s hitting .452/.500/.637 with men in scoring position, with a .473 BABIP. So obviously that won’t continue, even if he continues to strike out almost half as often with RISP as he does with the bases empty.

    About a month ago, I was looking at Craig’s splits for Bases Empty versus RISP, and one thing stood out at me as something that could represent a conscious and controllable change in approach — a drastically different K%. I had mentioned this K% phenomenon in a comment back then, so I’ve updated the numbers (amazingly, the split has actually grown slightly) and expanded my thoughts a little here.

    2013 K%
    Bases empty: 20.5% in 283 PA (105 wRC+ from a .262/.322/.400 with a .313 BABIP)
    RISP: 10.3% in 146 PA (216 wRC+, from that .452/.500/.637 line with that .473 BABIP)

    The really interesting thing is that it’s not a totally new behavior, as he showed similar tendencies last year as well:

    2012 K%:
    Bases empty: 20.1% in 264 PA (124 wRC+ from a .289/.330/.490 line with a .333 BABIP)
    RISP: 11.4% in 149 PA (199 wRC+ from a .400/.450/.680 line with a .393 BABIP)

    Adding up these past two seasons, and he’s 20.3 K% in 547 PA with the bases empty and 10.8 K% in 295 PA with men in scoring position.

    Strikeout rate stabilizes around 60 PA for hitters. So it does look like that we can say with some confidence that Craig is intentionally striking out less with RISP.

    The best counter-evidence to expecting such a disparity in strikeout rate to continue is the simple question of why Craig doesn’t decide to stop striking out with the bases empty. It’s a really good question, so I’d be curious to hear if people believe the K% difference is actually a random event.

    Vote -1 Vote +1

    • Mike says:

      You’ve just provided stats for nearly 2 years worth of performance for Allen Craig. Not only is he consistently better batting with runners on, he has been consistently better, and getting better over those nearly two years. Maybe he has discovered some particular skill that is replicable in runner on situations, or been taught it by the Cardinals hitting coaches (The Cardinals as a team are leading the league by a wide margin with runners on, and have been for several years).

      I see no sign of regression, nor do I expect one.

      Vote -1 Vote +1

      • Jason B says:

        “I see no sign of regression, nor do I expect one.”

        Can we wager? As much as you’re comfortable?

        +7 Vote -1 Vote +1

      • ralph says:

        Curious what you’re expecting not to regress? The absurd BABIPs with RISP? Or just the K% disparity?

        I have a hard time believing either will last (especially the BABIPs).

        I could maybe see the K% disparity lasting, but it seems more likely that if he really can cut his K% down that low at will and still put up a slash line with acceptable power, someone will eventually tell him to take that approach in non-RISP situations as well, so that the regression actually comes from his overall K% coming down.

        Vote -1 Vote +1

      • Hank says:

        Same thing with pitches too. While fWAR can be useful in pointing out sustainability of a pitchers performance I think an RA9 type of value system is better in identifying who had the best year (regardless of how susrainable it is)

        Vote -1 Vote +1

      • Nathan says:

        I think you’re wrong about “no sign of regression,” but that’s honestly not the important point here. The point here would be that, even if Craig’s RISP numbers don’t regress for 5 seasons, Craig is still inherently a less valuable player than the player that can produce a higher wRC+ in across-the-board situations, because Craig cannot control if he comes to the plate with RISP, and while his manager can kinda sorta put high OBP guys in front of him to maximize the chance that Craig hits in RISP situations, there’s no true control there.

        In other words, Craig could have 200 RISP PA in 2014 and post a similar wRC+ ~200, and then in 2015 could post that same wRC+ ~200, but only get 100 RISP PA. This means that even if Craig has some specific skill, talent, or mental makeup that enables him to hit exceedingly well with RISP, his true value can still fluctuate wildly since that production is dependent on context that anyone has little control over.

        Vote -1 Vote +1

    • tz says:

      I remember reading something about HR rates decreasing when runners are on base, presumably due to pitchers changing their approach. This came back to mind when I was reading Bradley Woodburn’s article on runs per HR.

      Perhaps Craig has found a way to exploit any such tendency. Or maybe the runner on second is signalling pitches back to Craig.

      Vote -1 Vote +1

      • ralph says:

        So here’s a funny thing — I had thought along those lines that maybe Craig was just happy to drive in runs without swinging for the fences, which would be consistent with such a K% drop.

        But he actually has better PA/HR ratios with RISP than with BE:

        BE: 7 HR in 283 PA (40.4 PA/HR)
        RISP: 4 HR in 146 PA (36.5 PA/HR)

        BE: 10 HR in 264 PA (26.4 PA/HR)
        RISP: 8 HR in 149 PA (18.6 PA/HR)

        The differences aren’t nearly as stark as in K%, so I’d hesitate to say he actually shows statistically significant greater HR power with RISP. But the numbers make it awfully hard to say his outstanding RISP numbers are in part attributable to not swinging for the fences with RISP.

        Unless of course he’s actually proving the old adage about how you hit more HR when you’re not trying to hit HR.

        Vote -1 Vote +1

    • Tim says:

      Interesting. If someone had found a systematic flaw in the way they were being pitched, that’s pretty much exactly what we would expect it to look like. (BABIP’s probably still inflated, but otherwise.)

      I don’t really have time or access to go through databases and video to try to identify whether/where it’s happening, but that could be a very productive thing for someone to do. And by someone I mean Jeff Sullivan.

      Initially I’d like to know whether the split comes apart with two outs, and also if there’s some strange way that Craig is pitched that would make sign-stealing extra effective. (For instance if nobody ever throws him off-speed pitches inside.) I suspect that by looking at base-out states and pitch values with RISP one could probably track down what’s going on.

      Vote -1 Vote +1

  11. JuanPierreDoesSteroids says:

    I’ve actually put RE24 into fWAR myself before just for fun! I’ve always thought it odd that nobody ever addressed the fact that it is essentially a context dependent wRAA. I would have thought that a WAR with RE24 instead of wRAA would have already existed and have been on the player pages.

    Also, I am a proponent of renaming it CBRAA. I’m tired of saying Are-EE-Twenty-Four. I’m also tired of saying Weighted-Runs-Created-Plus. You guys really need to make more marketable names for your stats!

    Vote -1 Vote +1

  12. Co-sign on calling RE24 COBRA. Will start doing so immediately. If nothing else, can we get “Cobra” into the “Getting to know RE24” post so I can link to it on other sites? The cooler the name, the more people will fall in love.

    Vote -1 Vote +1

  13. Damon Selman says:

    “His WAR is +5, but once you adjust for the Cobra, he’s off the charts!”

    Not foreign sounding to a GI Joe fan.

    +7 Vote -1 Vote +1

  14. Anon says:

    If you’re wondering why they keep winning even with a line-up of guys that don’t look to be having amazing seasons from first glance

    If you think the Cardinal lineup isn’t amazing, then amazing doesn’t exist in MLB. Look at team wRC+ for non-pitchers in 2013; the Cardinals and Tigers are tied in first at 116. The Cardinals lead the league in 2012 (115) and 2011 (119).

    Vote -1 Vote +1

    • Cool Lester Smooth says:

      They usually don’t have any one outstanding player, though.

      They just don’t really have any bad ones not named Pete Kozma.

      Vote -1 Vote +1

  15. jac says:

    I like RE24 or Cobra or whatever it is now, but what I really like is the difference calculation that you make. Can fangraphs add that as a stat? Cobra adjusted wins would be a nice thing to take a glance at at the dashboard level.

    Vote -1 Vote +1

  16. dennis says:

    I think there might be some skill component on a case-by-case basis. Here’s my reasoning:

    There was an article somewhere (maybe here?) about pitchers being able to change their pitching style to favor certain outcomes. Like for example, when the bases are loaded, pitchers tend to give up fewer HRs. If you take someone like Josh Hamilton, I suspect he might be at the bottom of the list here because the changes in pitching style with runners on are unfavorable to his hitting style. I think that if pitchers try to avoid meatballs with runners on, then it stands to reason that extreme free swingers with poor contact abilities would suffer in those situations.

    Vote -1 Vote +1

  17. Drew says:

    Isn’t it also possible that some guys hit a lot of solo homers, for example, because pitchers don’t give them much to hit with men on?

    Is it possible that Craig is not “respected” in this way?

    Feel free to shoot me down.

    Vote -1 Vote +1

  18. Jeremy B. says:

    Any specific reason you’re using RE24 as opposed to WPA? I would think the context-heavy crowd would find WPA more useful, given that it incorporates the score of the game as well as the particular base/out situation (unless I’m interpreting the two statistics incorrectly). Also, I’m betting people are more likely to warm to a metric called “Win Probability Added” than one that mirrors, as you put it, an oven’s error code.

    Vote -1 Vote +1

    • LeeTro says:
      FanGraphs Supporting Member

      In a 2-1 win with 2 solo homers, is a solo HR in the 9th really that much more valuable than the 1st inning HR? I want to give it a bit more credit, but not the ~30% difference WPA does. I’ve been trying to figure out a weighting system for the 3 stats, but context-neutral is probably the most important.

      Vote -1 Vote +1

    • Cool Lester Smooth says:

      It’s probably because WPA isn’t a very good stat, but who knows.

      Vote -1 Vote +1

  19. In some of these cases (like maybe Gerrardo Parra and his 10 CS) the disparity between batting runs and RE24 might be caused by SB/CS, which are included in RE24 but not in batting runs. Just something to keep an eye out for.

    Vote -1 Vote +1

  20. ksclacktc says:

    Looked at 1993-2013 and this is what I came up with as best seasons. A lot of Rockies . Am I missing something about Park effects? Why so many Rockies?

    Season Name Team RE24 Bat diff
    2000 Jeff Cirillo Rockies 33.5 -1.4 34.9
    2000 Todd Helton Rockies 93.2 58.4 34.8
    1996 Andres GalarragaRockies 51.9 20.4 31.5
    2002 Lance Berkman Astros 73.2 42.5 30.7
    2000 Tom Goodwin – – – 4.3 -24.8 29.1
    2000 Johnny Damon Royals 42.9 14.6 28.3
    1997 Tony Gwynn Padres 71.1 43.5 27.6
    1999 Matt Williams Diamond 37.7 11.2 26.5
    1996 Barry Bonds Giants 93.9 67.5 26.4
    2006 Clint Barmes Rockies -17.1 -43.4 26.3
    2000 Jeffrey HammondsRockies 33.0 7.0 26.0
    2001 Jeff Bagwell Astros 66.3 40.4 25.9
    2001 Eric Chavez Athlet 45.2 19.4 25.8
    2002 Jose Vidro Expos 48.1 23.0 25.1
    1996 Dante Bichette Rockies 31.7 6.7 25.0
    2006 Todd Helton Rockies 38.9 13.9 25.0

    Vote -1 Vote +1

  21. Doug Lampert says:

    Are the fangraphs RE24 tables for the 24 states for the current year or for last year available?

    Vote -1 Vote +1

  22. Ryan Howard says:

    Finally, you guys get my 2006

    Vote -1 Vote +1