There has been a lot of snark, in the wake of the Yankees' loss in the ALCS, about Joe Girardi and his little binder of pitcher-hitter matchup data (the use of which, as
John (New York, NY): Rob, the sample size of batter/pitcher matchups is of particular interest to me. Obviously a sample size of 5-10 PAs against a single pitcher does not yield any useful data. However, when you consider that in those 5-10 PAs, a single batter is only facing the repertoire of a single pitcher, my question is how many PAs are required before the data becomes significant? 20? 50? More? What do you think?Later, another New Yorker had a follow up:
Rob Neyer: More than 20. I'm not sure if 50's enough. I'm not sure if any batter has ever faced a pitcher enough times to show us anything truly meaningful. I think what makes more sense is looking at how a hitter has fared against *types* of pitchers.
Jeff (NY): While I agree that 0-20 from one batter against one pitcher is too small of a sample size, aren't the odds higher that the batter will continue to struggle than if he were 0-2? I don't think varying degrees of samples, with regards to educated predictions, is talked about enough. For example, 0-5 is meaningless, 0-10 is less meaningless, etc. Does this make any sense? Is there any research I can read on it?This piqued The Common Man’s curiousity, and he did some digging through Baseball Reference’s Play Index function to see just how much exposure a batter could get to a pitcher. First, though, a couple notes.
Rob Neyer: I don't know of any research, but yes I would assume you're correct: 0-20 is more useful than 0-2. The question is whether or not 0-20 is useful enough to use.
- The Play Index only has play-by-play data as far back as 1950, so all figures are from 1950-2010. This means that a few of the pitchers TCM looked at here (such as Warren Spahn, Robin Roberts, and Billy Pierce) have incomplete data.
- Second, TCM limited his examination to the Top 60 pitchers (measured by Innings Pitched) of the last 60 years. This is convenient for a number of reasons. For one thing, it limits the pool to a manageable size. For another, there are exactly 60 pitchers who have, since 1950, thrown more than 3000 innings in the Major Leagues. And finally, the opportunity for pitchers with fewer than 3000 innings to face a batter enough times to be significant is incredibly low.
- And third, TCM listed all batters who had more than 150 plate appearances against these pitchers, as well as a few select matchups that occurred 100-150 times that were particularly interesting.
Some of the results are not terribly surprising (you can look at TCM's spreadsheet here, if you want to look at the data yourself). First and foremost, 23 pitcher-batter matchups occurred more than 200 times each. Of these, only one of the pitchers and one of the hitters involved debuted after the 1963, Phil Niekro (1964) and Johnny Bench (1967). The reasons for this should be intuitive. Baseball has expanded greatly since 1961, meaning that batters and pitchers have had fewer opportunities to square off against one another. Also, since the onset of free agency in 1976, players have been far more mobile during the prime years of their careers, allowing them to switch leagues while they are still playing regularly. This, too, decreases their chances of facing any single pitcher. Finally, the use of the bullpen has evolved tremendously since the heyday of complete game madness in the 1960s and early 1970s. When starters leave the game early, naturally batters get fewer cracks at them.
The pitchers, though, had their revenge too. Don Drysdale absolutely owned Eddie Matthews, who hit just .194/.275/.335 with 7 homers over 229 appearances. Likewise, Willie Davis was completely lost against Juan Marichal in exactly 200 PAs, hitting just .181/.191/.321.
Maddux and Craig Biggio also locked horns more than 150 times (152), but Bidge had far less success, hitting .284/.331/.369. Biggio also struggled mightily against Tom Glavine (.234/.323/.351 in 127) and John Smoltz (.239/.297/.316 in 128) in significant exposure.
One could, TCM supposes, make the argument that batter/pitcher matchups are less prone to streaky runs by batters or pitchers, because the matchups are more spread out over time. Indeed, the kinds of fluky stretches players experience are even less likely as pitchers begin to know more about the hitters. After all, one of the reasons Boesch seemed to fall apart for the Tigers was that pitchers learned how to pitch against him. If a batter or a pitcher has sustained success against another over 100 or 150 or 200 plate appearances, one could assume that that dominance is informed by all the exposure that has come previously.
And the players involved in these matchups also have a sustained baseline of Major League performance we can reference to inform us about their abilities. We know, for instance, that Willie Mays was a good player, regardless of his hiting just .196/.315/.304 against Bob Gibson, which makes Gibson’s performance more impressive. And because of this, TCM thinks it’s likely that the results are more trustworthy in these cases when they are more extreme. Joe Morgan’s .226/.335/.415 against Don Sutton in 191 PAs isn’t particularly instructive, as it appears his plate discipline and his power are relatively intact. The main difference seems to be a few extra balls in play that were converted into outs. Meanwhile, Ron Fairly’s .302/.369/.440 performance against Bob Gibson in 179 PAs seems to be more likely the function of a few extra singles and doubles falling in.
However, when we look at, for instance, Rickey Henderson’s performance in 117 PAs against Frank Tanana, .350/.427/.767 with 11 homers or Gary Carter’s greatness against Steve Carlton (.310/.400/.672 with 11 homers in 135 trips) or Don Drysdale’s success against Orlando Cepeda (.224/.257/.301 with 1 HR in 167 PAs), we’re looking at a player who, for whatever reason, is simply capable of dominating the other. So where’s the line? TCM is tempted to put the number around 150, though obviously 200 is even more instructive. The trouble, as you can see above, is that players are increasingly unlikely to hit that mark in today’s game, making Hitter vs. Pitcher data little more than a fun exercise.
6 comments:
Very interesting post. One follow-up for perhaps a select number of these guys would be to look at their numbers at different intervals (after 20 PAs/50 PAs/etc) to look at whether there were significant shifts in the individual matchups.
The one problem with career-long samples is that the true talent levels of the pitcher and hitter are not static over the course of time it takes to build up these samples.
Great point, Kevin. In fact, TCM thinks there's some evidence of this in the data, if you look at Brooks Robinson. Brooks was normally a fine hitter, but put up really paltry numbers against some of the better right-handed pitchers of the 1970s, Catfish Hunter, Jim Perry, and Luis Tiant in particular.
FYI, it was Aaron over at HBT that wrote the post about hitter/pitcher data not being sabermetric, not Mr. Shyster. I doubt either would care about the credit, but for the sake of correctness I figured I'd pass it along.
Kind of unrelated but I remember reading a Bill James piece about how Craig Biggio was awful against the best pitchers. Bill still loved Biggio, how he did every little thing to get the most out of himself like HBP. But, but his stats against the Braves that you mentioned reminded me of that.
I like the Blyleven reference. Maybe this type of evidence needs to be considered for HOF candidates. Similar to "Iron Chef" standards, you have to beat one to be one. Frame a player's credentials against current HOFers as a reference. Not the end-all, be-all, but still interesting. Thanks.
Post a Comment