Batter's Box Interactive Magazine

Match-ups

As longtime readers may have noticed, I am enamored with batter-pitcher match-ups. I take them seriously, I believe in them absolutely.

I think it's safe to say that most serious analysts of the game, a group which certainly does not include me, do not share this view. Which is interesting because everyone - I mean, absolutely everyone - who is directly connected with the game takes these match-ups quite seriously indeed.

You know, the managers, the coaches, the players...

I don't know what, if anything, that actually means - but I sure think it's interesting.

Anyway. As I understand it, the objection to taking batter-pitcher match-ups seriously always invokes, like a talisman, the famous phrase "small sample size." I want to vamp a little on the notion of the sample size...

Why do we care about the sample size? What are we trying to do with it, anyway? What questions are we trying to answer?

What is a small sample size, anyway? When does a sample size become meaningful? Are there not times when we actually prefer the smaller sample?

Of course there are. The ultimate data sample is the complete package, a player's career numbers. But there's only a window of time, in mid-career, when we regard those as the most meaningful source of information. If your name is Ken Griffey or Travis Snider, we're not going to pay too much attention to those career numbers. At the very least, we're not going to pay too much attention if the question we're trying to answer is how much this particular player is going help our team right now (or, alternatively, over the next few years.)

That's not the only instance, of course. Every time we pay attention to a platoon split, we are announcing our preference for the smaller sample.

Which I'm cool with. Sometimes the entire season - thousands of games, thousands and thousands of at bats and balls in play, more than a million pitches - can actually come down to one single pitch. The sample doesn't get any smaller, or more meaningful, than that. Ralph Terry and Bill Mazeroski. Alejandro Pena and Gene Larkin. Mariano Rivera and Luis Gonzalez.

Now I believe, as I said, in batter-pitcher match-ups. But I will also tell you that I do not believe how a player performs over, say, one month tells you much of anything. When Lyle Overbay hit .171 this past April, I didn't think it meant he had completely lost his ability to hit major league pitching, and neither did any of you. (Well, neither did most of you.) We knew it was just one month and it didn't mean much of anything. What was the sample size? It was 96 plate appearances. Which means diddly, right?

As it happens, 96 plate appearances is twice as many plate appearances as Lyle Overbay has against any pitcher he's ever faced in his major league career. But, yes, I am going to take these match-ups, and that sample size, seriously.

It is unusual these days for a player to have as many as 100 plate appearances against anybody. Omar Vizquel, who has stepped up to bat more often than any other player currently active in the majors, faced Roger Clemens more times than anyone else, but he still had just 90 plate appearances against the Rocket (That was probably quite enough for Omar, by the way - he hit .198 and slugged .203 in those ABs.)

You're probably wondering which match-up between active players has occurred most often! Come on, you need to know!

It's a tie folks, but not for long.

Garrett Anderson has faced Jamie Moyer 112 times, and done okay (.312/.321/.550 with 5 HRs in 109 at bats.) They last faced each other about a year ago - while they're both in the NL, Garrett is a bench player now, and Moyer's future is in jeopardy.

Jason Giambi has also faced Tim Wakefield 112 times, and generally been befuddled by the knuckler (.163/.304/.359) - Jason's best move has been to keep the bat on his shoulder and hope for a walk. Giambi's now on an NL bench, while Wakefield rolls on in the AL.

Which is why we shall have a new leader quite soon. The Red Sox have 10 games coming up against the Yankees before this season is over, including a four game set in about two weeks. Wakefield has pitched to Derek Jeter 111 times. The Captain has held his own (.295/.324/.419).

What's the difference between Overbay's 96 meaningless at bats in April, and Giambi's 112 against Wakefield?

I dunno. What's the difference between one month and ten years?

The batter-pitcher matchup, as I've said before, tells us who - the batter or the pitcher - has the initial advantage in this particular match-up. Roy Halladay is a certifiably great picther, and given enough opportunities, he would eventually find some way to pitch to Johnny Damon that actually works. Eventually. But he's had 108 actual opportunites to pitch to Damon, and he never did find the answer. Still, he spent ten years trying. He just wasn't able to do it. His best solution has been to go the National League.

Anyway, here`s some information about your 2010 Blue Jays hitters. One of the things that interested me was to see if I could find anything in common among the guys that give a hitter difficulty, or the guys he destroys. At this point, I'm convinced that it's not the pitcher's stuff that makes the difference. I suspect it might be his delivery. Some hitters just don't see the ball as well against some guys.

Who the hell knows.

Vernon Wells

Most PA against - Tim Wakefield (72). V-Dub has scratched out some singles, but that is all (.258/.347/.306).
Most hits - Mike Mussina (21). Amongst active pitchers, it would be Mark Buehrle (19).
Most strikeouts - Jeff Weaver (10). Go figure. Wells has actually hit Weaver very well (.297/.350/.595)
Most HRs - Josh Beckett (5)
Likes to face - Buehrle (.500/.535/.658) and Beckett (.343/.425/.886)
Hates to face - Mark Hendrickson (.205/.244/.359)
No clue whatsoever - Jorge Julio (0-12)

See, that's what I'm talking about! Hendrickson and Buehrle are both LH finesse pitchers. One of them Has Vernon's number, and the other one - the really good one - Wells is able to hit more or less at will. He had an awful time with a mediocre RH flame throwers like Daniel Cabrera, but he pounds the crap out of Josh Beckett, a good RH flame thrower. He was helpless against Esteban Loiaza and Victor Zambrano...

Lyle Overbay

Most PA against - Josh Beckett (41). Opie's done okay (.314/.415/.429)
Most hits - Beckett and James Shields (11)
Most strikeouts - Beckett (11)
Most HRs - Brandon Claussen (4), all when he was with the Brewers in 2005. He's got 3 against Shields.
Likes to face - Zack Greinke (.381/.409/.714). He wishes Claussen was still around, and Greg Maddux, too.
Hates to face - Andy Pettitte (.103/.103/.138) and A.J. Burnett (.087/.125/.130)
No clue whatsoever - Mark Buehrle (0-16)

John Buck

Most PA against - Mark Buehrle (55). It hasn't been pretty (.157/.204/.353)
Most hits - Cliff Lee (12)
Most strikeouts - Buehrle (16)
Most HRs - Buehrle and Lee (3)
Likes to face - Cliff Lee (.353/.436/.706), and he shows up with a smile when Andy Pettitte or Nate Robertson is on the schedule.
Hates to face - Jose Contreras (.118/.158/.118)
No clue whatsoever - Zach Miner (0-13)

Aaron Hill

Most PA against - Scott Kazmir and Tim Wakefield (35). He's gone 10-33 (.303) against both of them.
Most hits - Josh Beckett (13)
Most strikeouts - Jon Lester (9)
Most HRs - Jamie Moyer (3)
Likes to face - Beckett (.406/.441/.688). And Hill's struggles in 2010 may simply be because Nate Robertson is no longer in the AL.
Hates to face - Andy Pettitte (.133/.133/.133), naturally. Lester and Hendrickson gave him fits as well.
No clue whatsoever - Runelvys Hernandez (0-9)

Edwin Encarnacion

Most PA against - Carlos Zambrano (48). EE has gone .238/.333/.381 with a couple HRs.
Most hits - Paul Maholm (13)
Most strikeouts - Zambrano (11)
Most HRs - Anthony Reyes (4) in just 7 ABs!
Likes to face - Besides Reyes? Jason Marquis (.500/.556/.938)
Hates to face - Roy Oswalt (.125/.160/.208)
No clue whatsoever - Rich Harden and Brad Lidge (0-8)

Jose Bautista

Most PA against - Andy Pettitte (33). It's been fun (.310/.394/.621)
Most hits - Pettitte (9)
Most strikeouts - Doug Davis, Rich Hill, Russ Springer (7)
Most HRs - Pettitte, Burnett, Kevin Millwood, Javier Vazquez, Kevin Slowey (2)
Likes to face - Jeremy Guthrie (.333/.412/.800). Pettitte, Chris Capuano, and Dave Bush are also welcome anytime.
Hates to face - Carlos Zambrano (.067/.222/.133), Not fond of Hendrickson, Ben Sheets, or Wandy Rodruguez.
No clue whatsoever - Russ Springer (0-12)

Yunel Escobar

Most PA against - Ricky Nolasco (27), and it went well (.370/.370/.593).
Most hits - Nolasco (10)
Most strikeouts - Jamie Moyer, John Maine, Max Scherzer, Ted Lilly (5)
Most HRs - Tim Redding (2)
Likes to face - Hendrickson! Finally, someone who can hit this guy (.750/.786/1.083). And Redding.
Hates to face - Kyle Kendrick (.063/.118/.063)
No clue whatsoever - Roy Halladay (0-9)

Fred Lewis

Most PA against - Aaron Cook (24), he's hit a few singles (.273/.333/.273)
Most hits - Dan Haren (9)
Most strikeouts - Haren (8)
Most HRs - 22 different guys (1)
Likes to face - Jason Hammel (.778/.818/1.444). And John Lackey, Jeff Peavy, Roy Oswalt, Adam Wainwright...
Hates to face - Hmm... Chris Young? (.167/.333/.167)
No clue whatsoever - Max Scherzer (0-8)

Adam Lind

Most PA against - Matt Garza (31)
Most hits - James Shields (9)
Most strikeouts - Andy Pettitte, Jon Lester, Daisuke Matsuzaka (8)
Most HRs - Shields and Scott Feldman (3)
Likes to face - Joba Chamblerlain (.571/.625/.929), and he's pretty happy when Feldman or Gavin Floyd is on the mound.
Hates to face - Matsuzaka (.050/.095/.050), of course. Millwood and Hendrickson (again!) have been a problem.
No clue whatsoever - Lind went 0-19 against Dice-K before lining a single the last time they met. So now it's Buehrle (0-12).

Travis Snider

Most PA against - Gavin Floyd (14)
Most hits - Carl Pavano (5)
Most strikeouts - Floyd (6)
Most HRs - Kevin Slowey and Sergio Mitre (2)
Likes to face - Slowey (4-5, 2 HRs).
Hates to face - Shields? (1-9, 4 KS)
No clue whatsoever - Zakc Greinke and Jake Peavy (0-6)

John McDonald

Most PA against - Tim Wakefield (31)
Most hits - Andy Pettitte and David Wells (8)
Most strikeouts - Scott Kazmir (6)
Most HRs - Jerrod Washburn (2)
Likes to face - Washburn, I would think!
Hates to face - Johan Santana (.077/.143/.154)
No clue whatsoever - Julian Tavarez (0-11)

Jose Molina

Most PA against - Freddy Garcia and Jamie Moyer (21)
Most hits - Garcia (8)
Most strikeouts - Randy Johnson (8). Molina had 11 ABs against the Unit - the other three were double, home run, groundout
Most HRs - 22 different guys (1)
Likes to face - Hendrickson! (.545/.545/.818)
Hates to face - Josh Beckett (.125./.125/.125) and Roy Halladay (.071/.133/.143)
No clue whatsoever - C.C. Sabathia (0-11)

DeWayne Wise

Most PA against - Justin Verlander (16)
Most hits - Verlander (6)
Most strikeouts - Roger Clemens (4)
Most HRs - Carl Pavano (2)
Likes to face - Verlander and Jon Garland.
Hates to face - Pavano's retired him in every other AB.
No clue whatsoever - Joe Nathan (0-6)

And, in the Gone But Not Forgotten category:

Alex Gonzalez

Most PA against - Tom Glavine and Kevin Millwood (57)
Most hits - Millwood (16)
Most strikeouts - Al Leiter (13)
Most HRs - Michael Bacsik (3)
Likes to face - Roy Oswalt (.435/.458/.696)
Hates to face - Livan Hernandez (.098/.156/.122). He's also pretty happy Maddux retired.
No clue whatsoever - Eric Gagne (0-10, 6 Ks)

And a Happy Birthday shout-out to my old friend Sally, my old mentor George Bernard Shaw, and that old rogue Mick Jagger.

Posted by Magpie on Monday, July 26 2010 @ 04:23 AM EDT.

Match-ups | 32 comments | Create New Account

The following comments are owned by whomever posted them. This site is not responsible for what they say.

Alex Obal - Monday, July 26 2010 @ 05:58 AM EDT (#219151) #

The batter-pitcher matchup, as I've said before, tells us who - the batter or the pitcher - has the initial advantage in this particular match-up. Roy Halladay is a certifiably great picther, and given enough opportunities, he would eventually find some way to pitch to Johnny Damon that actually works.

Right.

My crackpot take on batter/pitcher matchups is that I believe in them, but with the cause and effect backwards. In other words it's the results of the matchups, rather than the tendencies and quirks of the batter and pitcher, that affect future encounters and give whoever won the first few matchups the mental edge. In this insane theory, Vernon Wells isn't 19/38 off Mark Buehrle because he owns him. Rather, Wells owns Buehrle because he's 19/38 off him and they both know it. It's hard to pitch, or hit, with knots in your stomach. And getting away from your usual m.o. to change things up against a guy who owns you only puts you at a disadvantage.

I've always thought that soft-tossing lefty cutter artists drink Overbay's milkshake. 0/7 off Dinardo, 3/29 off Pettitte...

scottt - Monday, July 26 2010 @ 06:16 AM EDT (#219152) #

Awesome post, Magpie.

Chuck - Monday, July 26 2010 @ 07:27 AM EDT (#219154) #

I think it's safe to say that most serious analysts of the game, a group which certainly does not include me, do not share this view. Which is interesting because everyone - I mean, absolutely everyone - who is directly connected with the game takes these match-ups quite seriously indeed.

I don't know what, if anything, that actually means - but I sure think it's interesting.

Not to be glib, but could it simply mean that almost no one in the game -- managers, coaches, players -- has sat in a classroom learning about probability theory? (I imagine that several of the younger, highly educated GMs have.)

John Allen Paulos wrote a wonderful little book on the innumeracy that runs rampant in our society. And it's not just those that make up the small world of professional athletics that are afflicted. Highly educated people in many professions have a very modest grasp of even the most fundamental mathematical principles. Hell, I would argue that many of my fellow graduates from a university specializing in mathematics emerged far less mathematically competent than one would hope.

Innumeracy is everywhere, so it's not surprising that someone like Cito Gaston might make a big deal out of 3-for-6 or 1-for-7.

AWeb - Monday, July 26 2010 @ 07:49 AM EDT (#219155) #

My favourite batter/pitcher matchup that always seemed to point to more than just randomness was Gwynn vs. Maddox - Gwynn completely owned Maddox, hitting .429/.485/.538 off him in 103 PA. Gwynn was in turn owned by Dwight Gooden and Nolan Ryan - hard throwing right handers to get him out you would think, except that John Smoltz couldn't stop him at all (1.224 OPS).

With the new focus on divisional play, some players should start racking up even larger matchup numbers. Older players managed a lot - Aaron put up a .925 OPs in 249 PA against Drysdale, Pete Rose managed a 747 OPs in 266 appearances against Phil Niekro. Not sure if that's the highest in the babseball reference record, which seems to start in the 50's (retrosheet era, I assume). I would think the very highest numbers would be for Walter Johnson (long AL only career) against probably Ty Cobb (almost exact contemporary), but I don't see a source for that. Maybe 300-350 PAs for those two?

Matchups with "similar" hitters would seem to be a way to go with this type of information, although similar would have to be defined not by the handedness and hitting line, but by swing mechanics and approach (as I see it anyway).

Magpie - Monday, July 26 2010 @ 07:50 AM EDT (#219156) #

Every manager looks at the match-ups these days, but as far as I know the first guy to really make something of it was Earl Weaver with his little index cards. By the 1980s, Davey Johnson was running the Mets and keeping similar information.

There is a Keith Hernandez anecdote I remember, from his book about the 1986 Mets. Evidently Johnson looked at Hernandez's numbers against Nolan Ryan and thought this would be a good day to give his first baseman a day off. Now most players were extremely happy to get the day off when Ryan was pitching, but not Hernandez. I know my numbers suck, he told his manager, but I've had good at bats against him. I'm comfortable hitting against him. Johnson let him play, and Hernandez hit a decisive home run in that very game...

We can look this stuff up now! Over his career, Hernandez did go just 8-43 lifetime against Ryan. Up through July 1984, he was 2-21 against the Express. And on July 3, he hit a two-run homer against Ryan to give the Mets a 4-3 lead (that would be the final score). Two weeks later, in a 1-1 game, Hernandez hit another two-run homer against Ryan and the Mets would win 3-1.

Youneverknow...

Chuck - Monday, July 26 2010 @ 08:17 AM EDT (#219157) #

I know my numbers suck, he told his manager, but I've had good at bats against him. I'm comfortable hitting against him. Johnson let him play, and Hernandez hit a decisive home run in that very game...

Not trying to be a buzzkill and rain on this Disney moment, but how many times did Hernandez (or someone else) have that very conversation with his manager and then not play well? The failures don't seem to get recorded in the ledger and certainly don't find their way into autobiographies.

I don't know that this is exactly a case of confirmation bias, but it's at least similar. The successes get remembered, reinforcing the underlying belief, and the failures get forgotten.

Mike Green - Monday, July 26 2010 @ 09:43 AM EDT (#219159) #

Magpie, here's a simple thought experiment. BBRef lists Batter vs. Pitcher matchups for the day's games. Pick a bad hitter (Cesar Izturis say) who is 9-20 against a particular decent pitcher. And watch what he does. Repeat and repeat. You can do the same thing for hot hitters. The key thing is a priori, not a posteriori (before it happens, not after). That is essentially what the analysts have done, with the aid of computers.

If you do, I guarantee you that you'll be satisfied afterwards that it is better to look at a batter's general abilities and any particular extreme platoon tendencies for a LHB, rather than batter vs. pitcher or hot vs. cold to figure out what the batter is most likely to do.

christaylor - Monday, July 26 2010 @ 10:45 AM EDT (#219164) #

If memory serves, in Tango et al.'s book this issue was addressed in a unique and interesting way -- pitchers were grouped into families of similar pitchers (I forget the criteria used to generate a family) and then a batter v. family numbers were computed.

I think a promising approach for batter-pitcher match-ups could be to measure how hitters fare against certain types of pitches and then try to predict success of a batter against a pitcher who throws those pitches (and how frequently). I think the conclusions of that chapter of the book were pretty weak, some trends were found, but nothing in line with the stock baseball people put in match-ups.

The reason why I most baseball people think said match-ups matter, is that most hitters have holes in their swing. People who watch baseball for a living can spot these holes and that fans/stat-heads can not. The success/failure of batters v. certain pitches come from those pitchers being able to exploit those holes. Nothing earth shattering -- but I also think said match-ups have face validity and just saying "small sample size" adds nothing to the conversation.

christaylor - Monday, July 26 2010 @ 10:55 AM EDT (#219165) #

It seems hard to say anything about Gaston's numeracy with confidence, but I don't think he cares about the exact numbers. When justifies his use of batter pitcher match-up he often says, "X hits Y well" or "X has struggled against Y". He seems to frame it more subjectively.

If pressed in an interview, I'd bet that he would say "he trusts his eyes" rather than the match-up break downs he probably has at his finger tips... in fact that he doesn't doesn't like to platoon sort of back this up, very few hitter RH have enough ABs against LHP to make their numbers statistically significant.

Chuck - Monday, July 26 2010 @ 11:14 AM EDT (#219166) #

Nothing earth shattering -- but I also think said match-ups have face validity and just saying "small sample size" adds nothing to the conversation.

I think you'd find that many people agree that match-ups do have validity (given enough ABs), and that success and failure against certain pitchers is not entirely random. I think the "small sample size" argument is meant to dissuade people from drawing conclusions from a 4-for-7 match-up, say. I don't think it's meant to categorically dismiss the subject matter as a whole.

A hitter with a 4-for-7 history against a pitcher may well have a skill that allows him to do better against that pitcher than against all other pitchers. But such a conclusion can't be drawn from a mere 7 at-bats. Random chance alone can (statistically) explain most of that 4-for-7. Now a 40-for-70 match-up is a whole other matter. Random chance can only (statistically) explain so much. The probability of the 40-for-70 being the byproduct an underlying cause is much, much greater.

christaylor - Monday, July 26 2010 @ 11:31 AM EDT (#219167) #

Perhaps I misunderstand -- but hope that this approach is not done. This approach essentially uses Bayesian inference -- the problem here is that relative probability of observing a true .400ish hitter versus a .250 (for example) is minuscule. This which means that the inference will tend toward the hypothesis that the hitter is a .250 hitter (or whatever his career average is) very quickly.

Bayesian inference is a nice tool, but I can't see how it gets around the problem of small sample size in this instance.

Chuck - Monday, July 26 2010 @ 12:00 PM EDT (#219169) #

Bayesian inference is a nice tool, but I can't see how it gets around the problem of small sample size in this instance.

That's just it. I don't think you can get around the problem of a small sample size. I acknowledge that the phrase "small sample size" is blurted seemingly mindlessly when discussing match-ups, but my contention is that this is almost always justified.

What has been discussed in this thread -- the idea of measuring performance against families of similar pitchers -- is one way to perhaps collect a number of small samples into meaningful large ones. The LHP/RHP breakdown of pitchers into two very large families seems to have merit, though of course it isn't a universal that all hitters and pitchers perform better in "favourable" platoon match-ups.

christaylor - Monday, July 26 2010 @ 12:04 PM EDT (#219170) #

There's no doubt that random chance explains most of a 4/7 and one should always have in the back of one's mind that the variation is random chance, but the vast, vast majority batter pitcher match-ups are closer to 7 than 70. Getting to the the point where the data from a batter pitcher matchup has enough confidence to matter, for example, binomial 95% confidence on a hitter who has gone 10/20 is .272 to .728... and that's 20 AB which I'd guess is infrequent.

Thinking about it more -- batting average is probably the worst stat for this... because the probability of a hitter hitting a 2B/3B/HR is much smaller. That a batter has a high SLG ought become a much more reliable estimator of the batter pitcher much more quickly, statistically... but only for successes.

The default from a stats point of view is playing match-ups should really only be a factor if there's over-whelming evidence (40/70 -- has anyone been a greater than .500 hitter over 70 AB in the baseball history, even) or that all is equal. The trouble about the "all else being equal" part is that it never is, for example when deciding to start say one utility infielder over another.

That they are so used so ubiquitously requires a psychological explanation -- inummeracy is a good one, if a little dour. That baseball people are able to see things most can't (they are all highly trained experts) is kinder, but probably gives the baseball people too much credit (Cito does do many strange things that don't work out).

Match-ups are probably one of those issues that will forever remain, to use James' phrase "in the fog".

christaylor - Monday, July 26 2010 @ 12:18 PM EDT (#219172) #

"I acknowledge that the phrase "small sample size" is blurted seemingly mindlessly when discussing match-ups, but my contention is that this is almost always justified."

I don't think we disagree at all... most of the time, match-up numbers just seem to give color guys something to talk about. However, they're appealing, probably because so much of baseball is a battle between two individuals within the context of a team game.

AWeb - Monday, July 26 2010 @ 01:05 PM EDT (#219174) #

40/70 -- has anyone been a greater than .500 hitter over 70 AB in the baseball history

WEll, my quick look at this including the active leaders in BA
Ichiro - .405 against Millwood in 79PA, .512 against Padilla in 47 Pa
Pujols - .429 agaisnt Ian Snell in 44 PA
Helton - .465 against Livan Hernandez in 85 PA

So some guys have come close, it's possible...althogh 70 PA hasn't been done that I see.
Ted Williams is at the edge of player/pitcher matchups data (data starts in 1950 at bbref), and managed .512 against Vic Raschi in 55 PA.

Mike Green - Monday, July 26 2010 @ 01:41 PM EDT (#219178) #

Take Vernon Wells. The lefties he has faced the most often are in order: Pettitte, Moyer, Buehrle, Rogers and Hendrickson. He has the worst numbers (by far) against Hendrickson; the best by far are against Buehrle. If someone wants to bet that he'll fare better against Hendrickson than against Buerhle over the next 5 PAs, I'll take that one. It is kind of convenient that Hendrickson is coming to town.

Alex Obal - Monday, July 26 2010 @ 02:44 PM EDT (#219189) #

This thread seems to focus on the question, "Does a given batter-pitcher matchup tell us anything significant about how they will perform against each other in the long run, beyond the players' cumulative stats?" And I agree with pretty much everyone in this thread that the answer to that is probably, not really - the effect is probably so tiny in 99.94% of cases that you'll never be able to demonstrate it with any confidence. Kind of like clutch hitting.

I think the more interesting empirical question is, "Does recent history in a given matchup affect the probability distribution for the next at-bat?"

uglyone - Monday, July 26 2010 @ 02:58 PM EDT (#219190) #

I think the more interesting empirical question is, "Does recent history in a given matchup affect the probability distribution for the next at-bat?"

ah, the salient question.

well said.

Mike Green - Monday, July 26 2010 @ 03:06 PM EDT (#219192) #

I doubt it, Alex. For the sake of argument, let's say Vernon Wells goes 0-4 against Mark Hendrickson on May 16, 2010 and June 15, 2010. Does this mean anything if he faces him in July, 2010? I doubt it. If among those 8 at-bats, he has struck out 6 times or something, then maybe. But you very rarely see that kind of thing. FWIW, Wells' K rate against pitchers is pretty damn flat. The two pitchers he has struck out a lot against (Jeff Weaver and El Duque), he has also taken deep a lot.

Why does a hitter go 0-4 on a particular day? Bad luck. Minor Injuries. A pitcher particularly on his game that day. Batter in a funk or messed up mechanically. So, if you look at a starting pitcher who Lyle Overbay faced twice in April and struggled with, does that tell you anything about how Overbay is likely to do against him in July? I am pretty sure subjectively that it tells you nothing.

Alex Obal - Monday, July 26 2010 @ 03:26 PM EDT (#219194) #

For the sake of argument, let's say Vernon Wells goes 0-4 against Mark Hendrickson on May 16, 2010 and June 15, 2010. Does this mean anything if he faces him in July, 2010?

Well... (forgive the semantics game, but) what does 'anything' mean? Taken literally, how could it not mean anything? If Wells knows it, it'll affect him in some way, probably minor.

It's a good question. Intuitively, I'd 'expect' Vernon to hit (say) .287/.346/.502 over the next one plate appearance instead of (say) .290/.350/.510. And if there were 6 strikeouts in the 8 at-bats, I'd bias my expectation down even more. And if Vernon homered first time up, that would change everything. But without data, who knows.

Mike Green - Monday, July 26 2010 @ 03:48 PM EDT (#219201) #

Sure, it could mean something. Maybe Hendrickson has found a new pitch, or maybe Wells has a wrist injury that is affecting him. Almost always, these things can be seen by looking at something larger than a few confrontations.

Personally, I think that much more is to be learned by looking at particular pitches (movement, location and velocity) and batter's strengths and weaknesses.

dan gordon - Monday, July 26 2010 @ 03:50 PM EDT (#219202) #

I would love to see a proper statistical study of these pitcher/batter matchup numbers. Given how much they are used by managers to make lineup decisions, I would think that some team would be interested in doing it. If you took a large enough number of players and checked to see how batters performed in future AB's against pitchers they have performed well (or poorly) against in the past, I would be surprised if it turned out that the past performance had much predictive value for future performance. Perhaps for matchups of a large number of AB's (40? 50? 60?) you would see some predictive value, but I think that the smaller ones are likely to be meaningless. I just cringe when I turn on a Blue Jays game, and I hear that somebody is playing today because he is 5 for 9 or some such thing against today's pitcher and a clearly superior player is sitting because he is 1 for 10 or something like that.

John Northey - Monday, July 26 2010 @ 04:36 PM EDT (#219206) #

The only matchups I felt made sense were the extreme situations. Like any sane left handed hitter vs a young Randy Johnson.

Alex Obal - Monday, July 26 2010 @ 05:21 PM EDT (#219209) #

Sure, it could mean something. Maybe Hendrickson has found a new pitch, or maybe Wells has a wrist injury that is affecting him. Almost always, these things can be seen by looking at something larger than a few confrontations.

I'm thinking more about execution than individual quirks. Things like the pitcher being more likely to nibble or be 'too fine' against certain hitters, which is often self-defeating, and feel bulletproof against others. Or, like, Manny Delcarmen trashing his changeup against Frank Thomas because he killed one the other day. If stuff like that happens with any regularity, then in a big enough sample, it'll show up.

smcs - Monday, July 26 2010 @ 05:29 PM EDT (#219210) #

I think the more interesting empirical question is, "Does recent history in a given matchup affect the probability distribution for the next at-bat?"

I don't know enough about probability and whatnot to really dig into this question perfectly, but given what I know about baseball, a few inferences can be made in general. I would think that, the more plate appearances accumulated by a batter against a pitcher, the better the batter will do. For instance, we have seen Kevin Gregg struggle greatly when he faces a team multiple times in a series. More generally, if we look at the splits of the American League this season (link is giving me problems, but you can find the splits above the abbreviated standings on the homepage of baseball-reference.com) we see that batters improve as a game wears on against a starter. I think that the drop from 2nd time thru the order to 3rd time thru the order (by all of .001 OPS point) can be explained by exceptional pitching instances.

As to specific batter-pitcher matchups, I would expect batters to do better against pitchers as time goes on. A more simple explanation can be the style and type of pitches that someone throws. For instance, Mark Buerhle's worst pitch (all pitch data is according to Fangraphs) is his curveball. Vernon Wells just happens to like curveballs. But he hates cutters. Mark Hendrickson's best pitch just happens to be his cutter. But Vernon Wells also happens to do quite well against Mariano Rivera.

So overall what can we assume? I think we can make assumptions much along the line of platoon splits. Over time, the stats have shown that Vernon Wells is good at hitting curveballs. Therefore, we can expect that he will do better against pitchers who either throw lots of curveballs, or who throw very poor curveballs.

The good news? Brad Bergesen throws very poor curveballs. And very poor fastballs. And very poor changeups. And he doesn't even throw a cutter! We also know Wells' affinity for first-pitches. Good news again: Bergesen has an OPS against of 1.491 on first pitches.

To bring it back to the question at hand, I think there is more at work than random chance when it comes to the outcome of batter-pitcher matchups. However, there are too many variables to really significantly affect the probability of the next at-bat, but I think there is enough to show that there is an affect on the probability distribution of the next 10 or 20 at-bats.

Mike Green - Tuesday, July 27 2010 @ 10:25 AM EDT (#219230) #

After giving it some more thought, I realize what my main objection to the whole match-up thing is. We look at batting averages and home runs, or slash lines, to measure how well or poorly a particular batter has fared against a particular pitcher. Over 30-50 PAs, it is a very poor measure because of the variability of BABIP and the great uncertainty attached to whether fly balls become HRs or outs.

This is a place for xFIP. If you look at the particular pitcher's xFIP with platoon adjustment as required and reverse a hitter's line to get his xFIP (with adjustment to reflect the hitter's actual HR/fly rate rather than league averages) and then convert the batter and pitcher lines into an xxFIP (had to get the xx in there somewhere). You then run a Monte Carlo simulation, and see where the actual xFIP lands. It's quite a bit of work, and I am pretty sure that you won't learn much.

uglyone - Tuesday, July 27 2010 @ 02:33 PM EDT (#219250) #

Sometimes I could just do a comprehensive analsyis of average ball-off-bat speed & angle (combined of course with contact %), for both pitchers and hitters. I bet you it would be a helluva lot more accurate and predictive than most any other statistical model.

Mike Green - Wednesday, July 28 2010 @ 09:40 AM EDT (#219325) #

OK, Magpie. Here's a match-up contest for you, and the odds should be stacked in your favour. Wells and Bautista have raked against Jeremy Guthrie; Lind and Hill have struggled. I'll take Lind and Hill for tonight's game; you can have Wells and Bautista. OPS will be the measuring stick. The prize will be a cuttlefish eye. Are you in?

Mike Green - Wednesday, July 28 2010 @ 09:54 PM EDT (#219384) #

AWeb - Thursday, July 29 2010 @ 03:05 PM EDT (#219405) #

LInd's good day made me think about a study to test the importance of matchup history vs. platoon significance. If you just studied average hitters with a poor matchup record against a particular pitcher after 25 PAs despite the platoon advantage, I wonder how they do on the 26th on average.

Well, no one with more than 10 PA againsty Masterson, but Westbrook has been around a bit more. Wells with the most PAs, hitting .381 (OPS 867). Hill is 4/9 with a walk. Not many familiar faces coming up soon - AJ Burnett a week or more from now has owned Overbay (2/23, 1 walk, OPS .255) and Adam Lind (2/13, OPS .466). These are a couple of good test cases - platoon splits suggest they should do much better.

It would be an interesting database query to run, if I had access to such a thing (opposite handed matchup, 25 ABs with OPS below some pre-decided level, maybe .600?). Next PA - what is the OPS?

Mike Green - Thursday, July 29 2010 @ 03:52 PM EDT (#219406) #

I'd sure like to take my chances on Overbay vs. Burnett. Here's the BBRef report on their confrontations. Two deep line drives, one for a double and one for a single (off the wall?), three deep fly balls (all outs), eight ground balls (all outs), seven strikeouts, and one walk. It all adds up to a slash line of .087/.125/.130, but I'd be perfectly happy to give Overbay 25 consecutive at-bats (and send in a pinch-runner when he reaches) against Burnett to see how he would do.

Mike Green - Thursday, July 29 2010 @ 03:58 PM EDT (#219407) #

For fun, I checked Lind vs. Burnett. It was even more extreme, in some ways. 16 PAs, 3 walks, 4 strikeouts, four deep flyballs (all caught), a line drive to the outfield caught and four ground balls, two of which found a hole.

Match-ups | 32 comments | Create New Account

The following comments are owned by whomever posted them. This site is not responsible for what they say.