Archive for the ‘pitch f/x’ Category
Boy, I had no idea there was such hate out there for BABIP. I’ve gotten some emails that even made it sound like I actually discovered something new here, when I was really just trying to tell the mainstream to go re-read their SABR manual. I do have something else to add that has been rolling around in my head for ages, and it’s about BABIP for pitchers. Warning: it’s pretty spacey. And very nerdy. I’ll have some wacky photoshopped pictures of Bautista’s head on someone else’s body next week, I promise.
There was an outcry when it was discovered that BABIP was essentially the same for all pitchers. McCracken’s statement that: “there is little if any difference among major-league pitchers in their ability to prevent hits on balls hit in the field of play” simply made no sense to baseball minds.
Although that statement has been softened over the years (we now know that the ability is just very, very, small, and almost impossible to see through all the luck), it’s still deeply troubling. This is because it implies that the ball comes off the bat on a dirty slider on the outside corner is just as good as on a slow fastball down the middle. Pitchers can only hit, or miss, bats – otherwise it’s all the same. Pitching to contact does not exist.
That makes no intuitive sense, and even less if you’ve ever tried to hit a 90-mph fastball. If some major league pitcher was facing a team full of me’s, it’s obvious he would have a really low BABIP as I dribbled the occasional ball back to the mound and popped up a few (in addition to putting Pizza Pizza out of business). Ok, I’m an absurdly low-talent example, but the variety in quality of swings and pitches to hit in baseball is pretty wide, too. All contact can’t be the same.
And it isn’t. Different counts show different BABIP’s. Different locations, too. Breaking pitches each have different BABIP‘s. So why wouldn’t a pitcher who threw more breaking balls have a consistently and significantly lower BABIP? Or one with significantly better control? And how do all these different numbers in different situations possibly even out to be the same for everyone?
Important realization #1
At least to some extent, the reason is that these factors are in balance. Throw tons of sliders, or pitch to the corners, and on those pitches you’re going to get weaker contact (i.e. a lower BABIP). But you’re also going to fall behind in the count more often and have to come in with a fastball down the middle, which will be hit harder (and thus for a higher BABIP).
Conversely, if you’re pounding the zone with fastballs, you’re going to give up more hits because that’s fastballs do, but you’re also going to be getting ahead more often, since they’re easier to control. (And the closer to down the middle you aim them, the more hits you give up, but the more often you get ahead). The advantage of being ahead in the count will balance to some degree disadvantage of throwing pitches that are more hittable.
A good example is a guy like Jesse Carlson, who broke into the league throwing more sliders than anyone else in the league (Fangraphs says 56%, pitch f/x says higher), and maintained a .230 BABIP over his first season. His second time around the league made the adjustment and stopped swinging at so many his sliders, meaning he was constantly working from behind against guys sitting fastball. He went through a really tough time, and then adjusted. He is a far more normal pitcher now, with a normal BABIP.
Maybe why everyone’s BABIP is similar that it is the “golden mean” for how much you should nibble, how much of the plate you should on average be catching. If your BABIP goes below league average, that means your stuff is nasty enough that you should be challenging hitters more, and getting more strikeouts at the expense of a little BABIP. If it goes way above, you’re finding too much of the plate and should work the corners or throw a ball once in a while, even if it means your walk rate goes up a little.
My big shiny idea is that (underneath the mighty wallop of luck that runs through it) the reason pitchers see a very consistent BABIP is not that “all balls in play are the same”, but because the optimal level of aggression in pitching for most pitchers — neither getting ahead too often or too little — results in a BABIP of that value.
1) Imagine a pitcher whose BABIP happens to be exactly .300. Lets say we teach him a slider he can only throw once in an at-bat — but it’s the best slider in existence! Guys just can’t touch it, and when they do, they only manage a .100 BABIP off it. But our imaginary pitcher is an idiot, so he throws it only on the first pitch of an at-bat. As a result, his BABIP goes down, and statisticians everywhere say that he’s been getting really lucky.
But then his catcher and pitching coach have a long talk with him, and convince him that he’s got to start using this thing for better use – punching guys out. It’s going to be a lot better if he can get guaranteed outs rather than guaranteed first strikes. So he starts throwing it on 0-2, and punches out a whole bunch of guys who were behind in the count and more likely to put the ball in play weakly — so his BABIP starts going back up, as his nastiness is correctly being converted into a much better result: maximum strikeouts. See! We say, he stopped getting lucky!
2) Now take another guy who of a sudden becomes incredibly hittable. He is overnight Joshified (Towerized?), and as you would expect, his BABIP goes way up as hitters start gearing up and crushing the ball off him, over the wall, etc. Statheads scream “how unlucky!” (As we did in 2004, when Josh Towers put together one of the worst seasons in baseball, but had a BABIP of .341). After a barrage of towering upper-deck shots, he will almost certainly become more fine: picking out the corners and throwing more junk, effective or not, will lower his BABIP (at the expense of some walks). And since he can’t quite finish guys off as well he used to with his weakened repertoire, what used to be K’s on 0-2 now result in the kind of weak contact you get on 0-2 lowering his BABIP further (although not necessarily making him any better). “It’s starting to even out!”, we cry…
I don’t know exactly how much this is happening, but it’s almost certainly part of the answer. Otherwise, pitch f/x analysis tells us electric stuff would mean lower BABIP, just as you would expect.
Once again, the real excitement is hit f/x, which will make this look like child’s play. But it’s possible that we could use BABIP to understand when a pitcher is catching too much of the plate or working around hitters too often. Although obviously only to some extent, because there’s still so much luck. But think about a 2008 Matsuzaka. He had a really low BABIP, and I think a lot of baseball experts, or anyone who watched his games, felt like he was underperforming in some ways for his stuff, because he pitched in such an odd, backwards, not MLB optimal BABIP style (and a .258 BABIP).
Or Roy Halladay. There were a few seasons when he was almost certainly “pitching to contact” because because of concerns about his arm. In 2005 and 2006, he had a BABIP of .262, and .276. There’s no way that was a coincidence, or an unmaintainable result. He struck out about one less a game, and just by watching him pitch you know why his K rate was so low and his GB rate so high – he’d get guys to 0-2 and throw a third sinking fastball, stuff like that. The use of his power curve dropped dramatically. And the truth is, he wasn’t as effective – his ERA+ dipped to 143 and then 121. Sometimes those 0-2 tappers would go through.
Now the doc is back to converting those weak 0-2 grounders into K’s again, and his overall numbers have increased dramatically (up to a 165 ERA+) over the last three years, while his BABIP has stopped being abnormally low. I would suggest that he was negatively affecting his results by pitching too aggressively, and has now returned to optimal.
A possible alternative (or at least a contributing factor) to the reason that pitchers don’t have widely different BABIP’s being that they have extremely low “ability” to affect what happens to balls in play is that deviating from the norm, or the perfect balance between contact, walks, and K’s eventually interferes negatively with your results, and is therefore weeded out, or corrected by major league pitchers. It may be possible to peer through the randomness involved in BABIP in order to use it to make general statistical statements about pitchers being too straightforward, or too crafty for their own good.
A quick request from Mr. Roman who, after watching Travis Hafner way behind pitches in the ALCS, (striking out an ALCS record 12 times, capped off by going down on three straight heaters to Papelbon in the 7th game with two on) wondered if Boston was feeding him high fastballs, possibly taking advantage of him trying to go the other way off the Green Monster. Here are the fastballs that Hafner saw, and what he did with them (this includes cutters and two seamers- anything above 90 that didn’t break more than half a foot).
As you can see, Boston made a concerted effort to pitch him up and away. He took a lot of pitches right on the outside edge, fouled off a few right down the middle, and couldn’t catch up to anything at the letters. Kind of surprising for a batter who mashed those pitches all year. I wonder if the Red Sox, who have notoriously good advance scouts (check out Schilling gushing about them on his Blog) picked up something, or if his bat speed was just down at the end of a long season so they kept pounding him.
It’s kind of hard to see graphically, but between the first three games in the series (when Hafner went 3-11 with two BB and a HR) and the last four (when he went 1-19 with 11 K’s) the Sox changed from going low and away in the strike zone to challenging him with high heat. Hafner didn’t swing at the low and away pitches (that’s all those called strikes), but he went for the high stuff and couldn’t do anything with it. Here are his fastballs split between early and late in the series:
He may have gone a little too aggressively after the corners early in the count, but Fausto Carmona was also getting squeezed all night by the home plate umpire Dana Demuth. Here were the calls made by the umpire tonight (from the batters perspective). He wasn’t getting anything on the corner that his sinker dives towards. The zone was tiny all night, but Schilling didn’t have it near as rough.
Also marked is where the pitches in the first inning were- I think you can guess which pitch Victor Martinez and Wedge were getting in the umpire’s face about after the inning was over.
The umpire rankings I’m working on show Demuth as having the third-smallest zone in the majors, but it’s mostly because he’s really small on the other side of the plate and doesn’t call the low strike (which explains a few of the calls on both sides).
I have to admit, when I saw Curt Schilling’s first few fastballs float in at 88 mph, I predicted a quick exit for the veteran hurler and an end to the Red Sox in the 2007 playoffs. But the old man got the job done, allowing only 6 hits over 7 innings, striking out five and walking none. He never topped 93 on the radar gun (and only hit it once against Hafner), but kept the Indians off balance with his splitter and threw a lot of first-pitch strikes. Here’s his location over the night.
Amazingly, of those 14 called strikes, 9 of them came on the first pitch to the batter. He started almost every at bat with a fastball. Here’s the movement on his first pitches- there are 2 curveballs and one splitter, and the rest of them relatively straight fastballs:
Otherwise, his control was good but not great. He left a lot of pitches up and over the plate that the Indians were unable to take advantage of. Why? Take a look at his pitch movement over the night:
After the first pitch, he threw a ton of splitters and they were diving a over half a foot. His fastball was also moving up to eight inches to either side, and quite often he took up to 10 mph off of it (although I’m not sure about those pitches that pitch f/x says were under 70 mph- one of them was the 4th pitch of the game to Grady Sizemore and there’s no way that was a 59 mph fastball. Two of the release points in that at-bat were about a foot to the right so maybe they were still setting up the system).
He also tossed in a handful of curveballs mostly as a show-me pitch. I don’t think he throws a cut fastball, but some of his fast splitters moved almost like a normal pitcher’s cutter. The two parallel lines made by his fastball and splitter suggest his split-fingered pitch cuts in and down and he can modify it for speed vs. break in the same way as his fastball.
All in all it was a pretty predictable game plan for a reformed power pitcher. Schilling threw first pitch fastballs for strikes, and then threw offspeed pitches to induce weak contact and expand the zone for strikeouts later in the count. The Indians took a lot of early strikes and had trouble with his splitter and a variety of different fastballs. There was nothing special about either his control or movement tonight, but the crafty vet had the Indians on the wrong foot and guessing wrong all night. Expect Westbrook to attempt the same thing tomorrow night.
Along with all the whirring, buzzing, booming effects and that incredibly annoying bar halfway down the top of the screen, Fox is using a high-tech feature called the “pitchtracker” occasionally during broadcasts that shows dots for each pitch around a box signifying the strike zone. It looks like the same one that has been used for Blue Jays broadcasts all year, and it’s only now that I’ve been objective enough to notice that it is ridiculously inaccurate. Almost every outside pitch looks like a horrible call by the umpires, and the broadcasters are constantly making excuses for them. Sometimes a pitch is called a strike and it looks like it’s half a foot off the outer edge. It’s fueling the fire for complaints about the umpiring of games, but really it’s the system to blame. Whoever set it up didn’t do some basic research into displaying an accurate strike zone.
The first problem is it’s shape. Sorry, the strike zone does not look like Ted Williams’ classic one any more. No umpire in the world calls a strike as high as dictated by the rulebook, and they give a couple inches off the plate. If you want to see the real strike zone, check out John Walsh’s measurements for a general idea. I’m actually starting to think that Walsh might have calculated a very slightly-too-large zone, but regardless it’s not the high, narrow box that is shown on the broadcasts (or even the slightly more squashed ones that are seen on the gameday program). It’s not that umpires are making “mistakes” to the side by a couple of inches, but that it is universally accepted that the real zone in slightly smaller vertically than the rulebook and to compensate is slightly wider.
To test the graphic on TV against the pitch f/x data, I rewound a few at-bats using TIVO and compared it to where the pitches were actually going. Here’s an at-bat that they featured between John Fogg and Eric Byrnes in the top of the 6th of the final game of the NLCS:
The yellow line is the average strike zone as determined by statisticians and called by umpires. It’s about 2 inches to each side of the plate and slightly above the knees and below the letters. The blue box is the rulebook strike zone. The red line is my estimate of where FOX had it’s strike zone set based on the broadcast. Vertically it’s consistent, but the width of the strike zone they use is insane. The announcers even called the third pitch “paint” because it was outside. According to the width of the real strike zone, all of those pitches were over the plate, but that particular pitch was a strike because it was the only one that caught the bottom edge of the strike zone.
Notice that their zone is about three inches narrower than the plate on either side (which is 10 inches wide), and that’s not even considering the standard inch or two off the edge of it that umpires give. Maybe they’re not including the width of the ball? But even that would only account for about half of the discrepancy. The zone is so narrow the ball is very rarely thrown inside it. For example, the Torrealba home run from the Rockies series was a mistake that ended up almost right down the middle, yet the pitchtracker shows it on the inside corner.
Here’s another example that includes the vertical height of the zone as well:
This is Troy Tulowitzki’s strikeout just before Torrealba’s crucial home run, which FOX showed in its entirety on pitchtracker. Matching their strike zone to the real one shows much the same results as the previous example (and clarifies the top edge). The first pitch was shown as just off the top corner, and so gives an idea of how high the strike zone they’re using is (I’m guessing it’s just below the letters). The third pitch (that Tulo fouled off) was shown as being right on the edge of the strike zone, which confirms that the outer edge they’re using is substantially inside even the edge of the plate. The final strike to retire Tulo was clearly a strike, but shown on FOX as being a terrible call by the umpire.
It would be one thing to blindly call the rulebook strike zone even though that is understood by players and umpires not to be the real one. But the current TV broadcasts use a strike zone that is even narrower than that, and makes any calls towards the outside of the plate look terribly wrong. From what I’ve seen, despite some eccentricities the umpires are actually consistently accurate to within a couple of inches, and the system that the major TV broadcasts are using right now is set so poorly it is misleading, and making both the umpires and the pitch tracking systems look bad.
It’s sad, really. As I’ve tried to show on this blog over the last few weeks, there are some pretty amazing and intuitive things you can show about what a pitcher is doing during a game or how the game is being called using this data. But instead of some real analysis we get another box cramped into the broadcast window that is useless and wrong.
So far using pitch f/x it’s been pretty easy to break down what is important for Dustin McGowan, Jesse Litsch, and A.J. Burnett to be effective. Compare a really good start to a really bad one and things pop out like how much a fastball is tailing, a curveball is curving, or a changeup is working.
The next thing I want to look at is what happened to Roy Halladay in May (other than the obvious- his appendix exploded). The Doctor was the best pitcher in the league in April, and from June 10th on had an excellent ERA of 3.32. But he pitched like a little-leaguer for three starts around the time that he went on the DL with acute appendicitis.
In the two starts before he had surgery, Halladay allowed 16 runs. He came back quickly from the DL and had one good start, but in his next one was absolutely shelled by Tampa Bay for 8 runs in 3 1/3 innings, the only time this season he did not go 5 innings. Roy said afterwards that it was a mechanical, not a physical problem, and he wasn’t getting on top of his pitches. Later it came out as Sal Fasano suggested a different grip (which Halladay later abandoned) that he was having trouble with his cutter. Let’s take a look at what was different in those three starts to make him so hittable.
First, here’s the pitch movement in a good start by the doctor, from April 13. He went 10 innings to beat the Indians, allowing only 6 hits and one run – an absolute classic from the Jays ace on top of his game. Remember, this is pitch movement as compared to some mythical perfectly straight pitch. Those red dots on the left are diving, moving fastballs. Compare it to the movement of another pitcher to get an idea of how much- the scale is in inches.
This is what the “new” Halladay looks like, ever since he remade himself from a power pitcher into a ground ball pitcher. He doesn’t throw a big 12-6 breaking ball, just a little breaking pitch that is more like a slider. But his fastball sinks and tails away to a ridiculous degree- over 6 inches even when compared to McGowan and Burnett’s 2-seamers. Can you imagine trying to hit a 93 mph pitch that sinks and tails that much? And now here’s his location. Everything is low and he cuts his fastball in on the hands of left-handed batters.
Ok, enough drooling at Cy Young stuff. Here’s Roy’s first rocky start in May where he allowed 12 hits and 9 runs against the Texas Rangers.
Doesn’t look that bad, does it? All his pitches are sinking even more. There isn’t the same sort of crisp definition between his two fastballs, but his velocity and breaking ball is fine. And as you’ll see in a second, for the most part his location was there as well.
So what was different? Well, this game actually came down to one inning, the third. After a strong start, the Rangers strung together 8 hits and 6 runs. Then the Doc recovered and the only other runs credited to him were when Josh Towers allowed a couple of inherited runners to score. In the next graph you can see the pitches from the third inning with yellow dots inside them, and the pitches that went for hits are marked with X’s:
(You may notice that the scale is a little different- for the second time I used pitch “break” instead of movement. It’s very similar except it takes into account the “loop” of a pitch, which makes it a little easier to define breaking pitches and changeups).
As you can see, almost all of those pitches that blurred the line between the Doc’s two fastballs came in the third inning, and most of the hits that inning came off those pitches. So it looks like the Doc’s cut fastball deserted him for half an inning. Here’s his location with the 3rd inning pitches and hits marked. All the pitches from that inning were missing just off the inside corner, or were cutters that didn’t cut how he wanted them, ended up right over the plate, and were smacked for hits.
If you can stand looking at another graph about this start, you can see here that the Doc was right about what went wrong- his release point slipped down for that one inning.
Right before his Appendix Exploded, Halladay had a rough outing against the Red Sox. 5 innings, 11 hits- and again he had problems in the third inning, allowing 6 runs off 7 hits. Again, the problem was with his cut fastball, except this time in that inning it was cutting more than usual. I assume that he couldn’t control it, because all the hits came off his sinking fastball that Boston was sitting on. Again, the yellow dots are his trouble inning and the X’s are Boston hits. The final nail in the coffin was off a slider low but right down the middle (the blue + on both graphs).
And now his location, which wasn’t so bad, really: He was just facing a good hitting team and not hitting his spots dead on like he usually does. Amazing what a couple of inches can do to a pitcher’s plan.
The last blemish before being strong for the rest of the season came on June 5. After 7 shutout innings in his first start back, he was lit up by Tampa bay. There’s no secret what happened this start- too many pitches left up. On all of the Doc’s other starts, there are next to no balls in the upper half of the plate. Here is his location for that start and where Tampa got their hits (check out the Greg Norton single on a ball that was apparently a foot off the plate away from him).
Release point was the culprit for those pitches. It’s a little hard to see, but I’ve tried to show what happened to Hallday’s release point throughout the game in the next diagram. In the first two innings it was very consistent and centered. In the third and fourth, it started varying wildly- up to 3 inches to each side, down and up. And the only times that Halladay was able to go back to his old release point? They were those high pitches that were hit hard by Tampa. I know Doc’s too much of a warrior to admit anything was wrong physically, but I wouldn’t be surprised if this was the result of him rushing back and just running out of strength after a few innings so he ends up guiding the ball.
So there you have it- Halladay’s bumpy patch was caused by two different issues. First he had some problems with his cut fastball coming and going (while going up against two of the best offenses in the league- not a good combination) that lead to big innings. Then after being on the DL, he couldn’t keep a consistent release point and that led to him being unable to keep the ball down. Thankfully it only took him a few starts to self-diagnose himself, and for the rest of the year 2007 was another fantastic season for the doctor.
Vernon Wells is an extrememly aggressive hitter, who loves to swing at the first pitch. And with good reason – last year he hit .343 when making contact with it. This season that average plummeted to .233 and he seemed to pop out to deep second base on the first offering every time he came up with the bases loaded. So what the heck was he swinging at? Here are the pitches he put in play this season on the first pitch:
A fair number of pop outs (this year Vernon was 2nd in the majors in infield fly percentage, also known as the “hidden strikeout” for their utter uselessness) and a lot of ground balls. But most of those are strikes, and pretty decent pitches, although on the outside half. He just didn’t do anything with them- one of the main criticisms was he was trying to pull that pitch too much (which would explain all the grounders as he rolls over on the ball to the shortstop). As for the pitches he didn’t put in play:
Now that’s strange. Wells is as pull-happy as they come, but he let a lot of pitches go by for strikes on the inner half. And look at all those foul balls on pitches right down the middle of the plate. From the first graph you would think he was being exclusively pitched away, but that’s not the case. It’s just that he made weak contact on pitches middle-away, and fouled off everything else. A foul ball happens when a player is either a little bit off the pitch or a little bit behind it. Does that sound like a shoulder injury maybe?
The green misses are a lot of breaking pitches. As anyone who watched a game or two knows, the way to get Wells to chase is to pitch him something low and away, but he didn’t go for them that badly- a lot of those pitches he swung and missed at are in places that are being called strikes half the time, and he laid off even more.
At least on the first pitch, Vernon wasn’t overly aggressive this year. If anything it was the opposite, he took too many quality pitches. His real problem was he was missing his pitch when he got it. If a hitter is fouling back mistakes left up and down the middle that they normally feast on, they’re going to end up getting behind in the count and looking bad on nasty pitches later in the at-bat while they should be rounding the bases. Here’s another one of the results of fouling off (or staring at) those pitches- instead he had to make more contact while behind in the count on pitcher’s pitches, which he didn’t hit well even in his excellent 2006 season.
|Count||2007 AB||2007 AVG||2006 AB||2006 AVG|