July 27, 2010
Cracking the Pitch Sequence Code
An Arranged Marriage of Pitch Data and In-Game Psychology
What follows is a story of a pitcher who lost command of his fastball, and a hitter who approached him as if he could throw it to a teacup. The Mets were clinging to a 3-1 lead over the Giants on July 18 as their game entered the late innings at AT&T Park. After another eight-frame master class from Johan Santana, Mets manager Jerry Manuel called on Francisco Rodriguez to lock down a victory. It was a game the Mets desperately needed; they opened the second half of the season by scoring just four runs in their first three games, and if the week following this game is any indication, they aren’t good enough to waste Santana’s brilliance and still make a run at the postseason.
Now that we’ve set the scene, let’s think along with its principal players, and observe how Rodriguez and his opponents adapt—or fail to adapt—to the Mets closer’s uncharacteristic lack of a reliable fastball. We’ll follow K-Rod’s two innings in hopes of learning a thing or two about the mysterious art of pitch sequencing, and see how the information Rodriguez sends with each pitch of this outing may be more predictive of what he’ll throw next than simply relying on his overall tendencies.
Rodriguez enters the game to face one power hitter, Pablo Sandoval (whom we denote as a “power hitter” on league-wide reputation rather than performance) before taking on the bottom third of the Giants’ lineup: Juan Uribe, Eli Whiteside (the goat of our story, but that comes later) and the pitcher’s spot. With a pitcher’s park and a two-run lead providing insurance, K-Rod appropriately attacks Sandoval with nothing but fastballs.
Rodriguez is a good closer having a good season, but right from the outset it’s clear that he isn‘t quite himself. His first two pitches are low, and the next, a called strike, is up in the zone and center cut. He makes a quality pitch to even the count at 2-2, but Sandoval runs the count full before fouling off a very hittable fastball. Needing simply to put the ball in the strike zone with his 3-2 pitch, K-Rod misses badly, up and away. In baseball circles, a righty missing up and to the right with the fastball can be a telltale indication of fatigue, because the pitcher has failed to close his body off to home plate and has released the ball too early, without “finishing” his motion. Whether you interpret it that way or not, Rodriguez has failed in his one overriding objective: not walking the leadoff hitter—and we can reasonably infer that’s what he’s thinking, because of his eight consecutive fastballs to Sandoval.
If the antennae aren’t raised in the Giants dugout after Sandoval’s walk, they certainly should be as the inning progresses. K-Rod attacks Uribe, who likes fastballs almost as much as he likes cheeseburgers, with straight heat; he misses his spots and eventually gives up a lined shot into center for a single. Whiteside, in a bunt situation, fouls off a low and inside fastball (with Mets catcher Henry Blanco set up away) and then bunts the runners over on a high and inside fastball (with Blanco again set up away): second and third, one out. Pinch-hitter Travis Ishikawa sees a changeup miss badly, and ends up smacking a 2-2 fastball into center field for a two-run single to tie the game at 3-3. The next pitch is a fastball “right down the pipe,” as a baseball man might say through a wad of dip, and Andres Torres yanks it into the right-field corner for a double, Ishikawa’s potential walk-off run holding at third.
The rest of the inning, you may recall, is well-publicized: after a rather bizarre pitchout—why would the Giants squeeze after the laser show we’ve just witnessed, and against a pitcher who can’t locate his fastball? —Freddy Sanchez grounds one to third baseman David Wright, who throws high to home plate and Ishikawa gets under Blanco’s tag, only to be called out by Phil Cuzzi. (Blanco’s fantastic line after the game: “He was safe all the way. Good for us.”) K-Rod then gets Aubrey Huff to ground out to end the inning. Of the eight pitches to Sanchez and Huff, Rodriguez throws five off-speed, all for strikes, with both ground balls coming on off-speed offerings.
The Mets take a 4-3 lead in their half of the 10th, Manuel, the same man who thought so highly of Rodriguez that he anticipated a Giants' squeeze, somehow decides to bring Rodriguez back for the 10th. On the Manuel scale of dumb decisions, this is somewhere between start Jenrry Mejia’s arbitration clock so he can be a mop-up reliever on a team that isn’t going anywhere, and bring backup catcher Omir Santos out of the bullpen, and pinch-hit him against a guy who throws 100 mph with two outs in the ninth inning and the tying and winning runs on base.
But in the game K-Rod remains, and to his credit, he makes an adjustment. In fact, K-Rod’s philosophy in the 10th is virtually opposite his ninth-inning plan: with no insurance run and a disastrous inning still fresh in the memory, he is going with what works, and to same-sided hitters, that’s the curveball. (To a lefty, it would be the changeup— that’s how he got Huff in the ninth). Rodriguez throws two excellent curves to Buster Posey and gets him on a grounder to shortstop. He then misses with a curve to Pat Burrell, and gets him on a fly ball to center with his best fastball of the day, which on this day is like being the tallest midget. Against Edgar Renteria, K-Rod misses with a good curve and then comes down the line with fastballs, missing his spot each time, and lo and behold, Renteria slams the fourth consecutive heater into the left-field corner for a double. Manuel then intentionally walks Uribe to put the winning run on base (I’ll let the real sabermetricians assess that move). So it’s first and second, two outs, and Whiteside has a chance to be a hero. But he wouldn’t be a hero, not today, because he was not paying attention.
For Whiteside, this is an absolutely golden opportunity, the kind of chance you might receive only a few times over a long career. It’s not just the walk-off situation, of course, but the circumstances surrounding it that so greatly increase his chances of success. First, he is facing a two-pitch pitcher for the second time in one day. (He can’t be worried about a changeup because K-Rod hasn’t shown one to a righty yet, though his PITCHf/x data indicates he is generally willing to mix them in) A primary reason relievers can overachieve relative to their ability is that batters only see them once; to get a second look on the same day, especially in today’s game of specialty relievers, is something of a rarity. Second, for a short reliever, Rodriguez has given the Giants a wealth of information about his current form, and his confidence in his various pitches. Think about what Whiteside has seen since K-Rod entered the game: a) no fewer than four absolute rods, all hit off fastballs; b) a leadoff walk with a two-run lead where each pitch was a fastball; c) in the 10thinning, Rodriguez starting three right-handed hitters (Whiteside is right-handed) with a curveball; and d) four out of six curveballs thrown for strikes, with all but one out—excluding Whiteside’s previous appearance, a sacrifice bunt off a fastball—coming via an off-speed pitch.
At the big-league level, a situation like this is exceedingly rare: Rodriguez is afraid to have his fastball ripped a fifth time, and the whole stadium knows it. A hitter can always sit on one pitch up to a two-strike count, since he can afford to be guess wrong without being called out; here, at the very least, Whiteside should be thinking it inconceivable that Rodriguez could get to two strikes on him without bending in a curve.
Now, to the pitch sequence. You’re not going to believe this, but Rodriguez starts Whiteside with a curveball. Whiteside, similarly shocked, buckles his knees and takes a strike. To be fair, K-Rod broke off an absolute hammer— it’s a little above the knees on the outside black—but Whiteside’s reaction to the pitch reveals that he was sitting fastball. The next pitch is a fastball— another curve might be too obvious, even for Whiteside—and K-Rod misses away. Three curves follow, all away and in the dirt; Whiteside goes fishing for the first and just tips the second to keep himself alive. He manages to lay off the third, and a 2-2 fastball misses to run the count full. Needing to make his best pitch, Rodriguez dances with what brought him here, but with a full count it’s too late for Whiteside to be sitting curve: he strikes out swinging on a hittable curveball that broke sharply but over the middle of the plate.
I don’t think it’s bragging to say that I, sitting at home watching this unfold, called each pitch to Whiteside before it happened. I’m not reading tea leaves; Rodriguez had lost his fastball and didn’t want to get burned again, and I probably got a little lucky to think along with K-Rod and Blanco. But pitch sequencing in general is something of a virgin territory in the sabermetric community. Cracking the code is a tall order, since every pitcher on some level aspires to complete randomness in his pitch selection, assuming he could throw all his pitches with 100% efficiency. (As I write this, I’m watching Stephen Strasburg, and thinking he might become the first pitcher ever to simply pick his pitches out of a hat.)
Randomness in pitch sequencing isn’t so highly valued at every level of baseball. At the Division III college level, where I was a backup catcher for three years at Tufts we had a much different set of rules, which is to say that we had rules in the first place. The only thing D-III pitchers share with big-leaguers is the general axiom that if they execute their pitches just the way they want to, the batter has no chance, but at that lower level of play, pitchers are just trying to pitch to their own strengths, rather than keep the hitter guessing.
With that in mind, we had great success calling pitches based mostly on the batter’s reaction to the previous pitch. Jackknifed on inside heat? Get-me-over curve. Fights off a fastball away? Smoke him inside. Ahead on the changeup? Slow him up even more with a curve. Ripped an inside fastball foul? Inside changeup (now there’s a pitch call you’ll hardly ever see in The Show). With a little variation, the pitches would essentially call themselves, and no one at that level had the means to discern a pattern; we were just trying to throw the pitch for which the batter might be least physically prepared. We weren’t outguessing them.
You’ll notice this article has been devoid of statistics of any kind. That is for two reasons: a) I am not a sabermetrician (worse, I’m an English major), and b) in parsing the information a pitcher gives us to work out what he might throw next, much of the relevant data arrives not through PITCHf/x or other sabermetric means, but by the pitcher’s actions that same day. We have myriad fantastic resources to help us assess the average change in run value when following one pitch type with another. But those data struggle in short-term practical application; much of the information necessary to predict what a pitcher will throw next manifests itself on a day-to-day basis.
Pitchers will always have their reasons for why they can’t throw a certain pitch on a certain day: “It was too cold and my curve wouldn’t break;” “I just never had my splitter in the pen;” and recently, “I punched a door.” Obviously, “PITCHf/x told me to throw it” is pretty far down on that list, unless you‘re Brian Bannister or Max Scherzer. If we’re going to crack this code, it’s going to be in much the same way that we scout players, with a marriage of sabermetrics and old-fashioned in-game psychology. Pitchers are taking the mound every day and giving us information with each pitch that will be most relevant only that day, and if we’re going to complete a working, practical knowledge of the game, we must learn what criteria are predictive of a much smaller future sample size—the next pitch, the next at-bat, rather than the next month or full season. That Mets-Giants game was a great, if rather extraordinary example to show the importance of the day-to-day psychology of the game. Though incorporating this information with any consistency is at present a bridge too far—it can pop up in just a few pitches, and becomes irrelevant almost as quickly—to ignore it would be to neglect information that players and coaches spend years honing their skills to acquire. For now, we can ask Eli Whiteside what he thinks is coming, and expect the opposite.
Will Woods is a contributor to Baseball Prospectus