Baseball Prospectus is looking for a Public Data Services Director. Read the description here.

We keep waiting for Jacoby Ellsbury and Yadier Molina to have a showdown on the basepaths, but while we wait Pete Kozma and Craig Breslow snuck in to produce the most unexpected cat-and-mouse moment of the series.

Kozma stole third base at the front end of a double steal in the seventh inning. He represented the tying run, with one out, so the value of that stolen base was obvious at the time and obvious in retrospect: minutes later, Matt Carpenter hit a sacrifice fly, which narrowly scored Kozma; on the throw (which got away), trailing runner Jon Jay went to third, and on that throw (which also got away) Jay scored. A deficit was now a lead. The Red Sox goofball antics on the sacrifice fly had the largest win-expectancy shift of the inning, but none of it would have happened if Kozma hadn’t been on third.

The question is: What the heck was Kozma doing stealing third? Kozma had just entered as a pinch-runner for David Freese, but he’s not actually a burner, and he’s certainly not a basestealer. Kozma swiped three bags (in four attempts) this year, playing almost full-time. He’s got five steals in 185 regular-season games as a big leaguer (one of them on a busted suicide squeeze/wild pitch), and the last time he reached double figures in a season at any level was back in 2010, when he took 13. If stolen bases are your proxy for speed, the following players might have a better case for pinch-running over the past two seasons than Kozma: Vernon Wells, Todd Frazier, Albert Pujols, Jason Bay, Miguel Cabrera, John Jaso.

We’ve seen the Cardinals pull this off, in almost identical fashion, recently: on September 13th, Kozma pinch-ran for Brock Peterson. There was one out, and the Cardinals were trailing by one, and the runner was already on second base, just like in Thursday’s game. Kozma waited deep into the count and took off, stealing the base easily off Charlie Furbush. Furbush, in retrospect, was the crucial part of that play. He’s a big guy with a slow delivery, and in what amounts to a full season’s worth of innings he has allowed 27 stolen bases in his career, include seven (on nine attempts) of third base.

Maybe Matheny liked how well that worked and wanted to try it again. But Breslow is no Furbush. In his career, which amounts to two full seasons worth of innings, he has allowed just 22 stolen bases—to go with 24 caught stealings. Just 10 times has a runner tried to take third against him, and six times was thrown out. Breslow is close to elite at holding runners on.

Which brings us to the physics of the matchup in question. I wanted to see whether Breslow was predictable in his looks, giving Kozma a chance to get a huge jump; and whether Breslow was particularly slow to the plate. Here’s how the at-bat went, pitch by pitch. A capital L means Breslow took a long look at Kozma; a lower-case l means a short look; D means he looked down; d means a quick look down; and F means he looked forward, toward the plate.

1st pitch: L, D, l, d, L—pitch. It’s too soon to see whether Breslow will mix up the amount of looks he makes at the runner, but already we see that he’s mixing up the type of looks. The short looks convey to Kozma that, yes, the pitcher is paranoid; they make him look like he’s a twitching conspiracy theorist off his meds. There is no chance at this point that Kozma or Matheny would think Breslow is ignoring him. The pitch starts while Breslow has his eyes on Kozma, eliminating the possibility of a running start.

Non-pitch 1: L, D, F—aborted take-off by Kozma. The runner saw Breslow look toward home and took off, but Breslow wasn’t pitching yet. This seems like somewhat convincing evidence that Kozma hasn’t cracked Breslow’s code, if he’s trying. (Breslow turned too slowly to pick him off.)

2nd pitch: L, D, L, F, l, d, L—pitch. Four looks, of varying lengths.

3rd pitch: L, Dl, L, d, l, d, L—pitch. Four looks, against of varying lengths, and with varying looks away, too. Breslow continues to start his motion while looking at Kozma, or microseconds after turning away.

Non-pitch 2: l, d, l, D—leg lift and spin.

4th pitch: (inconclusive camera work, suggesting:) L, d, L—pitch

5th pitch: L, d, l, d, L—pitch.

Maybe, maybe there’s a pattern that Kozma cracked. It seems like he’d be safe with this conclusion: after Breslow goes through his quick check/quick look down/quick check routine, he takes one long look at the runner and then pitches. He doesn’t go through that routine on every pitch, but anytime you see this:

l, d, L

at any point in the sequence, Breslow went straight from the L to his pitch. I watched this at-bat four times with an index card, though, and it's hard to imagine that Kozma picked it up in real time. Maybe the Cardinals’ advance scouts picked this up, or maybe Kozma just took off #yolo style.

Breslow’s delivery time was about 1.6 seconds. That was consistent throughout the at-bat (once I got 1.5, but I’m bad at this), and it’s a slowish time to home. A pitcher who wants to shut down a runner tries for 1.2 or 1.3. But he doesn’t need to be nearly as quick for a runner on second, because the catcher’s throw to third is so much shorter. A good pop time for a catcher is about 1.8 or 1.9 seconds, but to third it’s more like 1.5 seconds. So Breslow’s 1.6 plus a reasonable pop time by the catcher would still add up to 3.1 seconds, too fast for average baserunners to steal. Kozma took 3.4 or 3.5 seconds to get to third. If not for Saltalamacchia dropping the pitch, he should have had a good chance to throw him out.

So why did Matheny do this? He’s got a pitcher who was unpredictable with his looks, who was paying attention to the runner, trying to stall him and trying to pick him off. He’s got a runner who is nowhere near elite. He’s got a situation where a caught stealing would cut his team’s chances of winning from about 38 percent to 27 percent. So why?

Unless the Cardinals had actually cracked Breslow’s looks (unlikely), I think the key was the trailing runner. In the scenario the Cardinals had, they still needed a base hit to tie the game. That’s tough to count on, but with just a runner on second base it would probably still be best to wait and hope and not get thrown out. With runners on first and second, though, the value of a stolen base is increased, because both runners move up; and the damage of a caught stealing is mitigated somewhat by the trailing runner’s advancement. If Kozma succeeds, then the Cardinals no longer need a hit to score the first (and most crucial) run; and, if they do get a hit, two runners score instead of one. If Kozma fails, then the Cardinals still tie the game with a single, which is the situation they already had, needing a hit. It’s more complicated than that, of course. An out cuts their chances of getting that hit in half; an extra-base hit scores two runs in a non-steal scenario, but only one if Kozma is thrown out; etc. But if you’re Matheny, you look at those baserunners and see a stolen base opening up new possibilities for you to gain a tie or a lead; you see a caught stealing as harmful but still preserving the same route toward a tie that you already had.

The Cardinals had 38 percent win probability at that point. With a caught stealing, it drops to 27 percent. But with a successful double steal, it goes up to 48 percent. Factoring in the added possibility of an errant throw advancing the runners further, the Cardinals needed to have just about a coin flip’s chances of succeeding to make it worthwhile. I’m not sure they had that coin flip’s chances, to be honest, given the pitcher and the runner and the fact that the Red Sox saw the steal coming. But maybe they did. It worked.

Or you could say that Daniel Descalso walking on the next pitch made the whole thing moot, and that Kozma would have scored on the sacrifice fly anyway, and that Matheny took a big chance for nothing. But would he have walked on the next pitch? Was Breslow thinking about the value of setting up a double play? Was Descalso more patient because he thought Breslow might be thinking about walking him to set up the double play? These are the unknown possibilities. The possibilities are part of the decision, and the decision is hard.


Matheny got to make a lot of decisions in this game. He got to decide to keep Carlos Martinez in to face David Ortiz (as the tying run!) in the eighth inning, rather than going to Randy Choate or his closer, Trevor Rosenthal. Martinez has been even better than his surface numbers suggest as a reliever, with about a 2.05 FIP in 33 innings (including postseason) in the majors. He hasn’t allowed a home run. It wouldn’t have struck anybody as odd if Matheny had gone to Rosenthal instead of a LOOGY; maybe he just figured Martinez is as good as any closer right now, and followed the same logic in leaving him in.

Martinez has had all the traditional platoon splits in his minor league career, and his whiff rate is 40 percent higher against righties than lefties, and Choate has truly been fantastic against lefties for the past five seasons. Choate’s true talent level against lefties right now is something like a .475 OPS; Martinez’s is, I’d wager, around .650. Odd move that worked out, more or less.

Matheny had a similar decision earlier in the game when he had to decide whether to leave Michael Wacha in to face Ortiz, or bring in Choate (or Kevin Siegrist) then. Ortiz homered. In an exceptionally small sample, Wacha hasn’t shown any trouble getting through the order a third time as a big leaguer, but for a pitcher who relies on a very simple repertoire—a repertoire that leans so heavily on heat that he might be especially wounded by routine velocity loss later in a game—a manager might wisely be especially sensitive to mid- or late-game struggles. There’s also Wacha’s age: 22. While almost all pitchers get worse as they go through the lineup a third time, young pitchers seem to have an even bigger split.

I looked at all 42 pitchers, age 22 or younger, who went deep enough in a game over the past three seasons to face a lineup a third time. Consistently, the pitchers’ FIPs go up the third time through, but, again, that’s typical. I weighted their increases by how many innings they threw (as good pitchers are more likely to make it three times through the order, skewing the totals) and, on average, a 22-or-younger pitcher sees his FIP go up .49 runs from the second time through the order to the third.

I repeated this with 30- and 31-year-old pitchers in the same time frame. There are 60 pitchers in the sample, throwing about 2,000 innings in the split. Their FIPs rose, on average, by .37 runs, making them perhaps safer bets the third time through the order than a kid like Wacha. This isn’t a huge difference. It’s also not a surprising difference.

Thank you for reading

This is a free article. If you enjoyed it, consider subscribing to Baseball Prospectus. Subscriptions support ongoing public baseball research and analysis in an increasingly proprietary environment.

Subscribe now
You need to be logged in to comment. Login or Subscribe
great breakdown - what about how they did a 'mock' steal a few pitches before? Was that an aborted attempt or just a deke? It seemed like either Kozema was going for the best jump ever or he just couldn't wait to get started, because it didn't seem like he was taking off for only a few steps.

2nd - wouldn't it seem likely that the aborted attempt would squelch the likelihood of an attempt, having killed the element of surprise?

I hang this one on Salty--that's what his teammates call him, per Tim McC.

Salty had plenty of time to gun down the runner at 3rd and the ball got stuck in his mitt. This goes back to some old-style love for the SB and aggressive calls--you make your opponent make a play. In this case Salty couldn't deliver.

Big respect to Matheny for sticking with Martinez against Ortiz. Interesting that 'only' giving up a single and getting to the righty is considered a victory.