Declining the Penalty

Sunday’s getaway-day game between the San Francisco Giants and the host Atlanta Braves featured a play we don’t see too often, and a choice we see even less often. Buster Posey was awarded first base on catcher’s interference by Braves backstop Evan Gattis, but manager Bruce Bochy waved off the award. If Posey had gotten a clean hit, such a move would be obvious, but the actual scenario made it a very interesting decision.

Catcher’s interference is that rare thing in baseball, an umpire’s call that acts like a football penalty because it can be declined. Go to a football analytics site, and you can probably find a learned opinion on whether it’s better to take a five-yard penalty and replay the down, or take the seven yards you gained on the play but expend the down. That’s a pretty close call, and what Bochy faced was also quite close.

The incident came early, top of the first inning. The count was 2-2 on Posey, Hunter Pence just having gone second to third on a wild pitch by Braves starter Alex Wood. Posey ticked Gattis’ glove as he bounced a ball to first base. Freddie Freeman made the play at first unassisted as Pence crossed the plate. Home-plate umpire Alan Porter’s call of catcher’s interference, though, sent Pence back to third (unforced runners don’t advance on CI) and Posey to first.

Bochy thought about this—how much, we can speculate—and came out of the dugout. This is something he needed to do: umpires aren’t required to offer a manager the choice, only to enforce the penalty. If he wants to decline, the manager must take the initiative. Bochy did that, taking the early run while giving up the second out of the inning.

(For a look at the play and its aftermath, and some of the Giants TV team’s commentary on the matter, look at the highlight here at MLB.com.)

Did this make sense? There are a few ways we can look at the matter.

First would be to use Expected Runs, the average amount of runs one would expect a team to score given its base-out situation. I will use 2013 numbers, because I have those available and because this season’s numbers would be a much smaller sample, with its attendant dangers. Runs per team per game last year were 4.166 in all baseball, as opposed to 4.013 this year in the National League, but since Turner Field plays around a 104 Park Factor, this almost perfectly balances out.

With runners at the corners and one out—the catcher’s interference accepted state—Expected Runs are 1.1383. With a run in, nobody on, and two outs—the catcher’s interference declined—Expected Runs are 1.0918. These numbers favor taking the CI, and by nearly five-hundredths of a run. For comparison’s sake, in the same run environment sending someone to steal second with a 78 percent success rate would net you the same gain. Not a game breaker, but a good play, so Bochy’s move looks bad in this light.

Let’s shift the perspective to our favorite toy here at THT: the WPA Inquirer. Dave Studeman announced on Friday that it was back in operation at the site, and better than ever. What a fine stroke of luck. Time to put it through its paces.

The Inquirer calculates Win Expectancy rather than Expected Runs, and is thus a more direct gauge of how a play affects a team’s game chances. I would expect almost no deviation from my Expected Runs calculation, though, since the play came so early that the overall state of the game is still close to an ideal situation for Expected Runs to mirror winning chances.

That turns out not to be the case. Or in layman’s terms, I’m wrong.

The WPA Inquirer can measure run environments in only half-run increments: 4.0, 4.5, and so on. But since I just got through figuring that the environment at Turner that day was effectively 4.166 per team, exactly one-third of the way between 4.0 and 4.5, it’s not tough to interpolate between the two and reach accurate numbers.

For interference accepted, Win Expectancy at 4.0 R/G is 57.0 percent; at 4.5 it’s 56.5; for Turner that day, call it 56.83 percent. For interference declined, the 4.0 Win Expectancy is 57.4 percent; for 4.5 it’s 56.4; in the Turner environment it comes to 57.07 percent. In a higher run environment, the Inquirer would have sided with the Expected runs numbers and taken the catcher’s interference. In today’s pitcher’s era, though, it says decline, by a margin of a quarter-percent of a win.

The Expected Runs table says accept; the WPA Inquirer measuring Win Expectancy says decline. I’d call it a split decision, except that Studeman didn’t resurrect the Inquirer from all the traumas of creating the new website format for it to render split decisions. I would be willing to yield to its verdict … were it not for one little detail.

Both these methods assume average situations underlying them: average pitchers on the mound, average batters coming up. But no mathematical average ever threw a ball or swung a bat. There were specific people doing those jobs, and they deserve at least a look before we close the books.

Due at the plate after Posey was the cleanup hitter, Michael Morse. Morse’s nickname is “The Beast,” and he’s been living up to it this young season, posting a .302/.352/.615 slash-line and a 175 OPS+ through Sunday’s game. You want someone performing that way batting in high-leverage moments, so his excellence can have maximum effect on your winning chances. The interference-accepted situation (first and third, one down) comes in at a Leverage Index around 1.9; the declined state (none on, two gone) is about 0.35.

Numbers like that swing the game your way, hard. Taking away Morse’s chance to clear the bases and keep the inning rolling was a mistake—if you believe in those numbers.

He didn’t have numbers like that last season, putting up a mere 84 OPS+ between Seattle and Baltimore. Lifetime, his OPS+ comes out at 121, which is certainly better, but not nearly as beastly. In fact, it almost exactly matches the season, and career, ERA+ of the pitcher he faced, Alex Wood: 123 and 122 respectively. (Wood’s peripherals do vary from that: better last year, worse this year, nothing decisive.)

If you trust this year’s numbers to reflect what Morse would have done, you should have accepted the catcher’s interference. If you think his true talent level is reflected by his career numbers, or last year’s numbers, you decline and take the run now. But come to think of it, if you believed last year’s numbers, why would you have Morse batting cleanup?

Average numbers are a fine tool, but they are not the end of the analysis. When it’s close either way, who you have performing in the specific situation ought to have a strong impact on what you decide. The whole image and cachet of a cleanup hitter is that you want men on base in front of him so he can drive them home. Morse had been doing that job as well as anyone could want, but when a crunch came, Bochy did not trust him with the job.

It may not be as bad as that. Bochy might have thought the odds favored taking the run so much that it wouldn’t really matter who was coming to bat. The incident may not reveal a sneaking lack of confidence in Morse’s hot first five weeks, even a warranted lack of confidence. And since the Giants won 4-1, it really is not a natural locus of second-guessing.

I did not mean to stir up trouble in the Giants’ clubhouse when I first started looking at this incident. I just wanted to look at an interesting move by a manager, and figure out whether he made the right call. The inquiry just went a little past the numbers where I thought it would end up.

(h/t to FanGraphs’ Wendy Thurm for cluing me in to this play.)

Print Friendly
 Share on Facebook0Tweet about this on Twitter8Share on Google+2Share on Reddit0Email this to someone
« Previous: Return of the WPA Inquirer
Next: The True Origins of Baseball (Scientifically Speaking) »

Comments

  1. Anon21 said...

    Additional context: the Braves are in the midst of an incredible streak of offensive futility, such that the first run essentially guaranteed the Giants no worse than a game that would have to be suspended after 36 innings.

  2. Steven said...

    “But come to think of it, if you believed last year’s numbers, why would you have Morse batting cleanup?”

    Brandon Phillips agrees.

  3. LHPSU said...

    Michael Morse has grounded into 5 double plays in 105 PAs this season. Even setting aside the effect that failing to score a runner from 3rd with less than two outs has, the fact that Morse is a legitimate double-play risk must be taken into account, Posey not being exactly Billy Hamilton.

  4. Steven said...

    Same thing happened with the Yankees that day they played a double header against the Cubs. Ellsbury grounded out bringing a run to the plate and Girardi declined the penalty.

    Girardi later said he encountered the rule in 1990 when Bobby Bonilla homered while hitting his glove but elected to keep the home run.

    • Aaron (UK) said...

      Not the most difficult decision imaginable for Bonilla. Though perhaps he could have taken 0.20 of a homer every year for 25 years instead?

  5. Shane Tourtellotte said...

    Anon21: There was another streak on at the time. All the runs the Giants scored in the first two games of the Atlanta series had come on home runs. You can hear the announcing team make oblique reference to this. That might have nudged Bochy toward the move he made.

    LHPSU: I saw this in his numbers, and would have noted it had I not already been running long. Morse’s GDP rate is running well ahead of last year’s or his career rate, much like his positive offensive numbers have been. If you think the good stuff is ready to regress, you ought to be thinking likewise for the bad stuff.

    Steven: I had not heard about Girardi and the Yankees before. It was the same situation in a way: accept and have runners on the corners with one out, decline and have a run in with two outs and the bases clear, with the cleanup batter coming up either way. Run Expectancy would thus say take the interference. But this time it was the bottom of the fifth, with the Yankees already ahead 2-0. In this case, declining and taking the run nets you a Win Expectancy gain of more than one percentage point over accepting. Also, cleanup hitter Alfonso Soriano was doing significantly worse than Michael Morse, with a .737 OPS before the day’s play began. Girardi’s move is more clearly the correct one than Bochy’s, which does not necessarily mean Bochy was wrong.

  6. tz said...

    Bochy’s decision parallels football coaches who decline a penalty on a short field-goal attempt so they can have a 100% chance of 3 points instead of a <100% chance of getting a touchdown.

  7. said...

    I too am surprised that the WPA Inquirer comes up with a different answer than run expectancy, but the numbers are so close I probably wouldn’t put much stock in the small difference.

    Regardless, I think this is a classic example of taking a bird in the hand over a more probabilistic two birds in the bush.

    • said...

      Ah, what he said, much more succinct and better said.

      Though I think I added some good points on your thought process that you could think about the next time you ruminate on a sequence of batters.

  8. said...

    There was another angle that was not covered in the article: the Giants this season was something like 15-1 at the time of this game when they scored first in the game, which they did with this score.

    Plus, is it just a matter of not trusting Morse to “do his job” or another matter? Maybe it is trusting that Bumgarner and the bullpen (and the offense later) could hold that one run lead, instead of not believing that Morse can do more with the situation. Maybe it is respecting the pitcher they were facing, Alex Wood has been a very good pitcher in the majors so far, who could maybe work out of the situation without giving up a run.

    That is the problem sometimes with an academic approach to a problem than a pragmatic one. This reminds me of the economist joke with the punch line, “assume you have a can opener,” where the pragmatic trumps the academic. Perhaps it is as simple as preferring the bird in the hand than two in the bush, as tz commented. And 1.13 vs 1.08, there is only a difference when you have many difference instances of this happening. For any one time instance, it is one run either way for the most part.

    And, since this is a thought experiment, maybe you can run a simulation of this situation, using Morse as the following batter. If you run the simulation 1,000,000 times, what percentage of the time falls into these three scenarios: zero runs, one run, two+ runs.

    Since Bochy wanted the run in hand, that makes two things important here. First, how often do you end up with no runs with Morse up? Sure, he has a great batting line, but even as well as he is doing, he’s still making outs 65% of the time. Plus, he’s facing Wood, who is a very good pitcher who gets outs at an even greater percentage, 70% at the moment. And when Morse walks, roughly 5% of the time, or make an out that do not bring the runner in, that brings up Brandon Belt, a lefty hitter who has been pretty good since the 2013 season started, but who has been in a bad stretch of hitting where he’s clearly searching (at one point something like 8-9 K’s in 12 ABs) and slowly working out of it, facing a left-handed pitcher in Wood. Followed by a hitter who has been struggling all season in Sandoval.

    As you astutely noted, the data deals with average situations, and you rightly questioned the move given Morse being next, but you didn’t take it far enough. The vast majority of the time, Morse does not drive in the run, and you are left with Belt and Sandoval needing to step up and drive in the runs. In the past, that’s probably a good bet, especially Panda in 2009 and 2011, or just him being healthy and hitting, as he hit well in 2012 and 2013 when he wasn’t injured or working to recover his stroke after an injury, and Belt was very good for most of 2013 himself. But right now, Sandoval isn’t hitting for much of anything, and Belt after a very good start, appears to be regressing to the mean relative to his very hot start, and is working to find his stroke again.

    The other thing important from the simulation is how often do the Giants end up with 2 or more runs? Let’s call that X%. Is it worth giving up the sure run, on the oft-chance that X% of the time, you end up with multiple runs?

    And if you are going to end up with one run most of the time anyway, which is the result I expect, why bust his chops (I know you are not intending this, but your readers might, amazingly, there are a lot of Giants fans who appear to not like Sabean or Bochy) for taking that route instead of gambling to get more runs?

  9. Sabean Wannabe said...

    I think the ancillary factors favor declining the CI. Morse is both a strikeout and GIDP threat, and runs had been hard to come by all series. Had this been in Colorado, I’ll bet Bochy takes the CI.

    I also think this is a situation where the saber community cannot see the forest for the trees. The Law of Large Numbers still applies. These types of statistical measures are meant to guide what to do over 162 games/1,500 innings/5,000+ at bats. The difference of 1.14 vs. 1.0 favors taking the CI, however that means you would on average need seven successes to make up for every one failure. Undoubtedly, this situation will not happen that many times in a season. Clearly, the risk of failure is too high. Take the run.

Leave a Reply

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>

Current ye@r *