October 31, 2015

Playoff Prospectus

Assessing the Managers' Moves in Game 3

by Rian Watt

A major-league plate appearance can resolve in a thousand different ways.

A base hit to left. A wild pitch that scores the game-winning run. A screamer down the first-base line, just off the glove of a diving fielder. That same screamer, but with the first baseman shaded just a hair too far over toward second base to get a glove on it.

A thousand different ways, all of which can be arrayed neatly before us, for any given batter, as a set of frequency distributions. We do it all the time: We call a batter a .300 hitter when his plate appearances have resulted in a base hit 30 percent of the time. We call him a .360 OBP guy when his plate appearances have resulted in a trip to first base or beyond 36 percent of the time. And although the names for the distributions get fancier as we grow more sophisticated—SLG, OPS, TAv—they’re all the same thing, at heart: records of the frequency with which certain results transpired.

And we make decisions about the future based on those records of the past: In spring training, we’ll advocate the award of 700 plate appearances—standard, for a full season—to the guy with the .360 OBP over his teammate with the .320 OBP. We know what we want, and, given enough plate appearances to work with, we know we’re going to get it from a guy who’s done it more frequently (there’s that word again!) in the past. The numbers tell us so, and they tell us ever better the more data we give them.

I bring this up now because I’m pretty sure that October brings with it—along with crisp air, fall leaves, and my annual commitment to a lifelong hatred of Halloween—a significant breakdown in the utility of frequency distributions. That’s because, looked at through the lens of October, there are actually only two ways in which a major-league plate appearance can resolve.

Out.

And, Not Out.

Let’s say you’re Terry Collins, and it’s the sixth inning of Game Three of the World Series. Your ace starter, Noah Syndergaard, is on the mound. He’s deep into his game, and he’s got the bases loaded with two outs in the inning. Your opponent's right fielder, Alex Rios, is striding to the plate, and you have a left-hander---Jon Niese---warming in the bullpen. Meanwhile, Kendrys Morales, the Royals’ switch-hitting designated hitter, is crouched on the top dugout step, waiting to see if you’ll stick with Syndergaard or jump to Niese. What do you do?

Well, one thing you could do is look at the numbers—those invaluable frequencies. And (lucky you!) there are a lot of those to consider. Here are a few of them:

• In his career, Syndergaard has allowed a .691 OPS to left-handed batters, and a .601 OPS to right-handers (like Rios).
• In his career, Syndergaard has allowed a .777 OPS in high-leverage situations---a mark far higher than those he put up in medium- (.597) and low-leverage (.647) situations.
• In his career, Syndergaard has allowed an .860 OPS to hitters recording their third plate appearance against him that game, as Rios was about to.
• In his career, Morales has put up a .734 OPS as a right-hander facing a left-hander (like Niese).

I could go on. I won’t, though, because the numbers, themselves, aren’t the point. The point is that, when the world has boiled down to its essence—Out, or Not Out—numbers (at least in the form of frequency distributions) cease to matter very much. Which of those numbers were you planning to rely on? And why did you pick that one over the others available to you? I’m not sure there’s a good answer to that question. Collins simply could not allow Rios to reach. He needed an out, and the increased probability of an out was no longer good enough.

Sure, if the Mets and Royals were able to play out the Rios-Syndergaard/Niese-Morales showdown a thousand times, then there would probably be an objectively correct move for Collins to make. I don’t have the technical skills to do the #gorymath to find it, but I’m fairly confident that it exists.

But I’m just as confident (which is to say, a hair under certain) that there was no objectively right move for Collins to make in the sixth inning of yesterday’s game. He needed an out. He knew his players. And he decided that Noah Syndergaard was the man he wanted in the game for that moment. That’s subjective, yes. But it’s also not necessarily wrong.

A great many people on Twitter last night—myself included—disagreed with Collins’ choice. Without the benefit of knowing Syndergaard personally, as Collins does, we saw that he’s been terrible against hitters the third time through the order, and advocated for the entry of a fresh arm out of the pen. I don’t think that answer is wrong either. But I think it is wrong to regard it as an answer with any more process validity than Collins’. The point is, I can’t defend my answer any better than I can defend Collins’. And he had more information.

Theo Epstein said something to Joe Posnanski, late in 2011, that’s stuck with me ever since I heard it: “The game is best understood at 10,000 feet. But it is best enjoyed from the front row.” I think that that’s highly relevant to the discussion at hand. The game is best understood at 10,000 feet. Part of the joy of a site like Baseball Prospectus is that we can do just that. And if you’re building a roster for a 162-game season and you’re not basing your decisions in large part on the best frequency distributions you and your front office team can possibly describe, you’re not doing a very good job of building that roster.

But I think we in the analytic community do ourselves a disservice if we try too hard to apply the logic of a frequency distribution to the exigencies of a moment. In the bottom half of the sixth inning, a few minutes after Collins had left Syndergaard in the game to face Rios, and watched the former retire the latter on a groundball up the middle, the Fox broadcast booth noted that Collins believed that “Syndergaard, even at 100 pitches, was better than any arm in the bullpen.” From 10,000 feet, that’s probably nonsense. From the front row, it might just be right.

This was meant to be a piece about tactics in Game Three of the World Series, and when I started writing it, it probably was. But as I analyzed each managerial decision I saw last night through the lens of the numbers I found to surround them, Epstein’s quote kept running through my head. Strategies are decisions you make at 10,000 feet. Tactics are decisions you make from the front row. And the same rules don’t apply.

And so I don’t think I’d be telling you much if I dove too deeply into the decision, by Ned Yost, to debut Raul Mondesi in the fifth inning, thereby burning a key bench option in a role he wasn’t suited for. And I don’t have a lot to add about the decision, by Collins, to let Syndergaard bat for himself with one out in the fourth and two runners on. I don’t feel like I have information appropriate for the analysis of those tactical decisions, when the game has been stripped so bare and the sample sizes are so small.

But Terry Collins has that information. And so does Ned Yost.

Let me leave you with one last thought. If you were standing near the Valley of Elah, and you saw a thousand Goliaths marching forth from the Philistine ranks, would you send a thousand Davids to face them one by one? Of course not. The first David would try the sling routine, win, and Davids two through 1,000 would be smashed to pieces by a sequence of Goliaths that had learned how to duck.

But Saul and the Israelites didn’t need a thousand Davids to beat a thousand Goliaths. They just needed one. And they found the right man for the job. On Friday night in New York, Terry Collins did just the same.

Rian Watt is an author of Baseball Prospectus. Follow @rianwatt
Click here to see Rian's other articles. You can contact Rian by clicking here