The Guardians Saved Their Closer by Ben Clemens September 23, 2022 © Ken Blaze-USA TODAY Sports Tuesday night, the White Sox and Guardians played to a standstill over the first eight innings of a game that would help decide the fate of the American League Central. Of course, getting late into the game with a chance to win suits both teams just fine. The White Sox have Liam Hendriks anchoring their bullpen, while the Guardians have Emmanuel Clase in the same role in theirs. Hendriks pitched a scoreless top of the ninth. But even with the Guardians’ two best setup men, Trevor Stephan and James Karinchak, out of the game, Clase cooled his heels in the Cleveland bullpen while Enyel De Los Santos matched Hendriks out for out. What was Terry Francona up to? He was, in fact, playing the percentages. “Never save your closer” is a modern analytical truism, but it wasn’t designed for the zombie runner rule. When every inning works the same, you should always get your best pitchers into the game post haste. The only difference between the ninth and 10th innings used to be that you might not get to play the 10th. That’s not the case anymore. As I examined in 2020, making the 10th inning a higher-scoring affair than the ninth changes optimal pitcher usage. When all the action is in the 10th, it follows that you want your best arm pitching then. Walking the tightrope in extra innings and escaping without a run allowed is quite difficult; it’s an inherently higher-leverage spot, which means using your best reliever pays off. The easiest way to think about it is with someone like Ryan Helsley or Edwin Díaz, a pitcher who frequently ignores runners on base thanks to strikeouts. Outs aren’t all created equal. With a runner on first and less than two outs, a groundout is the best kind of out thanks to the chance of a double play. With a runner on third and less than two outs, strikeouts and popups reign supreme. With the bases empty, everything is the same. The base/out state determines a lot about the optimal style of pitching. In regular innings, that doesn’t matter too much, because the bases start empty in every inning. The best pitcher is the one who gets the most outs. Guarantee a runner on second base with no one out, and strikeouts get markedly better. There’s no force play in effect, and in-play outs are more likely both to advance the runner to third with one out and to allow him to score from there. Clase doesn’t exactly fit that mold, but he makes up for it by getting a ton of outs. He’s a good strikeout pitcher, but he’s a great weak contact pitcher thanks to his 100 mph cutter. Weak fly balls are as good as strikeouts with a runner on second, and Clase is just so good overall that he also limits scoring via sheer weight of outs. By putting in De Los Santos for the ninth and saving Clase for a higher-impact spot, Francona was putting the idea of comparative advantage into practice, setting both of his players up so they were in the spot where they were best situated to help the team. Which managers do this overall? I came up with a rough way of estimating it. It will necessarily be rough, because I want to include every game played this year rather than merely a few anecdotal examples. On any given day, there are myriad factors impacting which pitcher goes in when, only some of which are related to the rules of the game. The closer might be unavailable, or need some work in a strange spot, or the opposing lineup might favor one pitcher over another. Coming up with a holistic way of examining every single bullpen decision would be more trouble than it’s worth. Instead, I went for something simple. I took two different scenarios: a tie game in the ninth inning, and any score in the 10th inning. For each of those two scenarios, I noted which relievers had appeared. Then, weighted by the number of batters faced accrued by each pitcher, I calculated each team’s aggregate strikeout rate in each situation. Why strikeout rate? Because while it’s not a perfect proxy — as I mentioned, Clase is a good choice for this gambit despite a middling strikeout rate — it’s the best one-number metric available for using pitchers best-suited to the new extra inning rule appropriately. That’s a bunch of words, so let’s pause for a quick example. Say, for the sake of argument, that the Mets had faced exactly four batters in a tied ninth-inning situation all year. Let’s say that Díaz took three of those and has a 50% strikeout rate on the year, while Seth Lugo and his 25% strikeout rate took the other one. That works out to (.5*3 + .25*1)/4, or a 43.8% aggregate strikeout rate. I ignored what actually happened in those plate appearances; I only looked at each pitcher’s strikeout rate this year. I did this for every single team, in every single game, in both of my two scenarios. From there, I took the difference between the two. That seems like the most important number to me; a good tactical manager should, in general, lean towards higher-strikeout pitchers in the 10th than in tied games in the ninth. Why tied games? Because if you already have the lead, you should just put your best pitcher in. Playing for an advantage in extra innings only makes sense if the game is likely to go to extras. Why difference rather than absolute number? A manager can only work with the pitchers he has. Unsurprisingly, to me at least, Cleveland (and Francona) fare well by this metric. The rest of the list is pretty interesting, too: Zombie-Killers Team 9(tie) K% 10 K% Diff PIT 22.5% 24.6% 2.1% CLE 26.2% 27.8% 1.6% MIN 23.9% 25.0% 1.1% BAL 21.2% 22.1% 0.9% BOS 25.1% 26.0% 0.9% WSN 26.0% 26.8% 0.8% TBR 26.3% 27.0% 0.7% SDP 29.5% 30.1% 0.6% NYY 24.5% 24.9% 0.4% CIN 24.2% 24.6% 0.4% SEA 28.4% 28.5% 0.1% ARI 21.7% 21.8% 0.1% HOU 27.8% 27.8% 0.0% OAK 26.0% 25.7% -0.3% COL 24.4% 24.1% -0.3% LAD 29.8% 29.5% -0.3% SFG 24.9% 24.3% -0.6% TOR 26.1% 25.3% -0.8% CHW 26.4% 25.4% -1.0% TEX 24.5% 23.5% -1.0% MIA 24.7% 23.7% -1.0% DET 25.9% 24.5% -1.4% PHI 29.3% 27.7% -1.6% STL 27.6% 25.9% -1.7% CHC 28.3% 25.5% -2.8% NYM 28.8% 25.1% -3.7% LAA 27.7% 23.8% -3.9% KCR 25.7% 21.5% -4.2% MIL 34.0% 28.1% -5.9% ATL 33.4% 27.3% -6.1% Change in K% between 9th inning tied situations and 10th inning situations, 2022 The Pirates! That’s no fluke of data; Derek Shelton has used Wil Crowe and his 20.2% strikeout rate more than any other pitcher in tied ninth-inning spots. He’s used David Bednar and his 33.5% mark most frequently in the 10th. Is that on purpose? I can’t be sure. But it agrees with the math, which is the point of the exercise here. On the other side of things, the Braves go completely by the pre-2020 modern book of using your best available pitcher in the ninth inning. Kenley Jansen and A.J. Minter have accrued the lion’s share of tied batters faced in that situation. In the 10th, the team has turned to Tyler Matzek and Dylan Lee. Optimal strategy would suggest reversing the two, or at least making a more concerted effort to get some heavy-hitting pitchers into the game in extra innings. Jansen has only faced four batters in extra innings all year; Minter has only faced three. I want to be clear: this isn’t the only thing that makes managers good at handling extra innings. It isn’t even necessarily the most important thing. It’s just a place where a good process adds a tiny bit of value. Adding a tiny bit of value sounds exactly like what I’d like my managers to do during a game, especially when it’s late and close. I planned on hunting down other ways that managers could make good or bad choices relating to the zombie runner rule, but there just aren’t that many. There have been a grand total of 15 bunts leading off the top of an extra inning this year, which is a tactical no-no. Only three teams – the Nationals, Rangers, and Dodgers – have multiple such attempts. And even then, there are mitigating circumstances. Victor Robles has both Nationals bunts. Hanser Alberto is one of the two Dodgers to do it, and he bunted for a hit rather than sacrificing (Austin Barnes is the other). These mostly look like player decisions instead of team strategy. Likewise, only six players have led off the bottom of the 10th inning with a bunt while trailing. Austin Hedges was one of them, for example, and fine, I get it, that might be a good bunt despite the situation. “Did you call for a bad bunt in extras” seems like a bad check of managerial competence because every manager is good at avoiding them. What about good bunts? I’m not really sure what to think of these. The math argues that sacrificing the zombie runner to third base is a good play if you start the bottom of an extra inning tied. No one really does it, regardless of the math. The Orioles have four, the Mariners have three, and the rest of the league combined has 15. A tip of the cap to those two teams is in order, but it’s hard to know, with so few observations, whether it’s a plan set by the team or just noise. Rougned Odor and J.P. Crawford have combined for 18% of all these opportunistic extra-inning bunts. When two players are 18% of all your observations, it’s hard to learn much about any league-wide trend. What can we learn about how managers squeeze the most juice out of the time-saving extra inning rules? Not a ton, most likely. The edges are tiny, the chances to deploy optimal strategy infrequent. What’s more, my metric paints with a broad brush; strikeout rate is a blunt analytical tool, merely one that’s easy to use at scale. That said, I feel confident that Francona is a good tactical manager, and I’m certainly intrigued to see Shelton so high. Oh yeah – Francona’s move didn’t really work Tuesday night. After De Los Santos pitched a clean ninth, the White Sox sent out Kendall Graveman for the 10th. He was undone by the rules; the zombie runner scored, and so too did José Ramírez, who Graveman intentionally walked to set up a double play earlier in the inning. He might easily have thrown a scoreless inning if it weren’t for that pesky runner. Clase gave back any advantage the Guardians had, though. In a rare off night, he gave up a single, a stolen base, and another single to allow Chicago to score two runs and re-tie the game. The White Sox had their pitchers in the wrong order, the Guardians had theirs in the correct one, and the aggregate edge was so small that it got washed away quickly. The Guardians won the game after hanging a five spot on Jake Diekman the next inning, but that’s less about pitcher ordering and more about the quality of each team’s bullpen. Should managers squeeze every last drop out of their roster by holding their best pitcher for extra innings as appropriate? Sure. Does it matter much? Nope. When I calculated how much this pitcher ordering mattered on a per-game basis, the result was quite small; perhaps 3% of a win at the high end, even with a huge difference in pitcher strikeout rate between the two innings. The game is played by the players on the field. There’s nothing wrong with picking up those small edges, and I find it very satisfying to see managers doing so. Just don’t go thinking this is what makes any given manager good or bad.