The Adam Duvall Debate by Ben Clemens October 10, 2019 Earlier this week, I wrote about some dicey intentional walk decisions from the two NLDS series. One worked and one didn’t, but both of them were decisions managers made in huge leverage situations, and both managers tried to put their thumb on the scale of fate and influence the direction one way or another. Today, I’m going to talk about a decision that didn’t end up mattering nearly as much. Brian Snitker tried to seize a small edge in yesterday’s Cardinals/Braves game, starting Adam Duvall in the outfield over Matt Joyce despite the fact that Joyce is left-handed and Jack Flaherty is a righty. It looked like a weird decision, one that might swing the game, but of course it didn’t. Nothing Snitker could have done, short of conducting a seance to channel the life force of a young Warren Spahn into Mike Foltynewicz, could have mattered. The Cardinals scored 10 runs in the first inning, an endless deluge of offense, run stacked upon run. They added another in the second, and by the time Duvall stepped to the plate, our game win probability stood at 1.4% for the Braves, though the system doesn’t work well with such extreme margins. The leverage index of that at-bat was a measly 0.11, and it didn’t get better from there; his subsequent at-bats came with leverage of 0.02, 0.01, and 0.00. So Snitker tried to make a small decision, and the gods of baseball laughed in his face. That doesn’t mean we can’t analyze his decision though, that we can’t evaluate it on its merits regardless of the actual outcome. If this game was played a million times, the vast majority would be closer than yesterday, and in some of those the difference between Duvall and Joyce would decide the game. You’re reading this on FanGraphs, so it won’t surprise you to hear that I don’t like declining the platoon advantage. But rather than just proclaim this from on high, let’s go through the math. In his career, Duvall has 1436 plate appearances against right-handers and 525 against lefties. Against the lefties, he’s an above-average hitter; .240/.318/.473, good for a .333 wOBA and 104 wRC+. Against righties, he’s been a more pedestrian .230/.283/.457; worse across the board, which works out to a .312 wOBA and 90 wRC+. This platoon split is actually slightly larger than the major league average; over the past two years, righties have hit for a .325 wOBA against lefties, but they have only put up a .310 mark they don’t have the platoon advantage. Given that Duvall doesn’t have all that many plate appearances against lefties, we have to regress his splits back somewhat heavily; in all, I expect that his true-talent edge is 5.3% against lefties. When you apply that to his projected batting line, that makes him a .307 wOBA hitter against righties and a .323 wOBA hitter against lefties. That doesn’t look so crazy; it’s slightly worse than what he’s done in his career so far. How about Joyce? Well, in addition to being a better hitter than Duvall (Depths Charts projects him for a .333 wOBA as opposed to .311 for Duvall), he has a huge career platoon split, and lefties generally have larger splits than righties. Even after regressing his numbers back towards the mean, he projects as a .339 wOBA hitter against righties. In one sense, the decision is obvious. Would you rather bat someone with a .307 wOBA (think 2019 Freddy Galvis) or .339 wOBA (2019 J.T. Realmuto or Marcell Ozuna)? That’s not a hard question. The Braves, presumably, aren’t dummies, which means they know that. What then could cause them to make this decision? There are two possible explanations, or possibly a combination of the two. First, there’s the defensive upgrade from Joyce to Duvall. This isn’t meaningless, not by a long shot; Joyce is a below-average defender, while Duvall is significantly above average in an outfield corner. What is that worth in a single game? It’s high-variance, of course, but we can take a guess. Call Duvall a 6 UZR/150 defender and Joyce -4. This works out to 10 runs per 150 games, or .067 runs per game. That’s just a naive expectation, and we can do slightly better. Foltynewicz is pitching, and he’s a fly-ball pitcher; bump it up to .1 runs per game based on his fly ball rate. That’s probably more than it’s worth, but we’re painting with a broad brush here, and I’m erring on the side of giving the Braves credit. Next, we subtract .1 from our projection for how many runs the Braves would allow in a normal game with Foltynewicz pitching. Using Pythagenpat, the Braves are a .574 team with Folty going against a .500 opponent per our projections; with the extra defense, that works out to a .584 team. That would, in theory, bump their odds of beating the Cardinals up from 54.8% to 55.7% per my model. Hey, nearly 1% of win expectancy from defense; not bad. Next, we have to work out the cost of the missing offense. That’s not quite as easy, because wOBA doesn’t combine additively into run scoring. To make an approximation, I calculated the team’s BaseRuns against an average right-handed pitcher, with both Duvall and Joyce in the lineup. Even this method isn’t perfect, and to get it right you’d really want to simulate a bunch of games with both of them to work out lineup effects, but buddy, we’re analyzing a 13-1 game here, cut me some slack. In any case, this works out to about .13 runs per game of sacrificed offense from playing Duvall over Joyce. The back-of-the-envelope wOBA math (wOBA difference divided by wOBA scale times plate appearances) comes out to .11, which suggests we’re in the right ballpark. If you use this .13 run answer, the team’s winning percentage with Duvall’s offense and defense taken into consideration falls to 54.4%, .4% less than if they’d played Joyce. Not bad, but not a good decision. So the defense makes it closer than it looks, but doesn’t flip the direction of the decision, even after using friendly assumptions. What if we’re missing something, though? What if Duvall is seeing the ball well; if he’s just so hot that he’s worth playing over Joyce? In a word: c’mon. In slightly more words though, this seems unlikely to be true. At surface level, sure: before yesterday’s game, Duvall had a .535 wOBA in the playoffs, Joyce .157. How about those sample sizes? The smallest imaginable! Duvall had come to the plate eight times (one an intentional walk), Joyce 10. Instead, let’s use their last 100 non-IBB PA for each; an arbitrary number, to be sure, but we’re far into arbitrary land here. Over those slices, Joyce had a .400 wOBA, Duvall .314. Wait, which one is the hot hitter?? If it’s not clear, I’m not a fan of playing Duvall over Joyce due to some imagined hot streak. There’s one option the Braves didn’t explore, however. What if the team instead benched Nick Markakis and played Duvall and Joyce together? Markakis isn’t as good of a hitter against righties as Joyce; he comes in at only a .333 wOBA projection. You can convince yourself he’s just as bad of a defender as Joyce — his DRS, UZR, and OAA all look quite similar this year. In our model from above, this gets the Braves up to almost exactly break even in win percentage; 54.7% to win the game with Duvall and Joyce starting in the corners. It also has the benefit of enriching the bench; the Braves don’t have a deep bench, but they’re particularly short of left-handed pinch-hitters, and being able to guarantee a platoon advantage in a key at-bat is a nice tiebreaker. In the end, I would have started Duvall and Joyce. Adam Duvall, by the way, went 0-3 with three strikeouts against Jack Flaherty, not that it mattered. Matt Joyce pinch-hit and grounded out against Giovanny Gallegos. Neither recorded a single putout in the outfield. Sometimes you scheme and wonder about small edges and then you lose by 12 runs and nothing matters.