Josh Hader Shouldn’t Have Pitched on Tuesday

Tuesday afternoon, Josh Hader took the mound at the start of the ninth inning. His Houston Astros were losing 3-0 to the Detroit Tigers in the first game of their three-game Wild Card series. Riley Greene smashed a one-hopper over the right field fence for a double, but Hader retired the other three batters he faced and departed with the three-run deficit still intact.

This was strange! That’s not how teams use their closers. It felt weird right away – to the broadcasters calling the game, to the chatters who flooded us with questions about it, and also to me. And it felt consequential the next day, too, when Hader was summoned for his usual job. This time, the Astros were tied, and there were runners on first and second with two away in the bottom of the eighth. It was the biggest spot in the playoffs for the Astros. Hader walked Spencer Torkelson on four pitches, then threw a fastball right down the middle that Andy Ibáñez tattooed for a bases-clearing double. That made it 5-2 Tigers, and just like that, Houston’s season was over.

If you want to, there’s an easy through-line to trace here. Hader made a low-leverage appearance, and then he had to pitch again on no rest. He didn’t have his best stuff in that second game, so he paid the price. Cause and effect, simple as that.

I don’t think it’s quite so simple, though. We’ll never know what would have happened if Hader hadn’t pitched in the first game. Maybe he would have been perfect in Game 2 and helped the Astros escape with the score tied. Maybe they would have added a run in the ninth to win the game and send things to a decisive Game 3 the next day. Or maybe he would have grooved the same fastball. Maybe he would have bounced a slider to allow a run to score on a wild pitch. The world isn’t deterministic.

That said, I found Hader’s Game 1 appearance strange, so I thought I’d try to do the math on it. Did it make sense? What did it cost the Astros in expectation to use him there, and what did they stand to gain? That’s the real question to answer, so let’s give it a shot.

I’m going to paint with a broad brush in trying to figure out what the Astros gained by sending Hader out in Tuesday’s game. They trailed by three heading into the top of the ninth. Home teams win about 3.1% of the time in that situation, according to our formulation of WPA. It’s 3% according to Baseball Reference’s accounting (they don’t use decimal points), and 3.3% if you use our WPA Inquirer. When the top half of the inning ended, the Astros were about 3.8% to win (3.8% in WPA, 4% in B-Ref’s WPA, 4.1% in the Inquirer). That’s a tiny difference. In other words, Hader’s appearance was extremely low leverage.

Here’s the kindest way of thinking about the benefits of using Hader there: Let’s assume that he was 100% certain to escape the inning without allowing a run. Let’s further assume that whichever pitcher the Astros would have used instead of him was 100% likely to surrender a run. That’s a wild assumption – no one in baseball gives up a run every time out – but I’m trying to estimate the largest amount of win probability you could credit Hader with for entering in that spot. Per our Inquirer, teams win 4.1% of the time when they enter the bottom of the ninth down three runs, and 1.8% of the time when they enter the bottom of the ninth down four runs. So Hader’s Game 1 appearance was worth no more than 2.3% of a win.

That’s the upper bound. Let’s make a more reasonable assumption next: The pitcher the Astros would have used instead of Hader was merely 70% likely to escape unscathed. That’s a below-average result, and one that tracks with reality. Caleb Ferguson, who Hader replaced, put up a zero in 71% of his appearances this year. That’s true of Héctor Neris, who had pitched earlier in the game, as well. Spencer Arrighetti was the likely alternative, but he was mostly a starter this season, so I don’t have a good split for him. Either way, 70% is a roughly league average level. Let’s say Hader will go scoreless 85% of the time. That puts the value of having Hader in instead of a replacement level reliever at 15% of the difference between being down four runs and being down three runs. That works out to 0.35% of a win added by pitching Hader in this spot. Let’s double it to be generous – pitching Hader increased the Astros’ chances of winning this particular game by 0.7%.

That’s the easy part. The next question is what that appearance cost them. First, let’s work out the chances that Hader would be asked to pitch again the next day. They’re not exactly 100%, because if the Astros went up big early, manager Joe Espada would likely have avoided using him to set up for the winner-take-all finale. This year, 60% of Houston’s games ended with a margin of victory of three runs or fewer. I think the odds of Hader coming in are actually a bit higher than that, for a few reasons.

First, that’s the score at the end of the game – some games that ended with lopsided scores were closer than that in the late innings. Second, we’re talking about an elimination game. When winning is the only option to keep your season going, bullpen usage changes. Closers pitch more often, even if the spot wouldn’t strictly call for them. There’s no margin for error, so teams use their best relievers more frequently. It makes less sense to sacrifice from the present to help out the future if losing in the present ends the season.

When Espada called for Hader in the ninth inning of the first game, I think that there was about a 75% chance of him pitching in Game 2. But that doesn’t answer the question of what his Game 1 appearance cost. To figure that out, we’ll need to estimate how much Hader’s effectiveness declines when he pitches back-to-back days. Take the change in effectiveness, multiply it by the chances of him getting used, account for the leverage of the situation where he’d be used, and then we’ll have our answer.

In 2024, Hader pitched on consecutive days 17 times. He posted a 3.94 ERA on zero days’ rest and a 3.76 mark the rest of the time. But that’s hardly conclusive. He had a 2.35 FIP on no rest, much better than his overall mark. And the sample sizes are tiny – 17 appearances, 16 innings. That’s no way to judge a pitcher’s true talent level.

To make the sample size work, I instead looked at Hader’s last four years of results. There, the picture is clearer. He’s roughly 0.8 runs per nine innings worse on one day’s rest, whether you’re looking at ERA or FIP. That result holds true whether you break each year out individually and deal in percentage change or sum everything up together. His strikeout rate is roughly five percentage points lower on zero days of rest; his walk rate is nearly four percentage points higher. He’s just worse, and that’s quite reasonable.

What did that worse-ness project to cost the Astros? I tried to approximate it in two ways. First, I did a back of the envelope check. Over one inning, 0.8 runs of ERA is about 0.09 runs. Hader’s average leverage index this year was 1.6, which means events have been about 1.6 times as meaningful to the outcome, relative to average, when he’s pitching. Multiplying 0.09 by 1.6 works out to 0.144 runs. There are 9.7 runs to a win this year, and 0.144 divided by 9.7 is 1.5%. So you can think of Hader’s decreased effectiveness on zero days’ rest as costing the team 1.5% of a win, on average, when he pitches. Couple that with our 75% estimate of how likely he was to pitch back-to-back, and you get a 1.1% decline as the expected cost, in Game 2 win probability, from having Hader on zero rest.

That’s a pretty abstract way of looking at things, so let’s calculate it another way as a sanity check. I took a situation that matches Hader’s average entry leverage index this year – up one run in starting the top of the seventh. I created a runs allowed distribution for him on regular rest, and another for him on zero rest. Those look like so:

Then I simply applied those outcomes, checked the resulting win probability, and multiplied the whole mess together. Let’s look at the regular rest distribution for an example. Hader doesn’t allow any runs 78% of the time, and the game goes to the bottom of the seventh with the Astros still up by one. They win 79.1% of the time in that scenario, per the WPA Inquirer. Sixteen percent of the time, he allows one run, and the Astros are tied in the bottom of the seventh (58.8% win probability). Four percent of the time, he allows two runs (36% win probability). Two percent of the time, he allows three or more — let’s call it three exactly, just for ease of calculation, and give Houston a 21.5% chance at a win.

Multiply those win probabilities by the chance of each outcome, add everything up, and you’ll get a 73% chance of a win. Do the same thing with Hader’s distribution on zero days of rest, and you get a 71.1% chance of winning. That’s a 1.9% decline in win probability, with 75% likelihood from up above, or a 1.4 percentage point decrease in win probability. Our estimate up above was 1.1%; that’s pretty good agreement between the two methods.

I tried a few other reasonable spots for Hader to enter the game, just to get an idea for how this could play out. Up a run heading into the ninth, tied heading into the ninth, down a run with one out in the eighth – I ran a variety of scenarios using this method. They produced results between 0.7% and 1.8%, after accounting for the 75% likelihood of him getting in the game. Sure, none of these replicate the exact situation where he came in, but I’m trying to capture what we should have expected when Hader entered the first game, not specifically what happened. Either way, a tired Hader is more likely to allow runs, and regardless of when exactly the Astros called him into Wednesday’s game, they were likely to need him in some high-leverage spot. These are all just rough numbers, but I think the idea is clear: Across a broad variety of situations, using Hader in back-to-back games hurts Houston’s odds of winning the second game by somewhere around 1-1.5 percentage points.

There’s just no chance that using him in the first game was that valuable. Teams almsot never win when they’re down three heading into the top of the ninth. Bringing in the ghost of Cy Young wouldn’t even be that helpful there; the salient fact is that you’re starting out down three, not how many further runs you allow.

It gets worse. Hader being ineffective on the second day of a back-to-back is bad enough, but even if the Astros had won on Wednesday afternoon, they would have needed to play another game against Detroit the next day. In that game, they’d probably want to use their best pitchers. After all, it would be winner take all, and they’d be using their worst starting pitching option after Framber Valdez and Hunter Brown started the first two games.

I don’t have enough data to say with any certainty how well Hader handles pitching on three straight days, because he’s almost never done it. That said, I feel comfortable saying that it wouldn’t be better than his results pitching on two straight days. Heck, he might not even be available, which would be a real disaster.

That’s just a “bonus” cost of using Hader for low-leverage outs in the first game of the series. The fact of the matter is, it was already a bad decision without taking that into account. The Astros hardly picked up anything by using him in a near-hopeless spot. They paid a high expected cost for it in Game 2, double or more the slim advantage they gained in Game 1.

Managers mostly don’t determine the outcomes of playoff series. The players on the field do the run scoring or preventing, and managers can only subtly put their thumbs on the matchup scales. Errors like this still aren’t huge; if this cost the Astros 1% of a win, a reasonable estimate, plenty of other things were a bigger deal. But this was a free 1%, a 1% that they could have picked up by simply not doing something silly.

Just to reiterate, this isn’t why Houston lost. A percentage point is tiny in the grand scheme of things. Kyle Tucker went 0-3 and left some runners on base, producing -.146 win probability added – his having an off day was about 15 times as consequential as Espada’s pitching usage. But a player having a bad day at the plate is inevitable and mostly unpredictable. A manager using his pitchers the wrong way isn’t. Bringing in Hader down three in Tuesday’s game didn’t swing the series – but as far as I’m concerned, it was clearly the wrong decision.

Source