Maybe The New-Manager Bounce Is Real After All
Firing the skipper is often a convenient scapegoat move — but the data says it might help more than simple regression would predict.
Regular readers of the Substack may be able to detect that I’m something of a head-coach (and by extension, manager) apologist. I tend to think teams should not change leadership lightly, and too often it feels like these decisions are made on an owner’s whim or — increasingly in both basketball and hockey — based on a late-season sense of panic.
Of course, I also once wrote that only a few managers in MLB history actually have a demonstrable effect on their players’ performance, which might seem at odds with the stance above. If it’s all just re-arranging deck chairs on the Titanic, why not take a shot at something different on the off-chance it works?
Well, there is evidence (from the NFL) that teams who make a bunch of coaching changes in quick succession tend to see their performance in subsequent seasons decline even after controlling for their previous quality.1 And when it comes to in-season moves, I’ve long suspected any perceived bounce a team receives immediately after a change — like the Phillies have gotten after dropping Rob Thomson for Don Mattingly — can mostly be chalked up to regression to the mean, which would have happened anyway under the old guy if given the chance.
But how true is that, actually? A quick look at the recent history of teams who swapped managers around midseason offers at least some mild evidence that these kinds of changes genuinely do help a team beyond simple regression.
To test this, I looked at all teams in full, 162-game seasons since the 1994 strike who changed managers before their 100th game of the year. (So 1995, 2020 and 2026 were ineligible.) I also tossed out cases where a third manager entered the fray, just for the sake of keeping things simple. Establishing the relationship between Elo ratings and future wins,2 I judged winning percentages in the month immediately following the switch relative to what we’d expect from a team with the same Elo in that particular moment.
In this (admittedly somewhat limited) sample of teams who changed managers, they ended up performing better — with a .472 winning percentage — over the month after the swap than we’d expect of a team in that particular Elo situation (.466). Is that a particularly large effect? No — and we can’t even attribute it completely to the coaching change, either, since other shakeups probably coincided with the decision to switch up skippers.
Still, it shows that improvement after a managerial change tends to go somewhat beyond what we’d expect from mere regression to the mean. And if we thought these upticks were short-term bursts in the wake of a fresh voice, there’s actually greater evidence that a manager change carries over through the rest of the year. If we perform the same analysis as above, but instead of predicting the month immediately after the change, we predict the remainder of the regular season after that first month, our sample of teams collectively posted a winning percentage of .479 during that span — compared with a .458 expected winning percentage following that first month, based on Elo:
(Why would the team’s expected winning percentage be lower following the first month? It could be a residue of bad teams’ tendency to trade away good players at the deadline in late July — though that’s just a guess.)
Either way, this is an even stronger effect to demonstrate that improvement from teams like the Mattingly Phillies and Chad Tracy Red Sox is, on some level at least, fairly normal. Teams tend to do better — if not exactly 7-1 better — after a midseason manager change, even after accounting for how teams with their Elo would be expected to improve over either the short or long run.
That being said, there are limits to all of this. While Elo attempts to be the best unbiased estimator of a team’s talent at any given moment, their rating at the time of a managerial change might catch them at one of their lowest-rated moments in a given season. If we set our expectations according to the team’s preseason Elo rather than its rating at the time its original manager was fired, any surplus winning percentage is erased:
I still think the former baseline — Elo at the time of the change — is the right standard to which an in-season replacement should be held. But this mainly serves to illustrate that while, on average, changing managers may be able to beat regression-to-the-mean, they can only bring you a fraction of the way back toward however good you were originally supposed to be before the year.
In other words, a manager change can help more than I was perhaps giving credit for, even if it’s seldom an outright magic-bullet fix. Yes, any gains relative to regression-to-the-mean are usually quite modest, and most of the time, it’s simply designed to give the front office a convenient way to scapegoat the easiest person to remove. But we can’t say a move is never worth it — especially if the clubhouse has truly gone stale, or if the alternative is watching a talented team drift aimlessly into oblivion, like we saw in Philly.
Filed under: Baseball
This, in theory, accounts for the selection bias of bad teams being much more likely to fire their coach (unless you’re a panicking NBA GM in the final days of the 2025 regular season, for whatever reason).
Specifically, I set up the expectations based on a team’s Elo after the 60th game of the season, since the average managerial change in our sample happened after a team had played 57.6 games.