pointzerofive.

· Sports data · 116 coach changes · 2012/13–2025/26

Does Firing the Coach Work?

Bundesliga clubs that change coach mid-season usually improve afterwards. So do clubs in the same slump that keep their coach. We compared 116 coach changes since 2012 with those teams.

After a change 0.68 → 1.28 points per game, 6 games each side
Same slump, coach stayed 0.68 → 1.21 matched on points before
Difference +0.07 95% CI −0.05 to +0.18

Abstract

We compare 91 mid-season coach changes in the Bundesliga (2012/13–2025/26) with six-game stretches in which teams with the same recent results kept their coach. In the six games before a change, teams averaged 0.68 points per game; in the six after, 1.28. Matched teams that kept their coach rose to 1.21. The estimated effect of a change is +0.07 points per game (95% cluster-bootstrap CI −0.05 to +0.18). Estimates under alternative windows and matching rules range from +0.03 to +0.12, all with intervals that include zero. The short-run improvement after a coach change is mostly regression to the mean.

1 Data

Results and the manager of record for all 4,284 Bundesliga matches, 2012/13–2025/26, come from transfermarkt-datasets (CC0 1.0; data from Transfermarkt.de) and were checked against openfootball (CC0 1.0); both sources agree on every match date. A coach change is the first league game under a different manager within a season. Stand-ins, where the previous manager returns later in the season, are excluded (10 cases); a caretaker followed by a permanent coach counts as one change. Of 116 changes, 91 have six games on each side within the season.

2 Results

All coach changes: points per game0123coach changebefore: 0.681.28last game before: 0.15−6−5−4−3−2−1+1+2+3+4+5+6Games before / after the changePoints per game Coach changes vs. same slump, same coachcoach changedsame slump, coach stayed0123coach change1.21before: 0.681.28difference: +0.07 points per game95% CI −0.05 to +0.18−6−5−4−3−2−1+1+2+3+4+5+6Games before / after the changePoints per game
Figure 1. Mean points per game in the six games before and after a coach change (n = 91). Left: changes only; the last game before the change is highlighted (0.15 points on average). Right: with matched controls (same points in the previous six games, no coach change in the twelve-game window). Bars mark six-game averages.

Results improve sharply after a change: 80% of teams took more points in the six games after than in the six before. Matched teams that kept their coach improved almost as often (76%) and almost as much.

Previous 6 games vs. next 6, with coach changes in red0123form continuesleague average 1.380 points in previous 6: 0.93 per game next (n = 19)1 points in previous 6: 1.16 per game next (n = 58)2 points in previous 6: 1.06 per game next (n = 81)3 points in previous 6: 1.17 per game next (n = 133)4 points in previous 6: 1.25 per game next (n = 263)5 points in previous 6: 1.28 per game next (n = 349)6 points in previous 6: 1.23 per game next (n = 367)7 points in previous 6: 1.30 per game next (n = 563)8 points in previous 6: 1.33 per game next (n = 502)9 points in previous 6: 1.34 per game next (n = 434)10 points in previous 6: 1.42 per game next (n = 521)11 points in previous 6: 1.58 per game next (n = 345)12 points in previous 6: 1.60 per game next (n = 282)13 points in previous 6: 1.69 per game next (n = 315)14 points in previous 6: 1.93 per game next (n = 147)15 points in previous 6: 1.87 per game next (n = 153)16 points in previous 6: 2.05 per game next (n = 165)18 points in previous 6: 2.23 per game next (n = 82)0369121518Points in the previous 6 gamesPoints per game, next 6 games
Figure 2. Points per game in the next six games against points in the previous six, for all 4,779 six-game stretches without a coach change (grey; dot area ∝ number of stretches) and for each coach change (red, jittered). Dashed: previous form continues. Solid: league average (1.38). Stretches with ≤ 4 points went on to 1.18 points per game.

3 Robustness

Effect of a coach change, points per game4-game windows (n = 103)+0.046-game windows (n = 91)+0.078-game windows (n = 74)+0.1210-game windows (n = 55)+0.04matched on season form too (n = 91)+0.12only after ≤ 4 points in 6 (n = 59)+0.03−0.20−0.100.00+0.10+0.20+0.300 = no effect
Figure 3. Estimated effect of a coach change in points per game with 95% cluster-bootstrap intervals (changes resampled; controls resampled by club-season; 2,000 replicates for the main specification, 1,000 for the others). Main specification in red.

Opponent strength does not differ: opponents after a change took 1.37 points per game over their season, against 1.38 for matched controls. Restricting to changes after ≤ 4 points in six games (n = 59) gives 0.42 → 1.18 points per game, an effect of +0.03.

4 Discussion

The result agrees with Heuer et al. (PLoS ONE 6(3), e17664, 2011) for the Bundesliga and ter Weel (De Economist 159, 279–303, 2011) for the Dutch league. Limitations: the data do not distinguish dismissals from resignations; clubs choose the timing of a change and may act on information not reflected in recent results; the analysis covers six games after the change, not longer-run effects; changes in the first or last six games of a season are excluded.

Data availability

Dataset: data/does-firing-the-coach-work.json; raw inputs: notebooks/raw/bundesliga/; pipelines: notebooks/bundesliga_matches_pipeline.py and notebooks/coach_firing_pipeline.py in the story repository. A notebook that recomputes every statistic opens in Google Colab via the "Notebook" link above.

Data source: transfermarkt-datasets (games, CC0 1.0; data from Transfermarkt.de); openfootball/deutschland (CC0 1.0) · Last updated: 2026-10-14