Patches · released Jul 28, 2026
06-30-2026 Update
18 hero win rates moved around this patch by more than the noise floor, out of 38 tested. These cleared a Bonferroni-corrected threshold — but read the caveat: this is what moved around the patch, not proof the patch caused it.
Read this first
This compares hero win rates after the patch against before it. A shift here is correlational: the meta drifts on its own, other things change on the same day, and win rate reflects who's playing a hero as much as the hero itself. We clamp both windows so they never cross a neighbouring patch, but when patches land close together the windows get short and the sample thin — noted below.
Before window
7.0d
3,576,651 matches
After window
7.0d
4,921,178 matches
Heroes tested
38
≥500 matches each side
Threshold
1.3e-3
0.05 / 38
Heroes that really moved
Reading the numbers: pp = percentage points (52% → 54% is a 2-point rise). p = how likely the move is just luck — lower means we're surer it's real.
53.7% → 50.3% win rate · 171,642 matches after
Haze's win rate fell 3.41 points across this patch. A swing this large over 171,642 matches is unlikely to be chance (p < 0.000001), but whether the patch caused it is a separate question this data can't answer.
47.8% → 44.2% win rate · 77,623 matches after
Mirage's win rate fell 3.62 points across this patch. A swing this large over 77,623 matches is unlikely to be chance (p < 0.000001), but whether the patch caused it is a separate question this data can't answer.
49.1% → 45.8% win rate · 74,866 matches after
The Doorman's win rate fell 3.30 points across this patch. A swing this large over 74,866 matches is unlikely to be chance (p < 0.000001), but whether the patch caused it is a separate question this data can't answer.
48.8% → 50.7% win rate · 154,408 matches after
Celeste's win rate rose 1.86 points across this patch. A swing this large over 154,408 matches is unlikely to be chance (p < 0.000001), but whether the patch caused it is a separate question this data can't answer.
51.6% → 49.7% win rate · 136,145 matches after
Lady Geist's win rate fell 1.86 points across this patch. A swing this large over 136,145 matches is unlikely to be chance (p < 0.000001), but whether the patch caused it is a separate question this data can't answer.
50.4% → 52.1% win rate · 136,992 matches after
Vindicta's win rate rose 1.71 points across this patch. A swing this large over 136,992 matches is unlikely to be chance (p < 0.000001), but whether the patch caused it is a separate question this data can't answer.
50.1% → 48.6% win rate · 111,363 matches after
Yamato's win rate fell 1.54 points across this patch. A swing this large over 111,363 matches is unlikely to be chance (p < 0.000001), but whether the patch caused it is a separate question this data can't answer.
54.2% → 55.5% win rate · 165,434 matches after
Seven's win rate rose 1.29 points across this patch. A swing this large over 165,434 matches is unlikely to be chance (p < 0.000001), but whether the patch caused it is a separate question this data can't answer.
48.5% → 49.7% win rate · 192,684 matches after
Infernus's win rate rose 1.13 points across this patch. A swing this large over 192,684 matches is unlikely to be chance (p < 0.000001), but whether the patch caused it is a separate question this data can't answer.
46.0% → 44.5% win rate · 104,311 matches after
Silver's win rate fell 1.48 points across this patch. A swing this large over 104,311 matches is unlikely to be chance (p < 0.000001), but whether the patch caused it is a separate question this data can't answer.
49.1% → 50.3% win rate · 110,430 matches after
Apollo's win rate rose 1.13 points across this patch. A swing this large over 110,430 matches is unlikely to be chance (p = 2.1e-6), but whether the patch caused it is a separate question this data can't answer.
50.9% → 51.7% win rate · 211,486 matches after
Drifter's win rate rose 0.75 points across this patch. A swing this large over 211,486 matches is unlikely to be chance (p = 5.9e-6), but whether the patch caused it is a separate question this data can't answer.
49.8% → 50.7% win rate · 166,657 matches after
Wraith's win rate rose 0.87 points across this patch. A swing this large over 166,657 matches is unlikely to be chance (p = 7.0e-6), but whether the patch caused it is a separate question this data can't answer.
50.4% → 49.5% win rate · 168,774 matches after
Billy's win rate fell 0.83 points across this patch. A swing this large over 168,774 matches is unlikely to be chance (p = 9.0e-6), but whether the patch caused it is a separate question this data can't answer.
46.7% → 47.8% win rate · 70,143 matches after
Grey Talon's win rate rose 1.09 points across this patch. A swing this large over 70,143 matches is unlikely to be chance (p = 2.9e-4), but whether the patch caused it is a separate question this data can't answer.
44.5% → 45.2% win rate · 149,799 matches after
Mina's win rate rose 0.70 points across this patch. A swing this large over 149,799 matches is unlikely to be chance (p = 6.4e-4), but whether the patch caused it is a separate question this data can't answer.
47.8% → 48.7% win rate · 96,781 matches after
Holliday's win rate rose 0.86 points across this patch. A swing this large over 96,781 matches is unlikely to be chance (p = 7.3e-4), but whether the patch caused it is a separate question this data can't answer.
46.9% → 47.6% win rate · 100,842 matches after
Viscous's win rate rose 0.76 points across this patch. A swing this large over 100,842 matches is unlikely to be chance (p = 0.0013), but whether the patch caused it is a separate question this data can't answer.
Aftermath: did it hold?
The same heroes, re-tested 2, 7 and 14 days after the patch. Each window runs from release to that day (cut short by the next patch or by today) and is compared with the same before-patch baseline. Windows that haven't elapsed say so instead of guessing.
- Haze
2d after
−3.35ppholds up · 42,028
7d after
−3.41ppholds up · 171,642
14d after
−3.28ppholds up · 322,452
- Mirage
2d after
−4.23ppholds up · 17,203
7d after
−3.62ppholds up · 77,623
14d after
−3.36ppholds up · 147,592
- The Doorman
2d after
−3.41ppholds up · 17,746
7d after
−3.30ppholds up · 74,866
14d after
−3.22ppholds up · 144,803
- Celeste
2d after
+2.08ppholds up · 36,679
7d after
+1.86ppholds up · 154,408
14d after
+1.84ppholds up · 296,081
- Lady Geist
2d after
−2.66ppholds up · 30,601
7d after
−1.86ppholds up · 136,145
14d after
−1.94ppholds up · 248,595
- Vindicta
2d after
+1.81ppholds up · 32,354
7d after
+1.71ppholds up · 136,992
14d after
+1.52ppholds up · 263,843
- Yamato
2d after
−1.36ppholds up · 24,170
7d after
−1.54ppholds up · 111,363
14d after
−1.74ppholds up · 216,589
- Seven
2d after
+1.27ppholds up · 34,430
7d after
+1.29ppholds up · 165,434
14d after
+1.00ppholds up · 306,531
- Infernus
2d after
+1.41ppholds up · 41,985
7d after
+1.13ppholds up · 192,684
14d after
+1.01ppholds up · 369,173
- Silver
2d after
−1.51ppholds up · 24,151
7d after
−1.48ppholds up · 104,311
14d after
−1.15ppholds up · 193,710
- Apollo
2d after
+0.90ppinside noise · 25,907
7d after
+1.13ppholds up · 110,430
14d after
+1.03ppholds up · 205,440
- Drifter
2d after
+0.71ppinside noise · 50,776
7d after
+0.75ppholds up · 211,486
14d after
+0.67ppholds up · 401,886
- Wraith
2d after
+0.94ppinside noise · 35,745
7d after
+0.87ppholds up · 166,657
14d after
+1.03ppholds up · 324,448
- Billy
2d after
−1.19ppholds up · 36,214
7d after
−0.83ppholds up · 168,774
14d after
−0.94ppholds up · 328,590
- Grey Talon
2d after
+1.46ppinside noise · 15,211
7d after
+1.09ppholds up · 70,143
14d after
+0.92ppholds up · 132,877
- Mina
2d after
+0.86ppinside noise · 33,146
7d after
+0.70ppholds up · 149,799
14d after
+0.65ppholds up · 284,804
- Holliday
2d after
+0.46ppinside noise · 21,212
7d after
+0.86ppholds up · 96,781
14d after
+0.76ppholds up · 187,512
- Viscous
2d after
+0.51ppinside noise · 23,969
7d after
+0.76ppholds up · 100,842
14d after
+0.96ppholds up · 192,713
Tested, moved, but inside the noise
20 heroes had a win-rate delta that didn't clear the bar. Shown by strongest evidence first — a reminder that a headline number and a real shift aren't the same thing.
Windows pivot on the release date (when the changelog thread went live). Two-proportion z-test per hero, Bonferroni-corrected across the 38 tested, same method as the weekly digest. Heroes under 500 matches in either window are excluded. Data from deadlock-api.com. Read the patch notes →