3,745 changes later: the heroes Valve cannot balance
We cross-referenced the full patch history since November 2024 with today's win rates. The number of changes a hero received predicts nothing at all — and two heroes prove it, each in their own way.
We took every change we have on record since the beta began — 3,745 modifications across 27 patches, from November 2024 to August 2026 — and cross-referenced them with each hero's win rate today. The idea was simple: find out who Valve has worked on the most, and whether that work paid off.
The answer is unsettling.
The most-changed podium
McGinnis leads with 109 changes: 31 buffs, 43 nerfs, 6 reworks, spread over 13 different patches. Behind him, Abrams and Mirage are tied at 98 changes each.
The consistency record goes to Shiv, who was touched in 15 separate patches — more than one patch in two since he arrived. Viscous follows with 14.
Mirage: two years of work, still last
This is the starkest case in the game. Mirage has received 98 changes, including 12 full ability reworks — the second highest total in the game. Valve buffed him 35 times, nerfed him 34 times, and rebuilt him a dozen times.
The result today: a 44.43% win rate. The worst hero in the game.
No other character has received that much attention for such a poor outcome. Not for lack of trying — for lack of finding the answer.
McGinnis: the opposite failure
The most-changed hero of the beta illustrates the mirror problem. Of his 109 changes, 43 are nerfs against only 31 buffs — Valve has never stopped trying to slow him down.
He still sits at 53.95%, the fifth best hero in the game. Two years of braking, and he never really slowed.
Ivy tells the same story more quietly: 35 nerfs against 21 buffs, the third harshest ratio in the game, and a comfortable 53.72% at the finish.
Changing a hero does not move them toward balance
This is the clearest finding of the analysis. We measured the statistical link between the number of changes received and the distance from a 50% win rate. The correlation is zero: 0.018, which is to say nothing at all.
Worse, the trend leans slightly the wrong way. The ten most-changed heroes sit an average of 2.74 points away from perfect balance. The ten least-changed, 2.21 points. The ones left alone are doing marginally better than the ones worked on constantly.
That number needs reading properly: changes obviously do not cause imbalance. It is that Valve concentrates its attention on the heroes that are a problem — and that attention is not enough to bring them back in line. The work is reactive, not curative.
What does work
Two successes deserve a mention, because they show the exercise is not hopeless.
Haze received 83 changes and now sits at 50.16% — a sixth of a point from perfect balance. Abrams, with 98 changes including 41 buffs against 18 nerfs, lands at 50.61%. In both cases, long sustained work eventually produced exactly the intended result.
Holliday is proportionally the most nerfed hero in the game — 16 nerfs against 7 buffs — and ended up at 48.65%. Again, intent and outcome line up.
Paradox, king of reworks
With 14 ability reworks, Paradox is the hero Valve has most often rebuilt rather than simply adjusted. That is an admission in itself: reworking something fourteen times means it is not doing what it was meant to.
His current win rate: 46.77%. The fourteenth attempt fared no better than the previous ones.
And then there is Sinclair
At the far end of the spectrum, one hero has simply been forgotten. Sinclair has received only 8 changes in total, across 3 patches, and nothing at all since November 21, 2025 — nearly nine months of silence.
That would be fine if he were doing well. He sits at 45.08%, the fourth worst win rate in the game. A weak hero nobody appears to have reopened since last autumn.
What it says about the game
Deadlock is in beta, and a beta gets fixed. The sheer volume of work is genuinely impressive: nearly 3,750 changes in under two years, across 39 heroes.
But the analysis shows that volume is no guarantee. Some characters stubbornly resist everything thrown at them — Mirage downward, McGinnis upward — while another stays stuck below average simply for never being reopened.
If you are looking for a stable hero to sink your hours into, the change history is not a good indicator. The only one that counts is the current win rate, and it moves with every patch.
Methodology. The 3,745 changes come from our own database, built by parsing Deadlock's official patch notes since November 2024, and cover the 27 patches for which detailed content is available. Each change is classified as a buff, nerf, rework, bugfix or neutral adjustment. Win rates cover matches played between July 30 and August 7, 2026, all modes combined, collected through the public deadlock-api.com API. A hero present in our database but absent from live game statistics was excluded from the analysis.