Damage simulator · beta
When we tested existing simulators against real fights, their numbers didn't match what players actually did in game. So we started building our own, from real fight data, and we keep updating it to help players get numbers they can trust. The simulator is a tool for you. It never grades or judges anyone.
Load a character, then change the stats to see what new gear would do. Testing trinkets and cast sequences on this page comes next. Only specs we have real recorded runs for are offered, and every profile shows how close it gets on fights it wasn't built from.
Leave one out: the profile is rebuilt without one run, then predicts that run. Split halves: built from one half of a run, it predicts the other half. Gear change: the character swapped gear and did another run; "unseen stats" means the prediction was made before the run, at stats the profile was never fitted on. Hit size is the same prediction using the run's real number of hits, so it tests only how hard each hit lands. The rest of the error is how often things hit, which varies from run to run with procs and cooldown timing.
Where it's accurate today: your own character, from one recorded run of yours. We change one thing (a piece of gear, or the cast sequence), predict the result first, then play it and compare. These were measured on our own characters, on training dummies:
| Spec | What changed | Predicted vs real |
|---|---|---|
| Shadow Priest | Gear, single target | within 1.1% |
| Unholy Death Knight | Gear, several targets | within 0.8% |
| Subtlety Rogue | Gear, single target | within 2.8% |
| Havoc Demon Hunter | Cast sequence | within 1–3% |
| Windwalker Monk | Gear (size of each hit) | within 1.5% |
| Windwalker Monk | Cast sequence | within 1.7–2.7% |
Where it isn't yet: real fights with moving bosses and many targets aren't measured to that level yet, and neither is a character the simulator has never seen, in a spec with little data. There, it currently misses the size of each hit by about 23% on average. It learns each spec from real fights, so a spec that few players have shared stays rough until more of them do.
Every shared fight teaches it more about that spec. Install the VoidScout addon and the uploader, and tick "share anonymous fight details for the damage sim". The more players of a spec share, the closer that spec gets to the accuracy above.
Before each Mythic+ key starts, the simulator predicts every player in the group, using only other players' data. After the run, we compare the prediction with what really happened, per spec. This tests the simulator, not the player. A live table of those results, per spec, is coming to this page.
Your own sharing for the simulator is off unless you turn it on. When another player shares a fight you were in, everyone in that group is part of it. Simulator data is always stored under anonymous IDs, with no names, whoever shared it. "Delete + go local" in the addon removes yours. Questions: [email protected]
See the accuracy table above for the specs on offer. A spec with little shared data will be rough, and sharing your fights is the fastest way to fix that.
No. The simulator never grades anyone. Scores come from how you played; the simulator only answers "what if".
Today: change your stats to see what new gear would do to your damage. Trinkets and different cast sequences come next.