DIRECT RESOURCE LOOP
Warrior
Build Rage by dealing or taking damage. Shield Block lets you defend while building that resource; stance choices add another layer as you learn.
Learn Warrior →MEASURE THE CLIMB
A tier list.
With the workings shown.
Nine classes put through the same automated challenge. See what the simulator achieved, then choose a class you enjoy learning.
THE BENCHMARK4,500 runs.
4,500 runs.
One consistent baseline.
Default difficulty. No allocated account talents. Fixed seeds. The upstream bot makes the decisions. These results measure this bot and these settings; they do not establish the strongest class for a human player.
SOURCE v1.0.3 · 68932e1
Class benchmark results
| Class | Win rate | Wins / runs | 95% interval | Avg. floors | Avg. act |
|---|---|---|---|---|---|
| 01Warlock | 20.4% | 102 / 500 | 17.1–24.2% | 68.0 | 5.40 |
| 02Mage | 19.8% | 99 / 500 | 16.5–23.5% | 74.2 | 5.75 |
| 03Paladin | 18.8% | 94 / 500 | 15.6–22.5% | 69.1 | 5.50 |
| 04Warrior | 18.4% | 92 / 500 | 15.2–22.0% | 73.7 | 5.75 |
| 05Hunter | 17.6% | 88 / 500 | 14.5–21.2% | 69.0 | 5.42 |
| 06Shaman | 16.8% | 84 / 500 | 13.8–20.3% | 61.9 | 4.89 |
| 07Priest | 14.2% | 71 / 500 | 11.4–17.5% | 66.6 | 5.29 |
| 08Rogue | 10.6% | 53 / 500 | 8.2–13.6% | 64.6 | 5.13 |
| 09Druid | 4.0% | 20 / 500 | 2.6–6.1% | 50.0 | 4.13 |
What this test can tell you
The game’s headless simulator runs the actual upstream engine. Every class uses the same seed sequence, normal difficulty and no allocated account talent points. The automated policy drafts cards, chooses routes and plays fights without human intervention.
Average floors and acts describe how far the bot progressed. An average act is not an act-clear percentage. The 95% Wilson intervals describe sampling uncertainty in the recorded win rates; they do not account for bias in the bot’s decision-making.
Account talents, manual sequencing, draft choices and different difficulty settings can change the outcome. The default starting continent and first eligible zone differ by class, so this is a full-run benchmark, rather than an isolated combat balance test. All class reward cards are unlocked. A class that needs careful resource timing may look weaker under an automated policy. Avoid treating a small gap as proof that one class is better.
Reproduce the benchmark
git clone https://github.com/Chrisl252/spiresofazeroth.git
cd spiresofazeroth
git checkout 68932e11873dec0a734c7dc9db3d7156b046b6bf
node tools/spires/sim.mjs --class all --runs 500 --seed soa-guide-20261101The published dataset is pinned to the source commit above. The site repository also contains the runner and detailed results.
Download all seeded runs (CSV) ↗ · Read the run summary ↗EDITORIAL STARTING POINTS
Pick a class by how it plays
These recommendations describe mechanics, not measured human win rates.
SPELL SCHOOL SYNERGY
Mage
Frost control, Fire finishers and Arcane Charges create distinct plans. Focus your draft around a school instead of collecting every powerful spell.
Learn Mage →A COMPANION TO MANAGE
Hunter
A guardian pet intercepts single-target attacks. Keep it healthy, establish your Aspect, then turn marks and shots into a focused damage plan.
Learn Hunter →Game data: Spires of Azeroth · 68932e1. MIT licensed. Sources & methodology