Improved Spirit
Patch 2026-09-29.
Win rate by purchase phase
| Phase | Win rate when bought | Edge over same-tier items |
|---|---|---|
| Bought before 10 min | 50.0% ± 0.1 (n=910,475) | -1.6 pp |
| 10–20 min | 48.7% ± 0.1 (n=769,831) | -1.5 pp |
| 20 min or later | 51.4% ± 0.1 (n=473,868) | +0.1 pp |
Edge over same-tier items, by hero
| Hero | Win rate when bought | Edge over same-tier items |
|---|---|---|
 Warden | 49.6% ± 1.1 (n=7,532) | +1.0 pp |
 Paradox | 47.2% ± 1.0 (n=8,893) | +0.9 pp |
 Lady Geist | 50.7% ± 0.4 (n=55,162) | +0.9 pp |
 McGinnis | 50.2% ± 0.7 (n=18,871) | +0.9 pp |
 Pocket | 48.1% ± 0.6 (n=26,798) | +0.9 pp |
 Apollo | 51.9% ± 0.4 (n=64,246) | +0.6 pp |
 Mina | 47.6% ± 0.3 (n=135,760) | +0.5 pp |
 Infernus | 47.5% ± 0.2 (n=161,402) | +0.5 pp |
 Deadman Danny | 43.7% ± 0.3 (n=111,862) | +0.2 pp |
 Graves | 55.8% ± 0.3 (n=97,398) | +0.0 pp |
By rank tier
| Tier | Games | Win rate when bought | Purchase rate |
|---|---|---|---|
| All tiers | 2,154,176 | 49.9% ± 0.1 (n=2,154,176) | 46.4% (n=4,643,618) |
| 1 Initiate | 150,906 | 50.1% ± 0.3 (n=150,906) | 52.6% (n=286,633) |
| 2 Seeker | 291,631 | 50.3% ± 0.2 (n=291,631) | 51.4% (n=567,145) |
| 3 Acolyte | 268,379 | 50.2% ± 0.2 (n=268,379) | 49.8% (n=538,601) |
| 4 Sentinel | 219,882 | 50.1% ± 0.2 (n=219,882) | 48.4% (n=454,663) |
| 5 Mystic | 195,286 | 50.1% ± 0.2 (n=195,286) | 47.3% (n=412,948) |
| 6 Ritualist | 190,637 | 49.7% ± 0.2 (n=190,637) | 45.7% (n=417,480) |
| 7 Emissary | 179,536 | 49.7% ± 0.2 (n=179,536) | 44.5% (n=403,465) |
| 8 Oracle | 231,584 | 49.6% ± 0.2 (n=231,584) | 42.4% (n=546,368) |
| 9 Phantom | 173,300 | 50.0% ± 0.2 (n=173,300) | 40.1% (n=432,077) |
| 10 Ascendant | 80,626 | 50.0% ± 0.3 (n=80,626) | 37.7% (n=213,670) |
| 11 Eternus | 28,078 | 50.0% ± 0.6 (n=28,078) | 36.0% (n=78,077) |
How we calculate this
In observed win rates, ± shows half the width of the likely range; n is the number of games or appearances counted. Percentage-point differences are labeled pp.
Win rate when bought is observed wins divided by games among buyers, with a 95% likely range. Small samples or wide ranges are withheld. Edge over same-tier items is how much better or worse the buyers did than a model of the game expected just before purchase, compared with the average for same-tier items. These cover all ranks, clean matches, and are computed daily. They are observational, do not establish cause and do not add up to a build win rate. Follow hero links for win rate impact comparisons among that hero's alternatives. Purchase rates use all player appearances in the selection.