WORLD CHAMPIONSHIP · S16
Read every Worlds draft.
Explore champion picks, bans and role trends.
173 champions5 roles26.20Worlds outcomes pending
Evidence-supported roles + separate patch priors. Unsupported roles stay exactly zero; original V2 is retained as a comparator. Pick bars are stacked by role.
Professional games2556Historical input
Champions173champions
Champion role details5roles
Shared features17Frozen model
Worlds outcomes pending · Comparators are forecasts or historical statistics. Differences are not Worlds prediction errors; actual outcomes remain blank.
01 / COMPARE
Select a champion for role detailsForecasts and references, side by side.
Top 20 · Role breakdowns
TOP 20All champions
17302 / EXPLORE
Explore every champion.
Sorted by selected metric · Select a row for details · Scroll horizontally on mobile
| Champion | Pick | Ban | Presence | Mean rank | Mean step | Historical reference | V3 95% |
|---|
Champion role details
03 / EVALUATE
Assess the evidence and its limits.
| Metric | V2 | V3 |
|---|---|---|
| Pick MAE / pp | 3.3161 | 3.2834 |
| Pick RMSE / pp | 5.3810 | 5.3696 |
| Ban MAE / pp | 3.7430 | 3.7435 |
| Presence RMSE / pp | 8.6146 | 8.5959 |
| Champion × role RMSE / pp | 2.4599 | 2.4542 |
| Pick-weighted rank MAE | 1.4622 | 1.4602 |
| Pick-weighted BP-step MAE | 2.4665 | 2.4612 |
Gains are small, not proof of better Worlds accuracy or actual26.19/26.20 effects. Parameters train/select on S15 only; S16 does not tune them.
Full method, training and S15/S16 errors · Per-window errors · Reproduction hashes04 / UNDERSTAND
Transparent parameters, traceable results.
Current model parameters
| ridge | 0.03 |
| residual_blend | 0.75 |
| position_blend | 0.5 |
Comparators are forecasts or historical statistics. Differences are not Worlds prediction errors; actual outcomes remain blank.