The International 10 (2021) · BO2
result vs predictionseries probability
70%30%
probability at match start · final 2:0
The prediction was locked in using data as of the match start: hero winrates, form and head-to-head were computed only from games before it began. This is how the model's honest accuracy is verified.
Teams' go-to heroes · 6 months
Who plays what and how often they win. "Signature" — from 6 games with a 60%+ winrate: such a pick statistically raises the win chance.
Elephant
No games played over 6 months.
SG e-sports
No games played over 6 months.
Patch meta · 3 months
The most picked heroes of the pro scene and their winrate. Picking a meta hero adds points to the team in the model.
Lion
299 picks · 173 bans
Mars
285 picks · 396 bans
Lina
279 picks · 543 bans
Pangolier
258 picks · 374 bans
Hoodwink
253 picks · 88 bans
Shadow Demon
245 picks · 256 bans
Tidehunter
227 picks · 386 bans
Luna
224 picks · 317 bans
Prediction components
Form (series winrate)58% vs 25%
Hero pool strength50% Elephant
Head-to-head0 – 0
Model total70% Elephant
Prediction by game vs result
Probabilities were locked in using statistics as of the match start: the draft was assessed by hero comfort, counter-picks and late-game potential. We compare against the actual result.
Game 1prediction missed31:25
Elephant 70%SG e-sports 30%
The model gave 70% on Elephant — the game was taken by SG e-sports in 51:09.
Game 2prediction missed34:18
Elephant 70%SG e-sports 30%
The model gave 70% on Elephant — the game was taken by SG e-sports in 30:27.
Model check result
Model favoriteElephant · 70%
Series resultElephant 2:0
Series predictioncorrect
Accuracy by game0 of 2