Stake Ranked Episode 5 · BO3
result vs prediction
magic#25 in the world · loss 1:2series probability
65%35%
probability at match start · final 1:2
BIG#33 in the world · win 2:1The prediction was locked in using data as of the match start: map winrates, form and head-to-head were computed only from games before it began. This is how the model's honest accuracy is verified.
Map veto prediction
For each map — both teams' winrate over 3 months. The model guesses who bans it and who keeps it: teams remove maps where they are weaker and pull toward their strong ones.
Ancientmagic pick
50%45%
magic · 18 mapsedge magic +5%BIG · 7 maps
CacheBIG pick
56%56%
magic · 5 mapsnearly even +0%BIG · 14 maps
Miragemagic ban
27%47%
magic · 7 mapsedge BIG +20%BIG · 11 maps
Infernopossible decider
50%62%
magic · 6 mapsedge BIG +12%BIG · 9 maps
Dust2BIG ban
54%50%
magic · 9 mapsnearly even +4%BIG · 4 maps
AnubisBIG ban
38%50%
magic · 4 mapsedge BIG +12%BIG · 0 maps
Nukemagic ban
50%57%
magic · 0 mapsedge BIG +7%BIG · 3 maps
Most likely veto scenario
1. Banmagic remove Nuke
2. BanBIG remove Dust2
3. Pickmagic take Ancient
4. PickBIG take Cache
5. BanBIG remove Anubis
6. Banmagic remove Mirage
DeciderInferno (50% / 62%)
Prediction by map vs result
Probabilities were locked in using statistics as of the match start (map winrate over the 3 months before it). We compare against the actual result of each map.
Ancientprediction correct13:10
magic 58%BIG 42%
The model gave 58% on magic — the map was taken by magic.
Cacheprediction correct9:13
magic 47%BIG 53%
The model gave 53% on BIG — the map was taken by BIG.
Mirageprediction correct11:13
magic 30%BIG 70%
The model gave 70% on BIG — the map was taken by BIG.
Model check result
Model favoritemagic · 65%
Series resultBIG 1:2
Series predictionmissed
Accuracy by map3 of 3