Model Leaderboard
Prediction accuracy across premium tier matches · 408/639 settled
#1
73.9%
Nexus
META: 3-way consensus of Apex + Oracle + Eagle with avg confidence >55%.
34 ✓ / 46
12 wrong
#2
71.4%
Oracle
Conservative: Elo HQ dampened when streaks disagree.
15 ✓ / 21
6 wrong
#3
70.8%
Streak
Win/loss streak signal blended with Elo.
17 ✓ / 24
7 wrong
#4
66.7%
Roster Change
Penalty for unstable rosters — fresh lineups underperform Elo.
38 ✓ / 57
19 wrong
#5
65.0%
Elo HQ
High-confidence Elo: smaller K-factor, scale=2000 — slow learner.
13 ✓ / 20
7 wrong
#6
65.0%
Consensus
META: triggers when ≥75% of base models converge on same side.
13 ✓ / 20
7 wrong
#7
64.3%
Fatigue
Dota-specific: penalty for back-to-back matches (<12h apart).
45 ✓ / 70
25 wrong
#8
63.8%
Elo H2H
Standard head-to-head Elo, scale=400.
44 ✓ / 69
25 wrong
#9
63.3%
SOS
Strength of schedule: form weighted by opponent quality.
19 ✓ / 30
11 wrong
#10
63.2%
Apex
Blend: Elo + form + streak — momentum-aware ensemble pick.
43 ✓ / 68
25 wrong
#11
63.2%
Eagle
Blend: Elo + WR10 + form — weighted toward top-tier record.
43 ✓ / 68
25 wrong
#12
61.9%
Patch-Aware
Elo with patch-boundary decay — stale form loses weight after Valve patches.
13 ✓ / 21
8 wrong
#13
60.0%
Form
Pure win-rate over last 10 matches.
15 ✓ / 25
10 wrong
#14
55.0%
Tier Elo
Elo computed only from matches against top-30 teams.
11 ✓ / 20
9 wrong
#15
55.0%
Synergy
Hero co-occurrence winrates within recent team lineups.
11 ✓ / 20
9 wrong
◉
Calibrating — fewer than 20 settled predictions on this tier. Ranking appears once more matches complete.
63.2%
WR10
Win-rate vs top-10 opponents, blended with Elo.
calibrating (19/20)
◉
Calibrating — fewer than 20 settled predictions on this tier. Ranking appears once more matches complete.
53.8%
Phantom
High-conviction: only commits when Elo + WR10 strongly agree.
calibrating (13/20)
◉
Calibrating — fewer than 20 settled predictions on this tier. Ranking appears once more matches complete.
53.3%
Meta Strength
Team's recent hero pool weighted by current meta winrates.
calibrating (15/20)
◉
Calibrating — fewer than 20 settled predictions on this tier. Ranking appears once more matches complete.
53.8%
Vanguard
META: fires only when Elo and WR10 strongly agree (>60% same side).
calibrating (13/20)
Accuracy is measured per tournament tier — Premium is the clean, high-signal subset (mirrors cs2predict's top tier). Models with fewer than 20 settled predictions on the selected tier are calibrating. Coin-flip predictions (48–52%) are excluded. Click a model for confidence buckets, Brier score, and recent picks.