Model Leaderboard
Prediction accuracy across premium tier matches · 1003/1494 settled
#1
79.3%
Consensus
META: triggers when ≥75% of base models converge on same side.
46 ✓ / 58
12 wrong
#2
75.8%
Oracle
Conservative: Elo HQ dampened when streaks disagree.
25 ✓ / 33
8 wrong
#3
75.3%
Apex
Blend: Elo + form + streak — momentum-aware ensemble pick.
55 ✓ / 73
18 wrong
#4
72.9%
Nexus
META: 3-way consensus of Apex + Oracle + Eagle with avg confidence >55%.
43 ✓ / 59
16 wrong
#5
71.8%
Streak
Win/loss streak signal blended with Elo.
79 ✓ / 110
31 wrong
#6
69.2%
Eagle
Blend: Elo + WR10 + form — weighted toward top-tier record.
63 ✓ / 91
28 wrong
#7
68.5%
Elo H2H
Standard head-to-head Elo, scale=400.
63 ✓ / 92
29 wrong
#8
68.4%
Fatigue
Dota-specific: penalty for back-to-back matches (<12h apart).
65 ✓ / 95
30 wrong
#9
68.2%
Form
Pure win-rate over last 10 matches.
60 ✓ / 88
28 wrong
#10
68.2%
Roster Change
Penalty for unstable rosters — fresh lineups underperform Elo.
58 ✓ / 85
27 wrong
#11
67.3%
Vanguard
META: fires only when Elo and WR10 strongly agree (>60% same side).
35 ✓ / 52
17 wrong
#12
67.0%
SOS
Strength of schedule: form weighted by opponent quality.
65 ✓ / 97
32 wrong
#13
66.3%
Patch-Aware
Elo with patch-boundary decay — stale form loses weight after Valve patches.
59 ✓ / 89
30 wrong
#14
65.6%
WR10
Win-rate vs top-10 opponents, blended with Elo.
80 ✓ / 122
42 wrong
#15
64.5%
Phantom
High-conviction: only commits when Elo + WR10 strongly agree.
49 ✓ / 76
27 wrong
#16
59.2%
Tier Elo
Elo computed only from matches against top-30 teams.
58 ✓ / 98
40 wrong
#17
55.6%
Synergy
Hero co-occurrence winrates within recent team lineups.
85 ✓ / 153
68 wrong
◉
Calibrating — fewer than 20 settled predictions on this tier. Ranking appears once more matches complete.
80.0%
Elo HQ
High-confidence Elo: smaller K-factor, scale=2000 — slow learner.
calibrating (5/20)
◉
Calibrating — fewer than 20 settled predictions on this tier. Ranking appears once more matches complete.
61.1%
Meta Strength
Team's recent hero pool weighted by current meta winrates.
calibrating (18/20)
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.