Model Leaderboard
Prediction accuracy across all tier matches · 2317/3421 settled
#1
77.0%
Consensus
META: triggers when ≥75% of base models converge on same side.
127 ✓ / 165
38 wrong
#2
75.7%
Oracle
Conservative: Elo HQ dampened when streaks disagree.
53 ✓ / 70
17 wrong
#3
73.1%
Apex
Blend: Elo + form + streak — momentum-aware ensemble pick.
141 ✓ / 193
52 wrong
#4
72.5%
Nexus
META: 3-way consensus of Apex + Oracle + Eagle with avg confidence >55%.
111 ✓ / 153
42 wrong
#5
71.5%
Streak
Win/loss streak signal blended with Elo.
193 ✓ / 270
77 wrong
#6
70.7%
Vanguard
META: fires only when Elo and WR10 strongly agree (>60% same side).
65 ✓ / 92
27 wrong
#7
70.6%
Roster Change
Penalty for unstable rosters — fresh lineups underperform Elo.
127 ✓ / 180
53 wrong
#8
70.5%
Elo H2H
Standard head-to-head Elo, scale=400.
141 ✓ / 200
59 wrong
#9
70.1%
Eagle
Blend: Elo + WR10 + form — weighted toward top-tier record.
148 ✓ / 211
63 wrong
#10
69.4%
Patch-Aware
Elo with patch-boundary decay — stale form loses weight after Valve patches.
136 ✓ / 196
60 wrong
#11
69.1%
Fatigue
Dota-specific: penalty for back-to-back matches (<12h apart).
141 ✓ / 204
63 wrong
#12
68.8%
Phantom
High-conviction: only commits when Elo + WR10 strongly agree.
88 ✓ / 128
40 wrong
#13
67.5%
SOS
Strength of schedule: form weighted by opponent quality.
187 ✓ / 277
90 wrong
#14
66.3%
Form
Pure win-rate over last 10 matches.
161 ✓ / 243
82 wrong
#15
65.2%
Meta Strength
Team's recent hero pool weighted by current meta winrates.
43 ✓ / 66
23 wrong
#16
64.2%
WR10
Win-rate vs top-10 opponents, blended with Elo.
120 ✓ / 187
67 wrong
#17
57.3%
Tier Elo
Elo computed only from matches against top-30 teams.
126 ✓ / 220
94 wrong
#18
56.1%
Synergy
Hero co-occurrence winrates within recent team lineups.
199 ✓ / 355
156 wrong
◉
Calibrating — fewer than 20 settled predictions on this tier. Ranking appears once more matches complete.
90.9%
Elo HQ
High-confidence Elo: smaller K-factor, scale=2000 — slow learner.
calibrating (11/20)
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.