Model Leaderboard
Prediction accuracy across all tier matches · 2793/4763 settled
#1
66.1%
Nexus
META: 3-way consensus of Apex + Oracle + Eagle with avg confidence >55%.
146 ✓ / 221
75 wrong
#2
64.6%
Streak
Win/loss streak signal blended with Elo.
184 ✓ / 285
101 wrong
#3
63.9%
Consensus
META: triggers when ≥75% of base models converge on same side.
117 ✓ / 183
66 wrong
#4
63.8%
Oracle
Conservative: Elo HQ dampened when streaks disagree.
157 ✓ / 246
89 wrong
#5
61.5%
Vanguard
META: fires only when Elo and WR10 strongly agree (>60% same side).
32 ✓ / 52
20 wrong
#6
59.4%
Elo HQ
High-confidence Elo: smaller K-factor, scale=2000 — slow learner.
130 ✓ / 219
89 wrong
#7
59.4%
Apex
Blend: Elo + form + streak — momentum-aware ensemble pick.
193 ✓ / 325
132 wrong
#8
58.9%
Eagle
Blend: Elo + WR10 + form — weighted toward top-tier record.
201 ✓ / 341
140 wrong
#9
58.7%
Fatigue
Dota-specific: penalty for back-to-back matches (<12h apart).
196 ✓ / 334
138 wrong
#10
58.3%
Roster Change
Penalty for unstable rosters — fresh lineups underperform Elo.
187 ✓ / 321
134 wrong
#11
58.2%
SOS
Strength of schedule: form weighted by opponent quality.
163 ✓ / 280
117 wrong
#12
57.8%
Patch-Aware
Elo with patch-boundary decay — stale form loses weight after Valve patches.
160 ✓ / 277
117 wrong
#13
57.3%
Elo H2H
Standard head-to-head Elo, scale=400.
193 ✓ / 337
144 wrong
#14
56.9%
Form
Pure win-rate over last 10 matches.
165 ✓ / 290
125 wrong
#15
55.9%
Phantom
High-conviction: only commits when Elo + WR10 strongly agree.
52 ✓ / 93
41 wrong
#16
54.8%
Tier Elo
Elo computed only from matches against top-30 teams.
125 ✓ / 228
103 wrong
#17
54.4%
WR10
Win-rate vs top-10 opponents, blended with Elo.
99 ✓ / 182
83 wrong
#18
53.9%
Synergy
Hero co-occurrence winrates within recent team lineups.
158 ✓ / 293
135 wrong
#19
52.7%
Meta Strength
Team's recent hero pool weighted by current meta winrates.
135 ✓ / 256
121 wrong
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.