Model Leaderboard
Prediction accuracy across all tier matches · 4999/7487 settled
#1
77.4%
Elo HQ
High-confidence Elo: smaller K-factor, scale=2000 — slow learner.
24 ✓ / 31
7 wrong
#2
73.5%
Oracle
Conservative: Elo HQ dampened when streaks disagree.
111 ✓ / 151
40 wrong
#3
73.0%
Consensus
META: triggers when ≥75% of base models converge on same side.
284 ✓ / 389
105 wrong
#4
72.9%
Vanguard
META: fires only when Elo and WR10 strongly agree (>60% same side).
151 ✓ / 207
56 wrong
#5
72.0%
Nexus
META: 3-way consensus of Apex + Oracle + Eagle with avg confidence >55%.
239 ✓ / 332
93 wrong
#6
70.9%
Phantom
High-conviction: only commits when Elo + WR10 strongly agree.
188 ✓ / 265
77 wrong
#7
70.4%
Streak
Win/loss streak signal blended with Elo.
426 ✓ / 605
179 wrong
#8
70.3%
Roster Change
Penalty for unstable rosters — fresh lineups underperform Elo.
277 ✓ / 394
117 wrong
#9
70.0%
Elo H2H
Standard head-to-head Elo, scale=400.
326 ✓ / 466
140 wrong
#10
69.5%
Patch-Aware
Elo with patch-boundary decay — stale form loses weight after Valve patches.
314 ✓ / 452
138 wrong
#11
69.0%
Apex
Blend: Elo + form + streak — momentum-aware ensemble pick.
296 ✓ / 429
133 wrong
#12
68.9%
Fatigue
Dota-specific: penalty for back-to-back matches (<12h apart).
323 ✓ / 469
146 wrong
#13
67.4%
Eagle
Blend: Elo + WR10 + form — weighted toward top-tier record.
302 ✓ / 448
146 wrong
#14
65.7%
SOS
Strength of schedule: form weighted by opponent quality.
414 ✓ / 630
216 wrong
#15
64.1%
Form
Pure win-rate over last 10 matches.
321 ✓ / 501
180 wrong
#16
64.0%
WR10
Win-rate vs top-10 opponents, blended with Elo.
236 ✓ / 369
133 wrong
#17
59.7%
Meta Strength
Team's recent hero pool weighted by current meta winrates.
92 ✓ / 154
62 wrong
#18
58.0%
Tier Elo
Elo computed only from matches against top-30 teams.
267 ✓ / 460
193 wrong
#19
55.5%
Synergy
Hero co-occurrence winrates within recent team lineups.
408 ✓ / 735
327 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.