Model Leaderboard
Prediction accuracy across professional tier matches · 1314/1927 settled
#1
75.7%
Oracle
Conservative: Elo HQ dampened when streaks disagree.
28 ✓ / 37
9 wrong
#2
75.7%
Consensus
META: triggers when ≥75% of base models converge on same side.
81 ✓ / 107
26 wrong
#3
75.0%
Phantom
High-conviction: only commits when Elo + WR10 strongly agree.
39 ✓ / 52
13 wrong
#4
75.0%
Vanguard
META: fires only when Elo and WR10 strongly agree (>60% same side).
30 ✓ / 40
10 wrong
#5
72.6%
Roster Change
Penalty for unstable rosters — fresh lineups underperform Elo.
69 ✓ / 95
26 wrong
#6
72.3%
Nexus
META: 3-way consensus of Apex + Oracle + Eagle with avg confidence >55%.
68 ✓ / 94
26 wrong
#7
72.2%
Elo H2H
Standard head-to-head Elo, scale=400.
78 ✓ / 108
30 wrong
#8
72.0%
Patch-Aware
Elo with patch-boundary decay — stale form loses weight after Valve patches.
77 ✓ / 107
30 wrong
#9
71.7%
Apex
Blend: Elo + form + streak — momentum-aware ensemble pick.
86 ✓ / 120
34 wrong
#10
71.2%
Streak
Win/loss streak signal blended with Elo.
114 ✓ / 160
46 wrong
#11
70.8%
Eagle
Blend: Elo + WR10 + form — weighted toward top-tier record.
85 ✓ / 120
35 wrong
#12
69.7%
Fatigue
Dota-specific: penalty for back-to-back matches (<12h apart).
76 ✓ / 109
33 wrong
#13
67.8%
SOS
Strength of schedule: form weighted by opponent quality.
122 ✓ / 180
58 wrong
#14
66.7%
Meta Strength
Team's recent hero pool weighted by current meta winrates.
32 ✓ / 48
16 wrong
#15
65.2%
Form
Pure win-rate over last 10 matches.
101 ✓ / 155
54 wrong
#16
61.5%
WR10
Win-rate vs top-10 opponents, blended with Elo.
40 ✓ / 65
25 wrong
#17
56.4%
Synergy
Hero co-occurrence winrates within recent team lineups.
114 ✓ / 202
88 wrong
#18
55.7%
Tier Elo
Elo computed only from matches against top-30 teams.
68 ✓ / 122
54 wrong
◉
Calibrating — fewer than 20 settled predictions on this tier. Ranking appears once more matches complete.
100.0%
Elo HQ
High-confidence Elo: smaller K-factor, scale=2000 — slow learner.
calibrating (6/20)
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.