Model Leaderboard
Prediction accuracy across professional tier matches · 1183/2012 settled
#1
66.7%
Nexus
META: 3-way consensus of Apex + Oracle + Eagle with avg confidence >55%.
60 ✓ / 90
30 wrong
#2
66.1%
Oracle
Conservative: Elo HQ dampened when streaks disagree.
72 ✓ / 109
37 wrong
#3
65.6%
Streak
Win/loss streak signal blended with Elo.
84 ✓ / 128
44 wrong
#4
63.7%
Consensus
META: triggers when ≥75% of base models converge on same side.
58 ✓ / 91
33 wrong
#5
62.9%
Elo HQ
High-confidence Elo: smaller K-factor, scale=2000 — slow learner.
61 ✓ / 97
36 wrong
#6
60.3%
Patch-Aware
Elo with patch-boundary decay — stale form loses weight after Valve patches.
76 ✓ / 126
50 wrong
#7
59.7%
Apex
Blend: Elo + form + streak — momentum-aware ensemble pick.
80 ✓ / 134
54 wrong
#8
59.0%
Roster Change
Penalty for unstable rosters — fresh lineups underperform Elo.
79 ✓ / 134
55 wrong
#9
58.4%
Fatigue
Dota-specific: penalty for back-to-back matches (<12h apart).
80 ✓ / 137
57 wrong
#10
57.9%
SOS
Strength of schedule: form weighted by opponent quality.
77 ✓ / 133
56 wrong
#11
57.7%
Form
Pure win-rate over last 10 matches.
79 ✓ / 137
58 wrong
#12
57.2%
Elo H2H
Standard head-to-head Elo, scale=400.
79 ✓ / 138
59 wrong
#13
56.2%
Eagle
Blend: Elo + WR10 + form — weighted toward top-tier record.
77 ✓ / 137
60 wrong
#14
54.0%
Synergy
Hero co-occurrence winrates within recent team lineups.
75 ✓ / 139
64 wrong
#15
53.2%
Meta Strength
Team's recent hero pool weighted by current meta winrates.
66 ✓ / 124
58 wrong
#16
52.3%
Tier Elo
Elo computed only from matches against top-30 teams.
45 ✓ / 86
41 wrong
#17
52.2%
Phantom
High-conviction: only commits when Elo + WR10 strongly agree.
12 ✓ / 23
11 wrong
#18
44.7%
WR10
Win-rate vs top-10 opponents, blended with Elo.
17 ✓ / 38
21 wrong
◉
Calibrating — fewer than 20 settled predictions on this tier. Ranking appears once more matches complete.
54.5%
Vanguard
META: fires only when Elo and WR10 strongly agree (>60% same side).
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.