Model Leaderboard
Prediction accuracy across all tier matches · 886/1474 settled
#1
69.5%
Nexus
META: 3-way consensus of Apex + Oracle + Eagle with avg confidence >55%.
57 ✓ / 82
25 wrong
#2
66.2%
Streak
Win/loss streak signal blended with Elo.
49 ✓ / 74
25 wrong
#3
66.2%
Oracle
Conservative: Elo HQ dampened when streaks disagree.
43 ✓ / 65
22 wrong
#4
65.6%
Elo HQ
High-confidence Elo: smaller K-factor, scale=2000 — slow learner.
40 ✓ / 61
21 wrong
#5
63.2%
Roster Change
Penalty for unstable rosters — fresh lineups underperform Elo.
72 ✓ / 114
42 wrong
#6
62.7%
Consensus
META: triggers when ≥75% of base models converge on same side.
37 ✓ / 59
22 wrong
#7
61.1%
Apex
Blend: Elo + form + streak — momentum-aware ensemble pick.
77 ✓ / 126
49 wrong
#8
60.8%
Eagle
Blend: Elo + WR10 + form — weighted toward top-tier record.
76 ✓ / 125
49 wrong
#9
60.8%
Fatigue
Dota-specific: penalty for back-to-back matches (<12h apart).
79 ✓ / 130
51 wrong
#10
60.5%
Elo H2H
Standard head-to-head Elo, scale=400.
78 ✓ / 129
51 wrong
#11
60.3%
Patch-Aware
Elo with patch-boundary decay — stale form loses weight after Valve patches.
44 ✓ / 73
29 wrong
#12
58.1%
WR10
Win-rate vs top-10 opponents, blended with Elo.
18 ✓ / 31
13 wrong
#13
57.8%
SOS
Strength of schedule: form weighted by opponent quality.
52 ✓ / 90
38 wrong
#14
55.2%
Meta Strength
Team's recent hero pool weighted by current meta winrates.
37 ✓ / 67
30 wrong
#15
54.3%
Form
Pure win-rate over last 10 matches.
44 ✓ / 81
37 wrong
#16
52.4%
Phantom
High-conviction: only commits when Elo + WR10 strongly agree.
11 ✓ / 21
10 wrong
#17
52.4%
Vanguard
META: fires only when Elo and WR10 strongly agree (>60% same side).
11 ✓ / 21
10 wrong
#18
51.0%
Tier Elo
Elo computed only from matches against top-30 teams.
25 ✓ / 49
24 wrong
#19
47.4%
Synergy
Hero co-occurrence winrates within recent team lineups.
36 ✓ / 76
40 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.