Model Leaderboard
Prediction accuracy across professional tier matches · 478/835 settled
#1
65.9%
Elo HQ
High-confidence Elo: smaller K-factor, scale=2000 — slow learner.
27 ✓ / 41
14 wrong
#2
64.0%
Streak
Win/loss streak signal blended with Elo.
32 ✓ / 50
18 wrong
#3
63.9%
Nexus
META: 3-way consensus of Apex + Oracle + Eagle with avg confidence >55%.
23 ✓ / 36
13 wrong
#4
63.6%
Oracle
Conservative: Elo HQ dampened when streaks disagree.
28 ✓ / 44
16 wrong
#5
61.5%
Consensus
META: triggers when ≥75% of base models converge on same side.
24 ✓ / 39
15 wrong
#6
59.6%
Patch-Aware
Elo with patch-boundary decay — stale form loses weight after Valve patches.
31 ✓ / 52
21 wrong
#7
59.6%
Roster Change
Penalty for unstable rosters — fresh lineups underperform Elo.
34 ✓ / 57
23 wrong
#8
58.6%
Apex
Blend: Elo + form + streak — momentum-aware ensemble pick.
34 ✓ / 58
24 wrong
#9
57.9%
Eagle
Blend: Elo + WR10 + form — weighted toward top-tier record.
33 ✓ / 57
24 wrong
#10
56.7%
Elo H2H
Standard head-to-head Elo, scale=400.
34 ✓ / 60
26 wrong
#11
56.7%
Fatigue
Dota-specific: penalty for back-to-back matches (<12h apart).
34 ✓ / 60
26 wrong
#12
55.8%
Meta Strength
Team's recent hero pool weighted by current meta winrates.
29 ✓ / 52
23 wrong
#13
55.0%
SOS
Strength of schedule: form weighted by opponent quality.
33 ✓ / 60
27 wrong
#14
51.8%
Form
Pure win-rate over last 10 matches.
29 ✓ / 56
27 wrong
#15
48.3%
Tier Elo
Elo computed only from matches against top-30 teams.
14 ✓ / 29
15 wrong
#16
44.6%
Synergy
Hero co-occurrence winrates within recent team lineups.
25 ✓ / 56
31 wrong
◉
Calibrating — fewer than 20 settled predictions on this tier. Ranking appears once more matches complete.
50.0%
WR10
Win-rate vs top-10 opponents, blended with Elo.
calibrating (12/20)
◉
Calibrating — fewer than 20 settled predictions on this tier. Ranking appears once more matches complete.
50.0%
Phantom
High-conviction: only commits when Elo + WR10 strongly agree.
calibrating (8/20)
◉
Calibrating — fewer than 20 settled predictions on this tier. Ranking appears once more matches complete.
50.0%
Vanguard
META: fires only when Elo and WR10 strongly agree (>60% same side).
calibrating (8/20)
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.