ORACLE v2 LIVE
Match Outcome · Gradient Boosting

Market-independent deep stats match predictor.

101
Predictions
27 pending
51.4%
Accuracy
74 settled
-3.1%
Shadow ROI
-17.6u P&L
51.4%
Bet Hit Rate
74 bets

About this Model

ORACLE is CounterQuant's primary market-independent match outcome predictor. Trained on 27,000+ historical CS2 matches with 538 engineered features spanning team Elo trajectory (overall, form, SoS, per-map), head-to-head record, map pool overlap, recent form windows (7/14/30d), and roster stability. Isotonic calibration ensures probabilities are accurate across the full 0–1 range, not just directionally correct. ORACLE is the default model when no Polymarket market exists for a match.

Performance Charts

Accuracy Over Time

Solid = cumulative; dashed = rolling-20 (shown after 5 predictions); gray = 50% baseline.

Reliability Diagram

Bars = actual win rate per confidence bucket; dashed = perfect calibration.

Cumulative Shadow P&L

Cumulative units of profit/loss across all shadow bets for this model.

Top Feature Drivers

Map-specific Elo differential 28%

Per-map team Elo (elo_before from match history) — no leakage

Overall Elo differential 22%

Baseline team quality gap at match time

SoS-adjusted Elo 15%

Strength-of-schedule normalized rating

Form Elo (recent 30d) 12%

Recency-weighted performance trend

H2H win rate (last 10) 9%

Direct head-to-head record on this map

Map pool overlap score 7%

Fraction of preferred maps in common

Roster stability delta 4%

Recent lineup change penalty

Win rate difference (20-match) 3%

Rolling win rate gap including tier weighting

Feature importances reflect normalized split gain across the training set. Values are approximate — the model uses all features simultaneously.

How to Interpret

ORACLE probability is purely internal — it has never seen a betting market price. When ORACLE differs significantly from NEXUS (the market-blended model), that gap represents potential mispricing. An ORACLE probability above 60% with a market implied below 50% signals genuine alpha. Note the AUC is measured on a held-out test set with strict chronological splitting — no future data leaks into training.

Per-Tier Breakdown

Tier Preds Correct Acc
T1 28 17 60.7%
T2 21 8 38.1%
T3 25 13 52.0%

Monthly Performance

Month Preds Correct Acc
Jun 2026 74 38 51.4%

Full Prediction History (74 settled)

Date Matchup Tier Predicted Winner Confidence Result
Jun 9 PsychoFace vs GenOne T2 GenOne
26.1%
✓
Jun 9 BIG vs B8 T1 BIG
2.6%
✗
Jun 9 Virtus.pro vs Oxuji T2 Oxuji
10.0%
✗
Jun 9 KOLESIE vs RUSTEC T3 RUSTEC
12.9%
✗
Jun 9 Legacy vs TYLOO T1 Legacy
2.6%
✓
Jun 9 Betclic vs AM T3 AM
10.0%
✗
Jun 9 Monte vs paiN T1 Monte
2.6%
✓
Jun 9 ex-RUBY vs RUSTEC T2 RUSTEC
12.9%
✗
Jun 9 Lavked vs Misa T2 Misa
10.0%
✗
Jun 9 Abyssal vs Rooster T3 Abyssal
2.6%
✗
Jun 9 CYBERSHOKE vs ex-RUBY T3 CYBERSHOKE
2.6%
✗
Jun 9 MGLBROS vs MARKandLARRY T3 MARKandLARRY
10.0%
✓
Jun 9 MANA vs Oxuji T2 Oxuji
12.9%
✓
Jun 9 INOX Division vs Bebop T2 INOX Division
37.8%
✓
Jun 9 Club 333 vs Chicken Coop T3 Chicken Coop
41.9%
✓
Jun 9 Brahmas vs Villainous T3 Villainous
12.9%
✓
Jun 8 HEROIC Academy vs NEW VISION T2 NEW VISION
26.1%
✗
Jun 8 DONSTU vs Banger Gang T3 DONSTU
37.8%
✓
Jun 8 Acend vs ex-RUBY T3 ex-RUBY
26.1%
✓
Jun 8 Misa vs eternal premium T3 Misa
32.6%
✗
Jun 8 ROUNDS vs OLDBOYS PL T3 ROUNDS
2.6%
✗
Jun 8 M80 vs Legacy T1 Legacy
10.0%
✓
Jun 8 XI vs Clutchain T3 XI
23.6%
✓
Jun 8 G2 vs BIG T1 G2
9.6%
✓
Jun 8 Ursa vs RUSTEC T3 RUSTEC
10.0%
✓
Jun 8 Monte vs BetBoom T1 BetBoom
10.0%
✓
Jun 8 TYLOO vs 9z T1 TYLOO
2.6%
✗
Jun 8 The Last Resort vs GenOne T2 GenOne
10.0%
✓
Jun 8 MIBR vs B8 T1 B8
35.1%
✓
Jun 8 brazylijski luz vs ex-Zero Tenacity T2 ex-Zero Tenacity
12.9%
✗
Jun 8 Astralis vs paiN T1 Astralis
2.6%
✗
Jun 8 Nuclear TigeRES vs CYBERSHOKE T3 Nuclear TigeRES
2.6%
✓
Jun 8 Leo vs HYPERSPIRIT T2 HYPERSPIRIT
10.0%
✗
Jun 8 KOLESIE vs ASTRAL T3 KOLESIE
23.6%
✓
Jun 8 Chicken Coop vs Third Prime T3 Chicken Coop
29.9%
✓
Jun 7 G2 vs FUT T1 FUT
10.0%
✓
Jun 7 WROTBERRY vs Noir Verse T2 WROTBERRY
9.6%
✗
Jun 7 GenOne vs BIG Academy T2 BIG Academy
12.9%
✗
Jun 7 TDK vs INOX Division T3 INOX Division
10.0%
✗
Jun 7 B8 vs GamerLegion T1 B8
9.6%
✓
Jun 7 Spirit vs 9z T1 Spirit
2.6%
✓
Jun 7 Acend vs Phantom T2 Phantom
10.0%
✓
Jun 7 Betclic vs Spirit Academy T2 Betclic
32.6%
✗
Jun 7 CYBERSHOKE vs SHISHKA T2 SHISHKA
10.0%
✓
Jun 7 eSuba vs LPH T3 LPH
10.0%
✓
Jun 7 Lilmix vs NEW VISION T2 NEW VISION
26.1%
✓
Jun 7 FlyQuest vs paiN T1 FlyQuest
37.8%
✗
Jun 7 ARCRED vs Walczaki T3 ARCRED
9.6%
✗
Jun 7 BetBoom vs M80 T1 BetBoom
32.6%
✓
Jun 7 MIBR vs BIG T1 BIG
12.9%
✓

Live Performance

Predictions 101
Accuracy 51.4%
Shadow ROI -3.1%
Bet Hit Rate 51.4%
Shadow P&L -17.6u
Avg AUC (live) 0.7180

Calibration Quality

Brier Score 0.246
Quality Fair

Brier score measures probability calibration (0=perfect, 0.25=random). Lower is better. A well-calibrated model's confidence matches actual win rates.

Data Coverage

First prediction Jun 5, 2026
Last prediction Jun 9, 2026
Active span 5d
Avg / week 103.6 preds

Selector Role

Default — runs on every upcoming match when NEXUS cannot be applied.

Version History

v1 2025-Q4

Baseline 238-feature ensemble; replaced APEX v3

v2 2026-Q1

538 features; exp-decay rolling windows; SoS features added

All Models

← TITAN
ORACLE (current)
NEXUS →
← All Models