Where ORACLE signal meets sharp market price.
About this Model
NEXUS blends ORACLE's internal probability with live Polymarket implied odds via log-odds averaging. When a liquid Polymarket market exists (>$5,000 volume), NEXUS weights the sharp money signal against ORACLE's model output, producing the tightest calibrated probability. The blend weight is dynamic: higher market volume increases Polymarket's weight. NEXUS identifies value bets as the delta between the blended probability and the raw market price — positive delta = model sees an edge.
Performance Charts
Solid = cumulative; dashed = rolling-20 (shown after 5 predictions); gray = 50% baseline.
Bars = actual win rate per confidence bucket; dashed = perfect calibration.
Cumulative units of profit/loss across all shadow bets for this model.
Top Feature Drivers
Internal model signal (converted to log-odds space)
Market-implied probability from current YES/NO prices
Dynamic weight scaling with log($volume)
Stale markets downweighted if >48h from match
Feature importances reflect normalized split gain across the training set. Values are approximate — the model uses all features simultaneously.
How to Interpret
NEXUS edge = NEXUS probability − market implied probability. Positive edge means CounterQuant rates the team higher than the market. The shadow bet system bets on positive-edge predictions using half-Kelly sizing. A large positive NEXUS edge (>5pp) is a strong signal; values under 2pp are noise.
Per-Tier Breakdown
| Tier | Preds | Correct | Acc |
|---|---|---|---|
| T1 | 5 | 2 | 40.0% |
| T2 | 1 | 0 | — |
Monthly Performance
| Month | Preds | Correct | Acc |
|---|---|---|---|
| Jun 2026 | 6 | 2 | 33.3% |
Full Prediction History (6 settled)
| Date | Matchup | Tier | Predicted Winner | Confidence | Result |
|---|---|---|---|---|---|
| Jun 6 | B8 vs M80 | T1 | B8 |
|
✗ |
| Jun 6 | G2 vs Monte | T1 | G2 |
|
✓ |
| Jun 6 | Astralis vs 9z | T1 | Astralis |
|
✗ |
| Jun 6 | G2 vs M80 | T1 | M80 |
|
✗ |
| Jun 6 | Johnny Speeds vs 1win | T2 | 1win |
|
✗ |
| Jun 6 | 9z vs FlyQuest | T1 | 9z |
|
✓ |
Live Performance
Calibration Quality
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
Selector Role
Preferred over ORACLE when a liquid Polymarket market exists (>$5k volume).
Version History
Fixed 50/50 ORACLE-market blend
Dynamic liquidity-weighted blending; Kelly bet sizing on NEXUS edge