Oracle

Ensemble probability forecasting from 6 signal sources. Surface markets where your model sees edge vs market price.

Ultra Feature. Requires an Ultra subscription. Get started at api.mathematicalcompany.com

Oracle

Combines six independent signal extractors into a single ensemble probability estimate for each market, then surfaces edge opportunities where the forecast diverges materially from the current market price.

Smart Money Flow

Holder analysis weighted by wallet score.

Microstructure

Kyle’s lambda + VPIN for informed trading detection.

Momentum

EMA crossover (short/long span ratio).

Volume Profile

Buy vs sell volume imbalance.

Holder Concentration

HHI of top holders as conviction signal.

Temporal Patterns

Hour-of-day buy activity patterns.

---

Forecast a Market

import horizon as hz

fc = hz.forecast_market(
    "0xcondition...",
    market_title="Will X happen?",
    market_price=0.65,
)
print(f"Ensemble: {fc.ensemble_prob:.4f}")
print(f"Edge: {fc.edge_bps:.0f} bps")
for s in fc.signals:
    print(f"  {s.name}: {s.normalized_value:.4f} (weight={s.weight:.2f})")

---

Scan for Edges

edges = hz.scan_edges(min_edge_bps=300, max_markets=30)
for e in edges:
    print(f"{e.market_title}: {e.direction} | edge={e.edge_bps:.0f}bps | size=${e.recommended_size:.0f}")

---

Pipeline Mode

hz.run(
    name="oracle-mm",
    markets=["0xcondition..."],
    pipeline=[
        hz.oracle(forecast_interval_cycles=20),
        model,   # reads ctx.params["oracle_edge_bps"]
        quoter,
    ],
    interval=5.0,
)

Injects ctx.params["oracle_forecast"] and ctx.params["oracle_edge_bps"] every N cycles.

---

Full Report

report = hz.oracle_report(max_markets=20)
print(f"Coverage: {report.coverage} markets")
print(f"Avg confidence: {report.avg_confidence:.4f}")
print(f"Calibration (log loss): {report.model_calibration:.4f}")
for edge in report.top_edges:
    print(f"  {edge.market_title}: {edge.edge_bps:.0f}bps ({edge.direction})")

---

Signal Weights

Default weights (configurable via OracleConfig):

Signal	Weight	Source
smart_money_flow	0.25	get_top_holders() + score_wallet()
microstructure	0.20	get_market_trades() + Kyle’s lambda + VPIN
momentum	0.15	EMA crossover on trade prices
volume_profile	0.15	Buy/sell volume ratio
holder_concentration	0.15	HHI of top holders
temporal_pattern	0.10	Hour-of-day activity analysis

---

OracleConfig

Parameter	Default	Description
signal_weights	See above	Weight per signal in ensemble
min_confidence	0.3	Minimum confidence to report forecast
edge_threshold_bps	200.0	Minimum edge to flag as opportunity
recalibrate_interval	50	Cycles between recalibration
holder_limit	50	Max holders to fetch per market
trade_limit	200	Max trades to fetch per market