The Science Behind Every Signal
AURUM runs a three-phase scoring pipeline combining classical technical signals, 9 intelligence layers, and 14 macro + ML modules — plus institutional alpha bundles for a total of 95 layers. Every signal is validated against an adaptive threshold and Monte Carlo probability bounds before delivery.
Technical Foundation
Baseline scoring on daily OHLCV produces a 0–100 raw technical score. Only setups that clear the foundational quality filter advance.
Intelligence Bridge · 9 Layers
Unified Scorer · 14 Layers
28 Additional Alpha Layers
Beyond the 24 core pipeline layers, AURUM runs 28 additional institutional-grade alpha modules — the dealer-flow, macro, and liquidity signals retail platforms don't touch.
Order Flow + Positioning
- · Gamma Exposure (GEX) surface
- · Dealer positioning + hedging flow
- · Quiet-signal detection (pre-volatility)
- · Modern tape classification
- · Auction market profile
- · VPIN (volume-synchronized toxicity)
- · Dollar volume imbalance
- · Liquidity absorption detection
- · Footprint delta divergence
- · Order flow exhaustion
- · Block trade imprint
- · Sweep vs iceberg classification
- · Opening rotation conviction
- · Session extension quality
- · Intraday regime shift
Regime + Capital Flows
- · Fed balance sheet liquidity
- · Real yield nowcast
- · Global liquidity composite
- · Dispersion regime (index vs constituents)
- · Repo stress index
- · ETF arbitrage discount/premium
- · VIX term structure (contango/backwardation)
- · Zweig breadth thrust monitor
- · Buyback blackout window awareness
- · Cash settlement bias
- · Commodity ratio regime (gold/silver, oil/nat-gas)
- · VVIX regime classification
- · Cross-margin systemic risk
Gates Before Delivery
Five Learning Loops Running Continuously
Every threshold, weight, and causal assumption is learned from live outcomes — not hardcoded, not backtested on synthetic data.
Per-asset online Bayesian updates. Thresholds tighten or loosen automatically from live signal outcomes — no hardcoded cutoffs.
Statistical test on rolling win rate. When an asset class diverges from its baseline distribution, AURUM auto-pauses signals until performance re-stabilizes.
Separate threshold + factor-weight sets for TRENDING, RANGING, and VOLATILE regimes. The engine learns which factors matter when.
Formal statistical lead-lag analysis on cross-asset pairs (DXY→Gold, VIX→SPY, BTC→ETH). Replaces hardcoded cascade assumptions with evidence.
Per-layer weight calibration from realized wins and losses — not synthetic backtests. The system learns from what actually happened.
The system improves with every closed signal.
Ten Archetypes, Automatically Classified
Each signal is tagged with its trade archetype before delivery — so you know the hold window and exit mechanics up front.
Default trade type when no specialized trigger fires
RSI extreme + order block alignment at structural level
Structure break + volume > 1.4× 20-day average
FVG present, no break, RSI in 40–65 range
Peak-liquidity sessions only; tight R:R ≥ 2.0 required
Session-aware setups during active market hours
Post-earnings announcement drift on 20 large-cap equities
NYSE open gap reversion on qualified US equities
Social sentiment cascade on retail-driven tickers
10 auto-selected strategies: Greeks + flow + IV rank
The Same Rigor, Applied to Sports
Sports runs its own ten-stage pipeline — substrate ingestion to CLV settlement — emitting an AURUM Composite Score with seven named drivers and an Anti-Fragility 4-gate contrarian cascade.
SDIO multi-league box scores, SGO sportsbook lines (open + close + live), official-API depth charts, weather, injuries.
Per-player + team projections from rolling form, schedule density, opponent matchup, and lineup state.
Schedule density · Rotation stress · Environment impact · Phase fragility · Series variance — each emits its own factor weight.
Multi-book consensus removes vig; EV strength scored against best line. Sharp money divergence detected per market.
7-driver attribution: EV strength, win-prob quality, devigging confidence, overlay context, variance penalty, substrate confidence, anti-fragility flag.
4-gate contrarian filter: AURUM disagrees with herding consensus AND favored team has structural fragility. When it fires, conviction is high.
Picks tagged STEADY_EDGE / HIGH_VARIANCE / CONTRARIAN — match selection to your bankroll temperament.
Grade C+ floor for BASIC, B+ for PRO, custom for ELITE. Telegram + Discord delivery with full driver breakdown.
DFS classic optimizer with correlation copula, ownership leverage, and stack constraint solver. Pick-em slates filtered for + EV.
Auto-grade vs final results, compute closing-line value, attribute alpha per overlay. Feeds back into the adaptive thresholds nightly.