46 Engines. Every Detail.

Levenberg-Marquardt nonlinear least-squares optimization. Multi-start fitting across rolling lookback windows. Parameter constraints: 0 < m < 1, 4 ≤ ω ≤ 15. tc stability scoring across multiple window sizes.

Closeness-to-resistance (30%), range compression (25%), volume accumulation (20%), candle microstructure (15%), catalyst alignment (10%), time coiling (bonus) — minus extension penalty, trap penalty, and pump penalty.

Bollinger Band width percentile, ATR compression ratio, volume dry-up detection, price coil tightness. Measures stored kinetic energy waiting to discharge directionally.

Drift (μ) estimated from weighted order flow momentum and trend persistence. Diffusion (σ²) from realized volatility, ATR, and microstructure noise. 2,000 stochastic paths per symbol over 15-minute horizons. 5-node confidence corridor extraction.

Orchestrates ALL sub-engine outputs into a unified forward path forecast. Composes technical momentum, structural health, institutional flow, sentiment extremes, and event catalysts into probability-weighted predictions.

Phase 1: parallel data ingestion. Phase 2: 46-engine parallel compute. Phase 3: 9 pure-function analyzers on shared context object. Phase 4: snapshot persistence for delta comparison and forecast learning.

Structural health scoring using topological data analysis principles. Computes persistence diagrams of price-volume phase space. Identifies liquidity voids (topological holes), cohesion decay, and hidden fragility.

Tracks capital movement across 8 market zones: large-cap growth, large-cap value, small-cap growth, small-cap value, bonds, commodities, crypto, cash. 5-day rolling window, 200 history snapshots. Detects money rotation before sector ETFs reflect it.

Fine-grained rotation detection below standard GICS classification. Identifies money flowing from defensive to cyclical, growth to value, mega-cap to mid-cap — at sub-sector and factor level.

Tracks all 5 Wyckoff phases simultaneously: Preliminary Support (PS), Selling Climax (SC), Automatic Rally (AR), Secondary Test (ST), Spring. Monitors support defense frequency, higher-low convergence, volume climax patterns.

Detects institutional accumulation through dark pool print patterns, block trade ratios, order flow imbalance persistence, and time-weighted volume distribution anomalies.

Real-time options unusual activity, put/call ratio extremes, short interest changes, borrow utilization spikes. Maps leveraged positioning pressure.

Views each symbol through 10 legendary investment philosophy lenses (momentum, value, growth, contrarian, etc.). Conviction score: strategy fit (25%), technical timing (15%), fundamental alignment (15%), regime fit (10%), flow (10%), risk/reward (10%), similar setups (7%), catalysts (5%), cross-agreement (3%).

Combines support proximity, volume profile POC, VWAP deviation, orderbook depth, microstructure noise level, and time-of-day statistics to score entry quality.

Aggregates signals from 1min, 5min, 15min, 1hr, 4hr, daily, weekly timeframes. Scores timeframe agreement, identifies divergences, and flags when all timeframes align.

Multi-factor scoring of breakout legitimacy: volume confirmation ratio, follow-through velocity, institutional participation, previous trap locations. Identifies bull traps, bear traps, fakeouts, manufactured liquidation events.

Monitors for fat-tail risk conditions: VIX term structure inversion, correlation breakdown, liquidity evaporation signals, circuit breaker proximity.

Combines short interest (SI%), borrow utilization, cost-to-borrow trends, days-to-cover, and buy-side flow acceleration to identify squeeze setups before they trigger.

Classifies current market into one of 6 regimes: trending up, trending down, mean-reverting, high-vol expansion, low-vol compression, transition. Adapts all other engine parameters based on detected regime.

Every prediction is persisted with timestamp, parameters, and input state. When the forecast window expires, actual outcomes are compared. Systematic drift triggers parameter recalibration. The system literally gets better with every cycle.

Computes multi-dimensional feature vectors for current setups. Searches historical database for statistically similar patterns. Reports outcome distribution of historical matches.

Decomposes signals into trend, momentum, and noise components across 7 timeframes. Identifies when trend signals align vs. conflict. Produces unified multi-scale view.