Full-Stack Quantitative Engineering

The IRSTTS platform is a comprehensive financial intelligence ecosystem. It integrates Direct SEC EDGAR Ingestion with a proprietary Genetic Optimization Engine and an institutional Technical Analysis library of over 100 indicators.

System Robustness: From multi-threaded data sorters to Disk-Backed Memoization, the architecture is engineered to handle terabytes of Polygon.io and Massive.com tick-level data with sub-millisecond precision.

1. High-Performance Data Layer

Our ingestion engine, S3DataDownloader.cs and BulkDataSorter.cs , manages massive datasets with institutional reliability. We utilize data from Massive.com and Polygon.io, normalizing varied asset classes into unified PriceRecord.cs and OHLC.cs objects.

Compliant SEC Access

Proprietary SECHttpClient.cs with 100ms asynchronous throttling for federal regulatory compliance.

Disk-Backed Memoization

State-aware caching using SlidingCache.cs prevents redundant calculations across multi-year backtests.

2. The Quantitative Indicator Engine

Our library implements a vast array of technical logic, allowing us to overlay Fundamental Insider Conviction with Mechanical Momentum.

Momentum & Trend

Full implementations of MACD, RSI, ADX, and Ichimoku Cloud, validated via IndicatorValueCalibrator.

Bill Williams Suite

Advanced Alligator, Fractals, and Gator Oscillator logic for trend-exhaustion detection.

Volatility & Options

Integrated Black-Scholes ImpliedVolatilitySolver.cs to price derivatives and measure market fear.

3. Genetic Optimization Layer

To avoid the "Backtest Illusion" of manual curve-fitting, our system uses the GeneticSolver.cs . This evolutionary engine discovers the optimal parameters for cluster weights and position sizing by simulating thousands of generations.

The WalkForwardScoring module ensures that parameters discovered in the "In-Sample" phase survive a strict "Out-of-Sample" validation before being promoted to production.

4. Systematic Execution & Rotation

Executed via SimplePnLTracker.cs , the strategy maintains an anchor in the VOO Baseline. High-conviction Form 4 detections trigger an automated rotation governed by the DynamicPositionSizer.

Open-to-Open Filling

We fill at the next Market Open (9:30 AM EST) to ensure 100% executable fills with zero lookahead bias.

10% Core Safety

Mandatory liquidity guard ensures the portfolio is never 100% concentrated in idiosyncratic risk.

5. Immutable Forensic Audit

Transparency is enforced at the code level. The Trade class and PositionId GUID system link every transaction to its source.

Accession Number Linkage: Every trade is tied to the original XML primary document at SEC.gov.
Maturation Auditing: PriceActionMetrics.cs tracks performance at T+1, T+7, up to T+90 intervals.
Risk Accountability: We record the Peak Adverse Excursion (Max Drawdown) for every staged trade.

Institutional Mathematical Framework

Our decision-making is governed by deterministic formulas that eliminate human bias.

1. Composite Cluster Score ($S$)

The ClusterBuyAnalyzer.cs determines the tier of an alert by evaluating aggregate volume ($V$), insider counts ($N$), and role hierarchy weights ($W$).

\[ S = \min(60, 15 \log_{10}(V_{agg})) + \min(50, 12 \log_{10}(V_{30d})) + (8 \cdot N_{insider}) + (10 \cdot \sum W_{role}) \]

2. Dynamic Fitness Function ($\Phi$)

The GeneticSolver optimizes the system by maximizing a multi-variable fitness score that penalizes volatility and rewards consistent Alpha.

\[ \Phi = \frac{TotalPnL}{MaxDrawdown} \times \sqrt{N_{trades}} \times (1 - \sigma_{daily}) \]

3. Volatility-Adjusted Rotation ($V_{rot}$)

The DynamicPositionSizer adjusts trade magnitude based on the ATR (Average True Range) of the target ticker to maintain a constant risk-per-trade.

\[ V_{rot} = \frac{K_{equity} \times Risk\%}{ATR_{20} \times Multiplier} \]