A comprehensive overview of our proprietary algorithmic trading system powered by custom built AI and multi-indicator composite signals for quantitative traders and developers.
Strategy Evolution & Current Architecture
Legacy System (2018-2019)
JSON configuration setups
Ticker-specific stochastic parameters
Manual fast/slow range controls
Individual time interval settings
Current Deployment (Nov 2024)
Custom AI built in PyTorch
GPU-accelerated processing
Autonomous whitelist generation
Dynamic entry/exit decisions
The newest R&D version represents a complete paradigm shift from manual configuration to fully autonomous AI-driven decision making, eliminating human bias in parameter selection.
DMF Core Concept
Directional Momentum Flux (DMF) is a compound indicator engineered to surface projected changes in directional momentum before they occur. Unlike traditional lagging indicators, DMF focuses on momentum inflection point prediction.
The system combines high-level technical indicators:
Double Exponential Moving Average Formula (DEMA)
Relative Strength Index (Double RSI; Fast & Slow, Variable)
Composite Commodity Channel Index (CCI+; Two Series)
Volume-Weighted Balance of Power (VWAP)
That also includes lower-level momentum signals:
Balance of Power (BOP)
Percentage Price Oscillator (PPO; Split, Two Series)
Range-adjusted Momentum Oscillator (RMOMO, our fancy MOM variant) in addition to some secret sauce and proprietary algorithmic enhancements developed in-house.
Extensive backtesting across multiple market cycles and randomized asset pools validates its cross-asset class effectiveness beyond simple bull market performance.
Multi-Indicator Architecture
Moving Average Composite
Double Exponential Moving Average calculations
Composite Simple Moving Average integration
Trend direction validation
Momentum Oscillators
Double Relative Strength Index (fast & slow, variable)
Composite Stochastic Relative Strength Index (dual series)
Range-adjusted Momentum Oscillator
Volume & Price Analysis
Volume-Weighted Balance of Power
Volume-Weighted Average Price integration
Percentage Price Oscillator split
Channel Analysis
Composite Commodity Channel Index (dual series)
Adaptive channel boundaries
Breakout confirmation signals
Signal Generation Framework
Two-Tier Signal System
Vertical Bands (VBs): Primary entry/exit windows that remain active while top-level signals show sufficiently strong BUY conditions. These semi-transparent overlays on candlestick charts indicate optimal trading windows requiring appropriate risk mitigation strategies.
Eggs: Validation signals that confirm conditions indicated by VBs. While a lit VB without an egg remains valid, waiting for egg confirmation can improve performance at the cost of potentially missing optimal entry points.
A VB followed immediately by an egg represents the highest confidence signal combination.
Multi-Strategy Portfolio
DMF (Primary)
Directional momentum flux algorithm with custom momentum and flux calculations. Optimized for volatility-driven markets with proven performance across multiple asset classes.
Trendy
Trend-focused algorithm designed for sustained directional moves. Complements DMF by capturing longer-term momentum when market conditions favor trend-following strategies.
Hercules
Market maker algorithm providing liquidity and capturing bid-ask spreads. Deployed when market conditions require liquidity provision rather than directional positioning.
Crabby
Bollinger Bands-based algorithm optimized for sideways, range-bound markets. Captures mean-reversion opportunities when volatility is contained within established ranges.
Live Trading Performance
Strategy has been actively traded since 2018-2019, with current deployment running since November 2024. Performance validated by registered broker-dealers with 35+ years industry experience.
Recent Performance Analysis
21.89%
Best Single Day
August 7th with 118 trades executed
3.5
Average Hold Period
Hours per trade, optimizing for quick momentum capture
18.4%
Peak Performance
August 8th with 86 successful trades
118
Max Daily Trades
August 7th achieving 21.89% returns
The strategy demonstrates consistent profitability with occasional drawdown days, maintaining an aggressive trade frequency that capitalizes on short-term momentum opportunities.
Market Regime Performance
Cross-Cycle Validation
High Volatility: DMF performs optimally during volatile conditions, as volatility creates the momentum flux opportunities the algorithm is designed to capture.
Range-Bound Markets: Crabby algorithm deployment provides mean-reversion strategies when directional momentum is limited.
The multi-algorithm approach ensures performance consistency across varying market conditions through strategic algorithm selection based on real-time market regime detection.
Exchange Infrastructure
Current deployment supports 94 exchanges, providing unprecedented market access and liquidity options. This multi-exchange architecture eliminates single-point-of-failure risks and optimizes execution across global markets.
Maximum estimated capacity: ~$20M per exchange before liquidity impact
Machine Learning Integration
01
Base Model Classes
BaseRegressionModel, BaseClassifierModel, and PyTorch implementations provide the foundation for custom AI development.
02
Reinforcement Learning
PPO, A2C, DQN, TRPO, and RecurrentPPO models enable adaptive strategy optimization based on market feedback.
03
Prediction Models
XGBoostClassifier and CatboostClassifier enhance signal generation accuracy through ensemble learning approaches.
04
Model Orchestration
Hierarchical and parallel model execution with dynamic selection based on real-time market condition assessment.
Performance Attribution Framework
Trade-Level Metrics
Entry/exit reason tracking
Profit/loss ratio analysis
Trade duration optimization
Risk-adjusted return calculation
Factor Analysis
Momentum factor decomposition
Balance of Power measurement
Stochastic momentum tracking
Trend strength quantification
The MultiMetricHyperOptLoss framework considers total profit, profit factor ratios, expectancy calculations, win rate coefficients, drawdown adjustments, and trade count penalties for comprehensive performance evaluation.
Risk Management Architecture
Position-Level Risk
Individual position monitoring with dynamic stop-loss adjustments based on volatility and momentum indicators.
Portfolio Risk
Maximum drawdown analysis, correlation monitoring, and exposure limits across all active positions.
Market Risk
Real-time market regime detection with automatic strategy switching during adverse conditions.
Execution & Infrastructure
Systematic Execution Framework
Signal Generation: Fully systematic based on DMF composite indicators and machine learning model outputs.
Order Execution: Automated execution across multiple exchanges with intelligent routing for optimal fill rates.
Infrastructure Requirements: It can run on EC2 instances, however, it is recommended to use GPU-accelerated PyTorch models, multi-exchange API connectivity, real-time data feeds, and redundant system architecture.
Average trade duration of 3.5 hours requires high-frequency monitoring and rapid execution capabilities with minimal latency tolerance.
Fee Structure & Cost Analysis
50%
Maker Orders
Providing liquidity to order books
50%
Taker Orders
Consuming existing market liquidity
Fee tracking separates maker and taker orders with conservative projections using higher fee calculations. The system calculates actual fees based on executed trades rather than estimates, providing accurate performance attribution net of transaction costs.
Alpha Decay Measurement
Current Measurement Mechanisms
Periodic Performance Breakdown: Daily statistics generation with rolling window metrics comparing recent versus historical performance across multiple timeframes.
Multi-Metric Decay Tracking: Trend strength analysis using linear regression slopes and volatility ratio calculations to identify signal degradation.
Market Regime Detection: Continuous monitoring of market condition changes that could impact alpha generation effectiveness.
While formal alpha decay expiry times are not implemented, the system includes mechanisms for detecting performance degradation and triggering strategy adjustments.
Portfolio Construction & Exposure
Concentration Limits
Instrument concentration managed through dynamic position sizing algorithms. Portfolio composition varies based on market opportunities and risk parameters.
Maximum position sizes determined by liquidity analysis and volatility-adjusted risk calculations.
No fixed maximum net or gross exposure limits - determined dynamically by market conditions and available opportunities.
Leverage Decisions
Signal strength assessment
Market volatility analysis
Portfolio correlation review
Risk-adjusted return expectations
Market Neutrality & Directional Risk
The strategy is not market neutral by design. DMF specifically seeks directional momentum opportunities, taking calculated directional risks based on momentum flux predictions.
Net exposure changes from long to short occur frequently based on signal generation - the system adapts to market conditions rather than maintaining artificial neutrality.
Position sizing and leverage decisions are made on a trade-by-trade basis using volatility-adjusted risk parameters and expected return calculations.
Stress Testing & Risk Scenarios
1
Historical Backtesting
Extensive testing across multiple market cycles including bear markets, not just bull runs, using randomized asset pools.
2
Volatility Stress Tests
Algorithm performance validated under extreme volatility conditions, with risk management systems tested for rapid drawdown scenarios.
3
Liquidity Risk Assessment
Multi-exchange architecture provides liquidity risk mitigation, with capacity limits established per exchange.
Drawdown Management Protocol
Current Challenge: 10% portfolio-to-terminal drawdown limit requires active monitoring and response protocols.
Proposed Modifications: Implementation of hanging position removal system since average win rate occurs within 3.5 hours. This reduces overnight risk and capital tie-up during extended drawdown periods.
Gross exposure adjustments during drawdowns would include: reduced position sizing, increased selectivity in signal confirmation, and potential temporary algorithm switching to more conservative strategies like Hercules market-making.
Infrastructure & Monitoring
Real-Time Alerts
Drawdown and exposure monitoring with immediate notifications for risk threshold breaches.
Redundant Systems
Multiple exchange connectivity ensures continued operation during single-exchange outages or connectivity issues.
GPU Computing
PyTorch models require GPU acceleration for real-time signal generation and position management.
Strategy Enhancement Roadmap
1
Q1 2025: Factor Attribution
Implementation of formal factor decomposition framework for value and carry factor analysis beyond current momentum focus.
2
Q2 2025: Alpha Decay Module
Development of dedicated alpha decay measurement system with signal expiry tracking and automated model refresh protocols.
3
Q3 2025: Enhanced Risk Management
Integration of advanced stress testing frameworks and real-time portfolio risk analytics with automated exposure adjustments.
Competitive Advantages
Technical Edge
Custom PyTorch AI implementation
94-exchange connectivity
Multi-algorithm portfolio approach
Real-time regime detection
Operational Excellence
6+ years live trading experience
Systematic signal generation
Rapid 3.5-hour average hold times
Continuous strategy optimization
Performance Validation
Strategy validation through multiple independent sources provides confidence in systematic performance:
Broker-Dealer Review
35+ year industry veterans have chartered and validated mathematical foundations, confirming sound principles that have worked historically and continue to perform.
Multi-Cycle Testing
Backtesting across various market conditions, not limited to bull markets, using randomized asset selection for unbiased validation.
Live Performance
Six years of live trading data demonstrating consistent methodology application across changing market environments.
Next Steps & Partnership Opportunities
Implementation Pathways
API Integration: Separate deployment with provided keys for performance verification and real-time monitoring capabilities.
Platform Integration: Direct integration with your existing infrastructure for seamless strategy deployment and management.
Custom Development: Tailored implementations to meet specific institutional requirements while maintaining core DMF methodology.
Ready for immediate deployment with comprehensive documentation, ongoing support, and continuous enhancement commitment.