Infrastructure for Systematic Strategy Development
NOVA is not a single trading strategy — it is infrastructure for the rapid, systematic development, testing, deployment, and refinement of trading strategies. Competitive advantage lies in the process of strategy creation itself, not in any single strategy.
Durable edge comes not from any particular model, but from the process by which strategies are systematically developed, tested, deployed, and refined.
Markets are non-stationary. Volatility regimes shift, correlations break down, and participant behavior evolves. Strategies calibrated to historical data degrade. NOVA systematizes the development cycle so that strategy turnover is continuous — multiple strategies run concurrently, and the portfolio engine shifts capital toward those performing well in current conditions.
Strategies, alpha models, risk models, and portfolio models are independent, reusable components that compose into new configurations.
Every state transition is recorded in an accumulate-only database. Any decision can be traced back to the data and model state that produced it.
Multiple strategies operate concurrently. The portfolio engine shifts allocation toward strategies performing well in current conditions.
Data flows through independent channels — market data, strategy signals, portfolio state, and execution events are decoupled via an internal event bus. Components subscribe to what they need and publish their outputs.
No single methodology dominates. Different approaches perform better in different markets and regimes. Running them concurrently provides diversification at the strategy level.
Filters, spectral analysis, wavelet decomposition — strategies built on mathematical transformations of price and volume data.
Supervised and unsupervised approaches including GRU networks, ensemble methods, and multifractal analysis features.
Explicit conditional logic derived from market microstructure, technical patterns, or fundamental triggers.
Signal processing feeding ML models, Bayesian priors constraining outputs, rules-based risk overlays attenuating signals.
NOVA runs strategies against historical data with the full platform stack — realistic portfolio allocation, execution simulation, and accounting. The same portfolio engine, executor, and ledger code that runs in production, with only external integration points swapped for mocks.
The system maintains a managed population of strategies at various lifecycle stages. Brier scoring measures each strategy's calibration quality — as believability declines, allocation shrinks proportionally.
Hypotheses from parameter search or AI-assisted ideation
Deployed and contributing to allocation. Continuously monitored.
Reduced allocation. Shadow-scored for potential reactivation.
Removed from active pool. Retained for analysis and resurrection.