Multi-Factor Futures System

Single factors perform well only in specific market conditions. Trend-following excels in trends but fails in ranges; mean reversion excels in ranges but fails in trends. By combining factors, you create a system that performs adequately across all conditions rather than brilliantly in some and terribly in others. The goal is consistent, sustainable returns with lower drawdowns.

For this strategy, calculate at fixed times: 10:00 a.m. ET (after opening volatility settles) and 2:00 p.m. ET (afternoon session). This provides two decision points daily. Avoid constant recalculation which leads to overtrading and noise-chasing. Some traders add an end-of-day calculation for next-day positioning. Consistency in timing is more important than frequency.

When composite score is near zero, factors are conflicting or all showing neutral readings. This is valuable information - the correct action is no position. Staying flat when signals are unclear preserves capital for high-conviction opportunities. Many traders lose money by forcing trades when signals aren't present. Zero score = wait for clarity.

Start with equal weights (25% each) to understand how each factor contributes. After 3-6 months of trading and tracking, you'll have data showing which factors contributed most to profits in different conditions. Then adjust weights based on this data, not intuition. Keep a detailed journal tracking factor scores and outcomes. Let data guide weight adjustments.

It works for both. For day trading: use hourly factor calculations, close positions daily. For swing trading: use daily factor calculations, hold positions 1-5 days. The core logic remains the same - combine factors, size by conviction, exit on score reversal or targets. Adjust timeframes of factor indicators to match your trading horizon.

Detect regime using ADX (trending vs ranging) and VIXC (low vs high volatility). When regime changes: 1) Adjust factor weights toward factors that perform better in new regime. 2) Reduce overall position size during transition (uncertain period). 3) Wait for new regime to be confirmed before full commitment. Transition periods are highest-risk; established regimes are easier to trade.

Yes, typically higher timeframes get more weight for direction, lower for timing. Example weighting: Weekly factor direction: veto power (if weekly bearish, don't go long regardless of daily). Daily factors: 70% weight for signal generation. Hourly factors: 30% weight for timing refinement. This hierarchy respects market structure while allowing tactical timing.

Key considerations: 1) Sufficient data (5+ years covering different regimes). 2) Walk-forward testing (optimize on window 1, test on window 2, roll forward). 3) Include realistic costs (slippage, commissions). 4) Test factor stability (do factors maintain predictive power?). 5) Stress test on crisis periods. 6) Check for overfitting (are you curve-fitting or finding real relationships?). Robust backtesting is essential before live trading.

Generally, avoid discretionary overrides - they defeat the purpose of systematic trading. However, consider override for: 1) Major news events model can't incorporate (e.g., a surprise Bank of Canada rate decision or commodity shock). 2) System clearly broken (data feed issues). 3) Unprecedented market conditions. Document every override and review outcomes. If you're overriding frequently, either the model needs improvement or you're not suited for systematic trading.

Process: 1) Research and define the new factor with economic rationale. 2) Backtest independently - does it have predictive power? 3) Test correlation with existing factors - too correlated = redundant. 4) Paper trade alongside live model for 2-3 months. 5) If validated, integrate with small initial weight. 6) Adjust weights after 6 months of live performance data. Never add factors without rigorous testing.

Methods: 1) Rolling window optimization: optimize weights on rolling 252-day window, apply to next period. 2) Regime-conditioned: maintain separate weight sets for each regime, switch based on detected regime. 3) Bayesian updating: start with prior weights, update based on recent factor performance using Bayesian methods. 4) Machine learning: train model to predict optimal weights based on market features. All methods require guard rails (weight bounds) to prevent extreme allocations.

Critical risks: 1) Factor crowding: if everyone uses same factors, edge erodes. 2) Regime dependence: factors that work in one regime fail in others. 3) Correlation regime: in crises, all factors correlate, destroying diversification. 4) Model overfitting: backtested performance may not persist. 5) Factor decay: factors can lose predictive power over time. 6) Execution slippage: in the relatively thinner Canadian futures market, systematic strategies can face execution challenges away from SXF/SCF. Mitigate through diversification, robust testing, and continuous monitoring.

Detection methods: 1) Rolling Sharpe degradation: if factor Sharpe ratio declining over 12+ months, potential decay. 2) Correlation increase: if factor correlations with other factors increasing, crowding may be occurring. 3) Signal frequency change: if factor generates more signals but worse outcomes, likely crowded. 4) Academic publication: if factor appears in research papers, expect crowding in 1-2 years. 5) Performance attribution: if factor P&L contribution declining without regime explanation. Regular monitoring essential.

Comprehensive stress testing: 1) Historical scenarios: replay 2008 crisis, 2020 COVID, 2022 rate hiking on current portfolio. 2) Hypothetical scenarios: simulate factor crashes, correlation spikes, liquidity events. 3) Sensitivity analysis: what if key factor fails completely? 4) Tail risk: calculate expected shortfall (CVaR), not just VaR. 5) Factor correlation stress: model factor correlations going to 0.8+ in crisis. 6) Drawdown analysis: calculate time to recovery under various scenarios. Update stress tests quarterly.

Critical documentation: 1) Factor definitions: exact calculation methodology, data sources. 2) Economic rationale: why each factor should have predictive power. 3) Backtest results: performance statistics, regime analysis. 4) Weight optimization: methodology and current weights. 5) Risk parameters: position limits, stop rules, correlation limits. 6) Review history: changes made, reasons, outcomes. 7) Code repository: version-controlled codebase. 8) Operational procedures: daily/weekly/monthly processes. Documentation enables continuity and systematic improvement.