037 - Kevin Davey Part II - Selecting Optimal Strategies for Peak Performance

From Strategy to Portfolio

In Part II with Kevin, he delves into the intricate mechanics behind his systematic futures trading approach, offering advanced quantitative traders a window into the finer points of strategy design, walk forward analysis, robustness testing, and portfolio construction. Drawing on decades of experience and a background in aerospace, he emphasizes practical best practices, highlights common pitfalls in back-testing software, and outlines a disciplined monthly routine for maintaining and evolving a diversified intraday futures portfolio.

Strategy Design Principles
Kevin’s strategy development process begins with clearly defined goals—establishing the desired return and acceptable drawdown before writing a single line of code. He has learned through trial and error that conflating objectives (e.g., maximizing return without regard for survivability) can lead to ruin. His workshop curricula cover these foundational steps, teaching traders to:
• Set realistic performance targets
• Design strategies around specific market behaviours, whether trend-following or mean-reverting
• Avoid over-optimization by preferring one comprehensive test over multiple iterative tweaks

Kevin stresses that the coding platform (TradeStation, Python, etc.) is secondary to the robustness of the methodology itself; traders must focus on building sound rules before seeking the perfect back-testing engine.

Walk Forward Analysis: Best Practices and Common Mistakes
Kevin distinguishes his approach from other practitioners by emphasizing a one-shot walk forward process: optimize over an “in-sample” period, test over an “out-of-sample” period, and never revisit the in-sample parameters. Repeated retesting qualifies as overfitting and undermines the validity of out-of-sample results. Key takeaways include:
• Single-pass testing: Conduct one optimization and one validation, resisting the urge to tweak parameters multiple times.
• Beware of “cluster analysis” traps: Some platforms (e.g., TradeStation’s early implementation) automatically scan parameter matrices to find a “sweet spot.” Validating that same sweet spot on identical data simply reaffirms the initial optimization, rather than offering genuine out-of-sample confirmation.
Through this disciplined framework, Kevin seeks to minimize the impact of data mining bias and ensure that performance metrics truly reflect a strategy’s forward-looking potential.

Robustness Testing Beyond Walk Forward
Beyond walk forward, Kevin employs several layers of synthetic and real-time validation to gauge strategy resilience:
1. Logic based models relying on understandable causation not machine learning
2. Monte Carlo simulations on equity curves to assess variability in returns and drawdowns across randomized sequences of trades.
3. Real-time paper trading (often for six to nine months) to confirm live market behaviour aligns with backtest projections.
4. Equity‐curve inspection: He examines curve shapes for alarming features (e.g., overly smooth or consistently linear growth) as signs of over-optimization.
Kevin acknowledges that no amount of robustness testing can eliminate all luck components in backtests; instead, he focuses on understanding and quantifying the role of chance, then building portfolios to mitigate adverse outcomes.

Tech Stack and Automation Tools
Kevin’s core toolkit comprises:
• TradeStation for strategy coding and initial back-testing
• Multiwalk (a specialized walk forward add-on developed by a former student) for rapid batch testing across markets and parameter configurations (only available to Kevin's workshop students unfortunately)
• Excel/VBA for customized Monte Carlo simulations, cluster analyses, and portfolio-level data aggregation
While he concedes that Python and other programming environments offer powerful alternatives, Kevin prefers sticking with familiar platforms to avoid the distraction of constant tool development. His guiding philosophy: “Focus on strategy, not on building the backtest engine.”

Portfolio Construction Process
Combining intraday futures strategies into a cohesive portfolio involves several sequential steps:
1. Correlation screening: Eliminate strategies with historically high inter-correlations, ensuring that each component contributes unique risk factors.
2. Preliminary selection: From a universe of ~200 strategies, rank candidates by recent performance trends and drawdown characteristics over rolling 6–12-month windows.
3. Sector-balanced diversification: Formulate a weight assignment rule (e.g., equal risk contribution) to allocate equal portions of the portfolio to 7 broad sectors (such as rates, equities, softs, metals, etc)
4. Position sizing: Determine the exposure of individual strategies to ensure diversification and equal-risk allocation is maintained across the portfolio
5. Real-time pilot: Trade the proposed portfolio in live simulation to observe actual behaviour before full deployment.
This “one-and-done” philosophy in portfolio testing parallels his walk forward rules: avoid iterative tuning on the final test to preserve genuine out-of-sample validity.

Monthly Maintenance and Rebalancing
Kevin dedicates the final trading day of each month to portfolio review and rebalancing:
• Update performance metrics for all strategies in the library (~200 in total).
• Select top performers (typically 20–30 strategies) and drop underperformers, aiming for sector diversification—ideally seven market sectors at roughly 14.3% allocation each.
• Conduct Monte Carlo on the chosen set to estimate potential worst-case drawdowns and compare realized returns against simulated expectations.
• Re-optimize individual strategies: He typically has 20–40 strategies per month requiring parameter updates via walk forward, scheduling optimizations across trading days.
This disciplined cycle ensures that the portfolio remains attuned to evolving market dynamics without succumbing to short-term noise.

Risk Management and Psychological Preparedness
A recurring theme in Kevin’s methodology is the interplay between statistical rigor and emotional resilience:
• Drawdown acceptance: Traders often overestimate their drawdown tolerance; real-time experience reveals that technical ability dissipates under actual equity drops. Kevin advises sizing strategies to limit maximum drawdowns to levels half of one’s perceived comfort zone.
• Monitoring and supervision: Regulatory requirements demand active oversight of automated systems. Kevin cautions that overseeing too many strategies increases logistical burden—arguing for a “sweet spot” of 20–30 algos for manageability.
By intertwining quantitative controls with human oversight, he fosters longevity in trading careers rather than short-lived performance peaks.

Performance Benchmarks and Goals
Kevin benchmarks his goals against institutional players (e.g., commodity trading advisors tracked by Barclay’s Hedge):
• Proprietary funds: the best of the best typically average 10–20% annual returns with matching drawdowns; they regard a 1:1 return-to-drawdown ratio as prudent.
• Kevin’s personal target: 50% annual return with a maximum 25% drawdown, acknowledging that some years meet both criteria while others fall short. That's a pretty high expected draw down from my perspective.
• Extreme contest performance (100%+ returns) involves high leverage and unacceptable risk for long-term trading; such feats, while impressive, clash with sustainable risk thresholds.
These benchmarks guide strategy design, ensuring that return objectives align with survivability imperatives.

Conclusion
Kevin’s systematic approach melds rigorous quantitative testing with pragmatic risk management and monthly maintenance protocols. By enforcing single-pass optimizations, extensive real-time validation, and lean portfolio sizes, he constructs a robust trading framework designed for consistency and longevity. Advanced traders can draw from his workshop principles to refine strategy design, navigate common back-testing pitfalls, and build diversified, adaptive portfolios capable of weathering market uncertainties.