Trade like Mulvaney: how Concretum Research reverse-engineered a systematic CTA

# What Concretum tried to do Concretum Research set out to reverse-engineer the trading rules behind Mulvaney Capital Management's Global Diversified Program by combining public clues (interviews, presentations and commentaries) with systematic experiments. They treat the task as a parameter-search problem: translate the qualitative descriptions of Mulvaney's process into concrete rule choices, then test many combinations and see which synthetic programs best explain MCM's monthly returns.

# The starting clues

# How the experiment was run

# Best-fit design features The top-fitting synthetic programs consistently share several concrete features:

• Donchian breakout with a long lookback roughly equal to six months (126 trading days).
• Symmetric treatment of long and short signals.
• Execution on the same or next trading day (Concretum reports execution within 1–2 days of the signal).
• Equal risk allocation across the traded markets rather than weighting by sector.
• Pyramiding: an initial smaller allocation with additional contracts added at predefined profit thresholds as trends develop.
• Two-layer stopping: a closer fixed initial stop and a trailing exit modeled by the Donchian midline, which moves outward as channel width expands (Concretum treats the midline as a volatility-accommodating trailing stop).

# How well the replicas matched MCM The best-fit synthetic CTAs achieved R² in the range 0.71–0.73, improving explanatory power by about 45% versus the SG Trend Index. Correlations for the top fits sat tightly between 0.84 and 0.85. Concretum's comparison uses pre-fee synthetic results against Mulvaney's post-fee series, and the group presents top-ten strategy fits and pre-fee versus post-fee performance charts.

# Practical takeaways for traders and researchers If you want to approximate Mulvaney-style trend following, start with a long-duration Donchian breakout (around 126 days), size risk equally across roughly 40 traded futures, use a small initial stake and pyramid into winners, and manage exits with a volatility-aware trailing stop rather than fixed profit targets. When testing hypotheses against an existing manager's track record, match volatility before comparing returns and explore a wide parameter grid to find robust fits.