RAR, R-Cubed, and Robust Sharpe Metrics for Strategy Backtests

Summary

This VeighNa community contribution describes additions to a backtesting statistics engine for evaluating strategies with regressed annual return (RAR), R-Cubed, and Robust Sharpe. RAR is calculated by regressing cumulative returns across time intervals and annualizing the result. The original R-Cubed measure uses RAR relative to the average of the largest drawdowns multiplied by their average duration, with an annualization factor. The author also tests a simplified variant that divides RAR by the average of the largest drawdowns alone. Robust Sharpe uses RAR over annualized return volatility.

The post includes implementation code and references optimized backtest curves, but the displayed evidence is limited and no numerical performance results are supplied. The author reports that the original R-Cubed optimization produced an unsatisfactory curve, while the simplified version appeared more reasonable; this is an observation, not validation that it generalizes. These metrics depend on drawdown definitions, sampling, and optimization choices, so strategies selected by them still require careful out-of-sample assessment.

Key ideas

• RAR estimates annual return by regressing cumulative returns against elapsed time.\nThe original R-Cubed denominator incorporates both the average of the largest drawdowns and their durations.\nA simplified R-Cubed version omits drawdown duration and uses only average drawdown magnitude.\nRobust Sharpe replaces conventional return in the numerator with RAR and scales by annualized volatility.\nThe reported optimization comparisons are qualitative and do not establish out-of-sample performance.

Tags

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.