In the latest Blackarbs research post, I dug into a question that influences a lot of systematic trading results:
What objective metric should your optimizer actually target?
Using my default leveraged ETF long-only strategy in Alpha Lab, I ran walkforward optimization experiments across common targets: Sharpe, Calmar, Gain-to-Pain, Omega, Total Return, VI Composite, and Geometric Martin Ratio itself.
The key distinction was that I did not judge objectives by how impressive they looked in sample. I judged them by real out-of-sample path quality, using Geometric Martin Ratio as the scoring rule because it rewards compounding while penalizing painful drawdown paths.
Out-of-sample path-quality results by optimizer objective.
The result
The surprising part: optimizing directly for the judging metric did not win overall.
Sharpe Ratio and VI Composite emerged as the strongest, most stable choices across the periods I tested. They were more predictive and consistent than simply optimizing for Geometric Martin Ratio itself.
That is the practical lesson: the best backtest metric is not automatically the best optimizer target. Optimization should prioritize objectives that reliably predict strong future behavior, not just objectives that produce flashy historical scores.
Read the full research post on the BlackArbs blog for the complete experiment, charts, and stability analysis.
You can also watch the walkthrough video for a visual tour of the setup and results.
Best,
Brian Christopher, CFA
BlackArbs LLC