Systematic Equity Factor Research
An attractive 28-stock backtest weakened when the universe broadened and concentration fell. Fresh data was still positive, but not statistically conclusive.
Central findingStronger methodology lowered the headline return and showed that positive holdout performance was not enough to establish reliable alpha.
Evidence
The result in context
- Original concentrated CAGR
- 32.41%
- Broad-development CAGR
- 17.46%
- Frozen-confirmation CAGR
- 22.73%
- Holdout alpha
- 3.25%
- HAC t-statistic 0.47; statistically inconclusive.
Question
What happens to an attractive concentrated momentum backtest when breadth, concentration, costs and genuinely unseen data are challenged?
A concentrated momentum result rebuilt with broader coverage, transaction costs, concentration diagnostics and a frozen confirmation period.
Original result
A strong return with hidden concentration
The original fixed 28-stock, momentum-weighted strategy produced 32.41% CAGR, a 1.33 Sharpe ratio and a −31.35% maximum drawdown. Sharpe relates excess return to volatility; maximum drawdown is the deepest peak-to-trough loss.
A robust rerun reproduced 32.41% CAGR with a lower 1.24 Sharpe, six average holdings and 84.23% in the top five. Those concentration diagnostics are not presented as the identical archived 1.33-Sharpe run.
Broad development
More breadth, costs and harder benchmarks
Across 2015–2024, the broader strategy produced 17.46% CAGR and 0.91 Sharpe with 92.8 average holdings. At 20 basis points of cost, CAGR fell to 16.36%.
Development alpha versus SPY was 4.04%, but its HAC t-statistic was 1.50 and the confidence interval included zero.
Robustness explorer
The headline falls as the test improves
Lower return is not weaker research when it comes from broader, harder evidence.
- CAGR
- 32.41%
- Sharpe
- 1.24
- Maximum drawdown
- −31.36%
- Holdings
- 6.0 average
- Concentration
- 84.23% top five
- Benchmark context
- Fixed 28-stock universe
The strongest headline, and the weakest breadth.
Frozen confirmation
Positive performance, inconclusive evidence
From 2 January 2025 to 24 August 2026, the frozen strategy produced 22.73% CAGR and 1.11 Sharpe. It beat SPY and broad equal weight on return, but not broad equal weight on Sharpe or drawdown.
Holdout alpha was 3.25% with a HAC t-statistic of 0.47. The right conclusion is positive but inconclusive.
Limitations
What this evidence does not establish
- The development universe is survivorship-biased because a 2024 S&P 500 snapshot was applied backwards.
- The frozen confirmation period is short and cannot establish a stable long-run premium.
Source and reproducibility
Trace the evidence
Source code, evaluation outputs and supporting material are available in the repository.
View repository- Concentration analysisoutputs/concentration.pngCommit / evidence ID: 2be4951c073fd8cafa7f20d59b6741f410f85d34
- Frozen confirmationoutputs/confirmation_benchmark_relative_performance.pngCommit / evidence ID: 2be4951c073fd8cafa7f20d59b6741f410f85d34