Filter by the benchmark before averaging
The upside subset contains rows where benchmark return is strictly positive; the downside subset contains rows where it is strictly negative. For either subset, capture equals 100 × mean(strategy returns)/mean(benchmark returns). Benchmark-zero rows enter neither calculation, regardless of the strategy return on those rows. This tool uses arithmetic conditional means. Some published capture statistics use geometric returns instead, so match the convention before comparing values taken from different reporting systems.
Calculate the two directions independently
Suppose benchmark returns are 2%, 4%, −2%, −4%, and 0%, with matching strategy returns of 3%, 5%, −1%, −3%, and 9%. In the up subset, strategy mean is 4% and benchmark mean is 3%, giving upside capture of about 133.33%. In the down subset, the means are −2% and −3%, giving downside capture of about 66.67%. The strategy's final 9% observation is excluded because the benchmark was flat.
A negative downside ratio needs careful reading
When the benchmark mean is negative but the strategy mean is positive in that subset, downside capture is negative: the strategy gained on average during those benchmark declines. This does not prevent losses during benchmark rallies or on excluded flat periods. A missing up or down subset yields an undefined ratio, not zero capture. Results can also depend strongly on one extreme observation. Review full-period performance, drawdowns, sample size, and benchmark relevance alongside both conditional ratios.