Upside and downside capture
Capture ratios split your benchmark relationship in two: how much of the benchmark's gains you captured on its good days, and how much of its losses you took on its bad ones.
On this page
Beta gives you one number for your relationship to the benchmark, averaged across every day. Capture ratios split that relationship into its two halves, because portfolios rarely behave the same way in both directions.
Upside capture is how much of the benchmark's gains you got on the days it rose. Downside capture is how much of its losses you took on the days it fell.
How it is calculated
Gylder sorts every day in the window by what the benchmark did, then sums both sides:
upside capture = Σ(your returns on benchmark up-days) / Σ(benchmark returns on those days)
downside capture = Σ(your returns on benchmark down-days) / Σ(benchmark returns on those days)
Days the benchmark was exactly flat fall into neither bucket.
How to read them
Both are expressed relative to 1, and the two are read in opposite directions.
Upside capture above 1 means you gained more than the benchmark on its good days. Below 1 means you lagged on the way up.
Downside capture below 1 means you lost less than the benchmark on its bad days. Above 1 means you fell harder.
The pair is more informative than either alone, because the shape of the combination describes the portfolio's character. High upside with high downside is amplification: you are essentially holding the benchmark with leverage. Low upside with low downside is damping. High upside with low downside is the asymmetry everyone wants and few sustain, and over a short window it is far more likely to be a sampling artefact than a property of the portfolio.
Two things to watch
The sign convention on the downside. Both sums are negative on down days, so the ratio comes out positive. A downside capture of 0.7 means you took 70% of the benchmark's losses, which is the favourable case, even though nothing about the number looks negative.
These are sums, not compounded returns. Gylder adds the daily returns within each bucket rather than chaining them. That is a simplification, and it means these figures should be read as a characterisation rather than as a precise attribution of performance.
As everywhere in the risk grid, the numbers depend on which benchmark you selected, and they recalculate when you change the window.
Was this helpful?