Median slippage tells a different story than mean slippage
Posted: Sat Oct 31, 2026 7:51 pm
Median slippage tells a different story than mean slippage
Average slippage is a convenient number, and a misleading one. A handful of terrible fills can drag the mean far from what typical trades experience.
Suppose 95 trades have zero slippage and 5 have 4 pips against. The mean is 0.2 pips. The median is zero. Both numbers are true, and each answers a different question.
The median describes the typical experience. The mean captures total cost. For budgeting, the mean matters. For understanding execution quality, the median and the tail matter.
The tail also tends to cluster around specific events: news, rollover, thin hours. Splitting the data by time shows where the damage comes from.
For a strategy that cannot afford big losses, the tail is the key. A bad fill on a stop is the kind of event that turns a good month into a break-even one.
At the 90th and 99th percentile, what do the slippage numbers look like in the logs?
Average slippage is a convenient number, and a misleading one. A handful of terrible fills can drag the mean far from what typical trades experience.
Suppose 95 trades have zero slippage and 5 have 4 pips against. The mean is 0.2 pips. The median is zero. Both numbers are true, and each answers a different question.
The median describes the typical experience. The mean captures total cost. For budgeting, the mean matters. For understanding execution quality, the median and the tail matter.
The tail also tends to cluster around specific events: news, rollover, thin hours. Splitting the data by time shows where the damage comes from.
For a strategy that cannot afford big losses, the tail is the key. A bad fill on a stop is the kind of event that turns a good month into a break-even one.
At the 90th and 99th percentile, what do the slippage numbers look like in the logs?