Average spread is a number from a brochure, entry spread is the one you pay
Posted: Sat Oct 31, 2026 7:49 am
Brokers publish averages. Averages are computed across every tick of a day, including the quiet hours when everything is tight. If a strategy fires mostly at 13:30 GMT, 14:30 GMT and the first fifteen minutes of London, the spread actually paid can be noticeably different from the advertised figure.
A quick audit shows how big the gap is. Export your trade history with entry timestamps, and for each trade record the spread at the moment of entry, not the daily mean. Most platforms store spread per bar in history data, which is crude but usable. Then compute:
- Mean entry spread versus mean daily spread
- The share of trades entered with spread above 1.5 times the pair's typical value
- Net result of that high-spread bucket compared with the rest
It is common for a small share of trades, say one in six, to carry most of the excess cost. When that is true the fix is not a new strategy, just a spread filter that skips entries when the current spread exceeds a threshold derived from your own data.
A cap like "no entry if spread is more than 1.2 times the rolling 20-minute median" is easy to code and stops the worst fills without reducing trade count much.
What entry spread versus advertised average did your own audit show, and was the gap bigger than expected?
A quick audit shows how big the gap is. Export your trade history with entry timestamps, and for each trade record the spread at the moment of entry, not the daily mean. Most platforms store spread per bar in history data, which is crude but usable. Then compute:
- Mean entry spread versus mean daily spread
- The share of trades entered with spread above 1.5 times the pair's typical value
- Net result of that high-spread bucket compared with the rest
It is common for a small share of trades, say one in six, to carry most of the excess cost. When that is true the fix is not a new strategy, just a spread filter that skips entries when the current spread exceeds a threshold derived from your own data.
A cap like "no entry if spread is more than 1.2 times the rolling 20-minute median" is easy to code and stops the worst fills without reducing trade count much.
What entry spread versus advertised average did your own audit show, and was the gap bigger than expected?