Bid-only data distorts tests of any strategy that trades both sides
Posted: Sat Oct 31, 2026 6:38 am
Many historical data sets contain only bid prices. Testers then simulate the ask by adding a spread. For long trades, entries occur at the ask and exits at the bid, and for shorts the reverse.
If the data is bid-only and the spread assumption is wrong, errors build up systematically. Short trades often show better results than reality, because their stops and targets are tested against the bid when real triggers use the ask.
The distortion matters most with tight stops and targets. A 2 pip target on a short might be hit on the bid chart but not reached on the ask in the real market.
Using data that includes both bid and ask, or at least a realistic variable spread, removes the bias. Some data vendors provide separate series for each side.
Checking long and short performance separately is a useful diagnostic. A large unexplained gap between the two sides often points to a data issue.
Does the data used for testing include ask prices, and were long and short results compared?
If the data is bid-only and the spread assumption is wrong, errors build up systematically. Short trades often show better results than reality, because their stops and targets are tested against the bid when real triggers use the ask.
The distortion matters most with tight stops and targets. A 2 pip target on a short might be hit on the bid chart but not reached on the ask in the real market.
Using data that includes both bid and ask, or at least a realistic variable spread, removes the bias. Some data vendors provide separate series for each side.
Checking long and short performance separately is a useful diagnostic. A large unexplained gap between the two sides often points to a data issue.
Does the data used for testing include ask prices, and were long and short results compared?