Bid-ask bounce makes tiny reversals look like signals
Posted: Sat Oct 31, 2026 6:44 am
Prices on a tick chart flicker between bid and ask as trades occur. A sequence of trades alternating between the two sides produces a zigzag pattern even when the true price has not moved at all.
That bounce creates negative serial correlation at the tick level. Up moves are followed by down moves not because sentiment changed, but because the last trade hit the ask and the next one hit the bid. Strategies based on one-second or tick patterns can pick this up and mistake it for mean reversion.
The issue appears most when the spread is large relative to price movement. If the spread is 1 pip and the true movement is 0.5 pip, the bounce dominates and every pattern looks like a reversal.
Testing on mid prices rather than bid or ask alone strips out the bounce and shows whether the pattern survives. If the edge disappears, it was a measurement artifact, and it would never have been tradeable after costs.
Has mid-price data been used to check a short-term signal, and did the edge hold up?
That bounce creates negative serial correlation at the tick level. Up moves are followed by down moves not because sentiment changed, but because the last trade hit the ask and the next one hit the bid. Strategies based on one-second or tick patterns can pick this up and mistake it for mean reversion.
The issue appears most when the spread is large relative to price movement. If the spread is 1 pip and the true movement is 0.5 pip, the bounce dominates and every pattern looks like a reversal.
Testing on mid prices rather than bid or ask alone strips out the bounce and shows whether the pattern survives. If the edge disappears, it was a measurement artifact, and it would never have been tradeable after costs.
Has mid-price data been used to check a short-term signal, and did the edge hold up?