The bid-ask bounce: why a flat market looks like it is trembling
Posted: Sat Oct 31, 2026 9:41 am
Imagine a market where the true price doesn't move at all. Buyers and sellers arrive randomly. Buyers pay the ask, sellers hit the bid, and the last traded price alternates between the two.
On a chart of last trades, that looks like price jumping up and down, with a size equal to the spread, even though nothing fundamental happened. The effect has a name, bid-ask bounce, and it appears in every market with a spread.
It has two consequences that matter for short-term trading.
First, short-term returns show negative serial correlation. A tick up is more likely to be followed by a tick down, purely because of the bounce. A system that reads that as mean reversion may be seeing nothing more than the spread.
Second, the noise floor of a very fast chart is about the size of the spread. Moves smaller than the spread can't be traded profitably, and they carry almost no information. A one pip pattern on a market with a one pip spread is mostly bounce.
A decent habit is to measure signals against the spread. If an indicator fires on moves comparable to the spread, it's detecting the cost of trading, not a trend.
Using the mid price, the average of bid and ask, removes most of the bounce from the picture. Plotting mid instead of bid gives a visibly cleaner chart on thin pairs.
That also helps in backtesting, where bid-only data can create fake mean reversion that disappears when the ask side is included.
A strategy that works on bid-only data but fails with realistic bid and ask fills is a common story, and bounce is often the culprit.
Has switching a chart or a backtest from the bid to the mid price changed what a strategy appeared to do?
On a chart of last trades, that looks like price jumping up and down, with a size equal to the spread, even though nothing fundamental happened. The effect has a name, bid-ask bounce, and it appears in every market with a spread.
It has two consequences that matter for short-term trading.
First, short-term returns show negative serial correlation. A tick up is more likely to be followed by a tick down, purely because of the bounce. A system that reads that as mean reversion may be seeing nothing more than the spread.
Second, the noise floor of a very fast chart is about the size of the spread. Moves smaller than the spread can't be traded profitably, and they carry almost no information. A one pip pattern on a market with a one pip spread is mostly bounce.
A decent habit is to measure signals against the spread. If an indicator fires on moves comparable to the spread, it's detecting the cost of trading, not a trend.
Using the mid price, the average of bid and ask, removes most of the bounce from the picture. Plotting mid instead of bid gives a visibly cleaner chart on thin pairs.
That also helps in backtesting, where bid-only data can create fake mean reversion that disappears when the ask side is included.
A strategy that works on bid-only data but fails with realistic bid and ask fills is a common story, and bounce is often the culprit.
Has switching a chart or a backtest from the bid to the mid price changed what a strategy appeared to do?