One-sided ticks and what they might reveal about thin pair order flow
Posted: Tue Nov 03, 2026 12:10 am
One-sided ticks and what they might reveal about thin pair order flow
On thinly traded pairs, quotes sometimes update on only one side. The bid moves up while the ask stays put for several ticks, or the reverse. The spread narrows or widens as a result, and the mid price shifts in a way that does not reflect a balanced market.
Such patterns hint at an imbalance. A liquidity provider adjusting one side suggests a change in inventory or a read on incoming flow. Repeated one-sided updates in the same direction can precede a move in the following seconds.
The signal is noisy and depends on the feed. Aggregated retail feeds often smooth away such detail, but raw ECN data shows it more clearly. It is also easy to overinterpret, since single-side updates occur for mundane reasons like a provider refreshing a stale price.
To test, label each tick as bid-up, ask-up, bid-down, or ask-down, and check whether sequences of one type predict the next mid-price change. Require a sample of hundreds of thousands of ticks before believing the result.
Has anyone looked at which side updates first, and did it add any predictive value?
On thinly traded pairs, quotes sometimes update on only one side. The bid moves up while the ask stays put for several ticks, or the reverse. The spread narrows or widens as a result, and the mid price shifts in a way that does not reflect a balanced market.
Such patterns hint at an imbalance. A liquidity provider adjusting one side suggests a change in inventory or a read on incoming flow. Repeated one-sided updates in the same direction can precede a move in the following seconds.
The signal is noisy and depends on the feed. Aggregated retail feeds often smooth away such detail, but raw ECN data shows it more clearly. It is also easy to overinterpret, since single-side updates occur for mundane reasons like a provider refreshing a stale price.
To test, label each tick as bid-up, ask-up, bid-down, or ask-down, and check whether sequences of one type predict the next mid-price change. Require a sample of hundreds of thousands of ticks before believing the result.
Has anyone looked at which side updates first, and did it add any predictive value?