Rollover spreads aren't just wider, the quotes themselves get weird
Posted: Sat Oct 31, 2026 10:45 pm
Everybody knows spreads blow out around 5pm New York. What gets talked about far less is what the quotes themselves do during that window.
Some feeds go quiet for a few seconds and then print a jump that looks like a gap on a one-minute chart. Others keep ticking, but only one side moves. The bid drops while the ask barely budges, so the spread doubles without the mid price changing at all.
That matters because plenty of scalping setups key off the bid chart. A "breakout" printed at 5:00:03 is often just the bid catching down to a repriced ask. No buyer or seller showed up. The chart simply reflects the broker's quoting engine rearranging itself.
Indicators fed by that data will happily fire signals on it. A stochastic cross or a Bollinger break during rollover is a measurement artifact more often than a real move.
Check the tick log for the window between 4:55 and 5:10 New York on a few different days. The pattern repeats almost to the second on some brokers.
A quick way to see it: export a day of tick data and sort the 4:55 to 5:10 New York window by timestamp. On some feeds, the bid column shows a staircase of lower values while the ask column barely changes for twenty or thirty seconds. On others, both sides freeze and then jump together. Those are two different quoting engines doing two different things, and they produce different false signals.
The Bollinger Band squeeze that "breaks" at 5:00, the moving average cross at 5:02, and the RSI spike all inherit the artifact. A filter that simply ignores signals inside a short window around rollover removes a surprising number of losing trades, and nothing else is lost, since fills in that window are poor anyway.
Does anyone else see the quote freeze, or does the feed just widen and keep running?
Some feeds go quiet for a few seconds and then print a jump that looks like a gap on a one-minute chart. Others keep ticking, but only one side moves. The bid drops while the ask barely budges, so the spread doubles without the mid price changing at all.
That matters because plenty of scalping setups key off the bid chart. A "breakout" printed at 5:00:03 is often just the bid catching down to a repriced ask. No buyer or seller showed up. The chart simply reflects the broker's quoting engine rearranging itself.
Indicators fed by that data will happily fire signals on it. A stochastic cross or a Bollinger break during rollover is a measurement artifact more often than a real move.
Check the tick log for the window between 4:55 and 5:10 New York on a few different days. The pattern repeats almost to the second on some brokers.
A quick way to see it: export a day of tick data and sort the 4:55 to 5:10 New York window by timestamp. On some feeds, the bid column shows a staircase of lower values while the ask column barely changes for twenty or thirty seconds. On others, both sides freeze and then jump together. Those are two different quoting engines doing two different things, and they produce different false signals.
The Bollinger Band squeeze that "breaks" at 5:00, the moving average cross at 5:02, and the RSI spike all inherit the artifact. A filter that simply ignores signals inside a short window around rollover removes a surprising number of losing trades, and nothing else is lost, since fills in that window are poor anyway.
Does anyone else see the quote freeze, or does the feed just widen and keep running?