Crypto’s 5% hourly drops: 57% of next hours rose, vs 49%
A 5% candle on the hourly chart pulls traders both ways: ride the momentum, or wait for it to turn back. We checked 50 Bybit coins over a year: which way did the next hour go?
What we measured
Bybit hourly candles for 50 coins, every hour from 16 October 2025 to 15 October 2026, in UTC: 437,950 coin-hours. A 5% hour is a close at least 5% above or below the previous hour’s close. Then we took the next hour’s move. To test luck, we shuffled the next hours across the hour slots 2,000 times, the same shuffle for every coin. We wrote the rules and the bar down before finding a single 5% hour.
What we found
There were 1,197 drops of 5% or more in an hour. The next hour rose 57.2% of the time, while any hour rose 48.6% of the time. Only 8 of the 2,000 shuffles did as well, and both halves of the year agree. It passed the bar we set.
One coin, AKE, made 527 of the 2,887 spikes, so we added two checks after that first run. Without AKE the lean held, and 25 of the 32 coins with 10 or more drops leaned the same way on their own.
The lean was small: the typical next hour was +0.42%, next to a 5% drop. And 42% of the time the next hour fell again.
The 1,690 rises of 5% or more said nothing: the next hour looked like any other hour. Bitcoin and BNB never moved 5% in an hour all year.
On the chart
AKE, the coin with the most 5% hours, from 6 to 9 October. The grey band runs 5% either side of the previous hour’s close, so a candle that ends outside it is a 5% hour. Three did on 8 October, each with a dot on its close.
The rise of 5.9% was followed by another 4.6%. The 6.4% drop was followed by +0.1%, and the 5.9% drop by a further -4.4%. One drop bounced a little and one kept going: a lean is not a rule.
What to do with it
After a 5% hourly drop, the next hour leaned up, but by little and not reliably. Treat it as a tilt in the odds, not a trade on its own.
What this does not tell you
Whether it holds next year, after costs, or on a coin’s first days. One year, one exchange, hourly closes. Past data.
Observations, not recommendations.


