▲ Work in progress. This site is built and researched in the open: pages, numbers and findings may change or be wrong, and the record has gaps. Not advice. More
The Lab
Ideas traders use, put to the record
← the Lab · signals tested · write-up
#34 Do coins have a clock? Hours, weekdays, funding times, the US open No reliable edge
Past prices; no real orders were placed. Every figure is a share of the money, or a count.
Last checked: 15 October 2026.
Where the idea comes from
- Calendar and time-of-day patterns were studied in stocks by Kenneth French (1980), on the weekend, and Lawrence Harris (1986), within the day and the week.
- Testing many hours and days at once means some will look good by luck. Why, and how the bar is raised, is on our theory page "Hard-to-beat markets, and why every test needs a random twin".
The question
Traders often say crypto tends to rise or fall at certain times: the Asian morning, the US stock market's open, weekends, or around the funding payments every eight hours. Is any of that real, and is it big enough to pay for the trade?
How we tested
- Ten large coins: Bitcoin, Ether, Solana, BNB, XRP, Dogecoin, Cardano, Avalanche, Chainlink and Litecoin. Hourly futures prices, January 2021 to September 2026.
- 35 recurring times:
- each hour of the day (UTC);
- each weekday;
- the hour before and the hour after each funding payment;
- the hour of the New York open, moved for US summer time;
- the weekend.
- As a trade: in at the start of that time, out at its end. Long or short, whichever way it moved on average in the first years. The exchange's market-order fee (0.055%) is paid both ways, 0.11% a round trip.
- Fair judging:
- patterns were found on January 2021 to September 2025 only;
- then judged once on October 2025 to September 2026;
- because 35 times were tried, a pattern had to be strong enough to stand out among 35 tries;
- to pass, it also had to beat the fee in the last year, on at least 7 of the 10 coins.
The maths, step by step
1. An hour's habit. In the first years, the hour from 05:00 UTC rose on average +0.050% across the ten coins: on 100 dollars, 5 cents.
2. The fee. In and out costs 0.055% twice, 0.11%. So +0.050% − 0.11% = −0.06% a trade, even if the habit held.
3. Strong enough to count? The usual bar is a score (the t-statistic, from "Student", 1908: the average divided by how much it varies) of about 2. With 35 tries, one will pass that bar by luck more often than not. So the bar is raised (a Bonferroni correction) to 5% ÷ 35 = a 0.14% chance, a score of 3.19. Hour 05 scored 2.80.
4. The year it hadn't seen. That hour then went −0.044% on average, with none of the ten coins rising. Bought each day it would have lost 0.044% + 0.11% = 0.154% a trade.
What we found
1. No time of day passed the bar.
- The strongest were 05:00 UTC (+0.050%, score 2.80) and 04:00 UTC (+0.042%, 2.45). Both turned down in the last year.
- Every hour's average move, in either period, was under 0.07%: well below the 0.11% fee.
- Even at the cheaper resting-order fee (0.04% a round trip), every hour lost in the last year.
2. The weekdays only followed the market.
- In the first years every weekday was up on average, because the coins rose. None stood out (the best, Saturday, scored 2.34).
- In the last year, when the coins fell, Tuesday and Thursday lost 0.56% and 0.82% a day on all ten coins, and Monday gained 0.42%.
3. Funding times and the US open moved almost nothing: 0.004% to 0.020% on average. The US open hour fell in the first years and rose in the last.
What it means
- There's no clock to trade. Whatever habits coins have at certain hours are hundredths of a percent, below any fee, and they don't carry from one year to the next.
Limits
- Ten large coins that move together.
- Hourly prices, so nothing shorter than an hour is seen.
- The weekday and weekend trades last a day or two, so the market's own trend dominates them.
Update, 18 October 2026: a chart
Finding 1 as a picture: each hour of the day's score (step 3's t-statistic, across the ten coins) in the first years and in the last year, between the raised bar of 3.19 either side.
No hour reached the bar in either period, and the strongest, 04:00 and 05:00 UTC, turned negative in the last year.
Credits
- Kenneth R. French, "Stock Returns and the Weekend Effect", Journal of Financial Economics, 1980; Lawrence Harris, "A Transaction Data Study of Weekly and Intradaily Patterns in Stock Returns", Journal of Financial Economics, 1986: calendar and time-of-day patterns in stocks.
- "Student" (William Sealy Gosset), "The Probable Error of a Mean", Biometrika, 1908: the t-statistic.
- Carlo Emilio Bonferroni, Teoria statistica delle classi e calcolo delle probabilità, 1936, and Olive Jean Dunn, "Multiple Comparisons Among Means", Journal of the American Statistical Association, 1961: raising the bar for many tests.
- The hours, the funding-time and US-open tests are our own design.
- Data: Binance's public archive of 1-hour futures prices.
- PerpGremlin and Claude, Do coins have a clock? Hours, weekdays, funding times, the US open, TerminalGremlin, 2026: the tests, the code and the words.
← The S&P 500's RSI and BitcoinForced selling and crowded trades →
Educational research from public market data; any trading discussed is paper, never real. Nothing here is advice. The full notice.