Study · Original data

When is the order book thinnest? Weekends and the daily break

By Swiftedge · Published 5 Nov 2026

From 30 October to 5 November 2026 (one weekend), resting liquidity within 1% of price on the five 24/7 contracts on traditional markets fell at the weekend to 48-74% of the level on the neighboring weekdays, while no hour of the day moved the four crypto order books more than 9% from their usual level.

Weekend and daily-break liquidity, by market

Resting bids and asks within 1% of the mid price. "Weekend vs Friday and Monday" compares each weekend hour with the same hour on the neighboring weekdays. For the traditional markets, the thinnest and thickest hours are for weekdays only; in brackets, liquidity in that hour as a share of the market's median over the whole week. Full hourly data: CSV.

Hour by hour

Red bar = thinnest hour, green = thickest.

Gold

Weekdays only, as a share of the market's median over the whole week, per UTC hour

Silver

Weekdays only, as a share of the market's median over the whole week, per UTC hour

WTI oil

Weekdays only, as a share of the market's median over the whole week, per UTC hour

US 500

Weekdays only, as a share of the market's median over the whole week, per UTC hour

US 100

Weekdays only, as a share of the market's median over the whole week, per UTC hour

Bitcoin

All days, share of the median per UTC hour

Ethereum

All days, share of the median per UTC hour

Solana

All days, share of the median per UTC hour

XRP

All days, share of the median per UTC hour

How it was measured

We used Swiftedge's own recording of perpetual futures order books for nine markets: four crypto markets and five 24/7 contracts that track precious metals, oil and US stock indices. All nine trade around the clock, including weekends and the hours when the reference markets for the metals, oil and the indices are closed. The recording takes a snapshot of the visible order book several times a second and stores, for every minute, the average resting size at each price level during that minute.

For each minute we added up the dollar value (size times price) of all resting buy orders (bids, below the mid price) and sell orders (asks, above it) within 0.5% and within 1% of the mid price. This edition uses every full hour from 30 October 2026 9:00 AM ET (14:00 UTC) to 5 November 2026 8:00 AM ET (13:00 UTC): 143 hours and 8,575 minutes per market, the same window for all nine, with no minutes rejected. For each UTC hour of the day we took the median of all minutes in that hour across all days and divided it by the market's median over the whole window. A ratio of 0.80 means the book in that hour typically held 20% less than usual for that market, so markets of very different size can be compared.

Weekend means Saturday and Sunday in UTC. Sessions are fixed UTC hours, start included and end excluded: Asia 00:00-07:00, London 07:00-13:00, New York 13:00-21:00 and late 21:00-24:00. Because liquidity levels drift over time, we checked the results in two extra ways: the hourly profile was recomputed against a rolling 24-hour median, and weekends were compared hour for hour with the Friday before and the Monday after rather than with the whole window. For the markets that were already recorded before 30 October 2026, from a different set of order books, we also compared the hourly profile with that earlier recording. The study is re-run automatically every week on the latest 28 days of recording.

Limits of this study

• Window: 30 October 9:00 AM ET (14:00 UTC) to 5 November 13:00 UTC (143 hours), so each hour of the day rests on 5-6 daily observations, and the window has one weekend (31 October-1 November). With a single weekend, the weekday-versus-weekend split, for crypto in particular, is a first look.
• The window never starts before 30 October 2026 9:00 AM ET (14:00 UTC). Before that, four of the nine markets were recorded from a different set of order books and the other five were not recorded at all; mixing the two would compare different books, so the earlier recording is used only as a cross-check.
• Liquidity levels drifted during the window (comparing full weekdays, Bitcoin's daily median rose from $331 million on Monday 2 November to $362 million on Wednesday 4 November; US 500's rose from $44 million on Monday 2 November to $53 million on Wednesday 4 November). Ratios against the whole window include this drift; the hour-for-hour comparison with the neighboring Friday and Monday is less affected by it.
• The crypto hour-of-day differences are small (within about 9%) and changed between recordings, so no single thinnest hour for crypto stands out as a rule; Bitcoin's thinnest hour was below its rolling 24-hour level on 4 of 6 days.
• US clocks changed on 1 November 2026. Before that, the daily break of the main US futures markets and their weekly close fell at 4:00–5:00 PM ET (21:00-22:00 UTC), afterwards at 5:00–6:00 PM ET (22:00-23:00 UTC), so weekday hours in this window mix both times.
• For the traditional markets, the weekday dip at 5:00 PM ET (22:00 UTC) rests on only three weekdays with the daily break in that hour (Monday to Thursday); on Fridays that hour already falls after the weekly close.
• This measures visible resting orders near price, averaged over each minute. It does not include hidden orders, does not show how fast orders are pulled or refilled, and is not the cost of executing a trade.
• Dollar levels depend on which order books are combined in the recording, so the ratios are more comparable across markets than the dollar amounts. The results for the traditional markets describe these 24/7 perpetual contracts, not the underlying exchange-traded markets.

Cite this

Common questions

Keep reading