Crypto never closes, but it still has tides. Volume rises at the US open, drains through the Asian night, and ebbs every weekend. This page measures those tides across Bitcoin, exchanges, DEX chains and launchpads, and marks which patterns hold at least 80% of the time.
Share of each day's volume traded in each hour. A flat day would put 4.17% in every hour. Shaded bands are 95% confidence intervals across days.
Sessions: Asia 00–08 UTC, Europe 07–16, US 13–21. Hourly candles can't separate 13:30 from 14:30, so the US equity open shows up across hours 13 and 14.
Source: Binance public market data.
Each cell is that hour's volume divided by the average hour of the previous 28 days, rescaled so the map averages 1.0. It opens on all crypto combined (exchanges and DEX together); switch to exchanges only, DEX only, or a single market, chain or launchpad. Hover a cell for its value.
Source: Binance public market data.
Volume by weekday relative to its trailing 28-day average. Weekend ratio = average weekend day ÷ average weekday. Days in these charts are UTC days; the heatmap above follows your chosen time zone.
Source: Binance public market data. Data: DefiLlama.
Bars show weekend ÷ weekday. Blue bars pass the 80% test (the weekend was quieter in at least 80% of weeks); amber bars don't, or have too little data. Hover for the share.
Source: Binance public market data. Data: DefiLlama.
Saturday and Sunday were below every weekday in all 9 years since 2018. The weekend dip deepened from about 0.85× to 0.5–0.65×, and sharpened further after the US spot ETFs launched in January 2024.
Source: Binance public market data.
pump.fun and PumpSwap on Solana compared with Solana DEX volume over the same weeks, plus Pons on Robinhood Chain.
Hourly volume from the deepest DEX pools on each chain over the last 180 days (GeckoTerminal), plus top memecoin pools. Robinhood Chain uses its WETH/USDG pools from July 2026.
Big WETH/stablecoin pools are dominated by arbitrage against exchanges, so their timing mirrors exchange hours. Memecoin pools are shorter histories picked from today's top lists, so treat them as indications. Solana memecoins peak later (21:00 UTC) and show no weekend dip.
On-chain data provided by GeckoTerminal.
Every DEX trade on Ethereum, Base, Arbitrum, BNB Chain, Polygon, Solana and Robinhood Chain over 24 months, from Dune's trade tables. On-chain USD values include some mispriced and wash trades, so each hour shows the median share across days, which one bad day can't move.
Launchpads here are pump.fun (bonding curve), PumpSwap (after graduation) and Raydium LaunchLab on Solana. Robinhood Chain has 14 weeks of data, too few for confident weekday conclusions.
Data: Dune (dex.trades, dex_solana.trades). Queries run 28 Sep 2026.
A launch "graduates" (migrates) when its bonding curve fills and its liquidity moves to a DEX pool. The rate here is graduations in a period divided by launches in the same period, so a sudden flood of bot launches pushes a month's rate down.
Data: Dune (decoded launchpad events: token created vs graduated, migrated or listed). Queries run 28 Sep 2026.
When a launch never graduates, the money its buyers paid in stays behind: in the bonding curve, or in a locked pool on launchpads without curves. It still belongs to the token's holders, who can sell back into it, but on tokens nobody trades any more it just sits there. In the launchpads we studied, we did not find a mechanism that winds these down. Only the pair asset is counted (the SOL, ETH, BNB, stablecoin or stock token buyers paid in); the launched tokens themselves are valued at zero. Every figure below was read from each chain's own contract state or event logs, not taken from a published estimate.
Data: direct reads of Solana (via Helius), BNB Chain, Robinhood Chain and MegaETH state and event logs; Dune for BNB Chain token lists and LONG trades. USD at on-chain prices on 28 Sep 2026: Pyth SOL/USD, Chainlink ETH/USD and PancakeSwap BNB/USDT. Read 28 Sep 2026.
Each month's volume relative to the trailing 12 months. Only 2–8 years support each month, so wide bands are expected. No month passes the 80% bar except November on DEX, which matches the 2020, 2021 and 2024 rallies.
Source: Binance public market data. Data: DefiLlama.
2026 covers 270 days.
Source: Binance public market data.
Twelve months of exchange-reported daily volume.
Twelve major shocks from 2020 to 2026. Pick one to see Bitcoin's price and hourly volume (× its normal level) from 3 days before to 4 days after.
Source: Binance public market data. Event dates and causes from contemporary news reports.
"Holds" means the pattern appears in at least 80% of the independent weeks, days, years or events tested, or has a 95% confidence interval that excludes no effect.
We want you to know exactly what you are reading, so here is the whole process, including where it can go wrong.
The EOTM Labs team chose the questions, the sources and what to publish. The research work was done by AI agents (Anthropic's Claude, running in Claude Code): they wrote the data queries and scripts, ran them, cleaned the results, drew the charts and drafted the text. People reviewed the output and asked for changes; not every number was re-derived by hand.
Only public sources: exchange and aggregator APIs, Dune Analytics queries over decoded on-chain tables, and direct reads of contract state and event logs from public blockchain nodes (Robinhood Chain, BNB Chain, Solana via an RPC provider). Each chart names its source underneath. Nothing is estimated from private data.
Where possible the same figure was measured two ways and compared, for example launch and graduation counts from Dune against a direct scan of the launchpad's own event logs. When the two disagreed, the difference was traced before publishing: one early on-chain scan counted the wrong event as a "graduation" and was caught this way. Several agents worked in parallel on separate pieces, and their scripts and outputs were kept so each result can be rerun.
AI agents can misread a contract, pick the wrong event or table, or make an arithmetic slip, and a check can miss it. The underlying data can itself be incomplete, delayed, revised or distorted by bots and wash trading. Some figures are sampled or partial, and those are marked. Treat every number as an estimate that shows the size and shape of a pattern, not as an exact or audited value.
TideWatch is provided "as is", for general information and research only. EOTM Labs makes no promise that it is accurate, complete or current, and accepts no responsibility for decisions made using it. If you rely on a figure, check it against the original source. Spotted an error? Tell us on X @EOTMofficial and we'll look into it.
TideWatch is refreshed by hand from these sources; it is not a live feed. Each source's newest data point is listed so you can judge how current a chart is.
Why we built it. The globe is made to run around the clock on a screen in the EOTM Labs office, so the team can see at a glance where the crypto day's tide stands and when the next peak is coming.
For viewing only. An illustrative view to explore daily patterns, not live data and not precise. Dollar figures are typical values: recent daily volume (CoinGecko exchange-reported, month shown; DefiLlama DEX, last 30 days) spread by our measured hour-of-day shares (Binance, Dune). Places are approximate and do not show where traders are. Not financial advice. Earth imagery: NASA.
TideWatch is Instrument No. 001 by EOTM Labs · Economy of the Many.