115,006 Large-Trade Alerts: Where and When Crypto Whales Trade, by Venue and Hour

Vultax logged 115,006 large-trade alerts on six exchanges from 18 June to 3 September 2026. Binance printed the most large trades, 20,022, but HTX carried the most money: 19,617 large trades worth $2.88B, 56% of large-trade notional, with a 90th-percentile print of $364,000 against Binance's $84,000. The largest single print was a $7.06M BTC-USD sell on Coinbase on 27 June.

Direction barely moved. 51.1% of alerts were buys, between 47.5% and 52.6% on every venue and between 49.2% and 53.4% in every hour of the day. What moved was how much printed: the four hours from 13:00 to 16:00 UTC carried 23.6% of all alerts, the busiest hour ran 1.9 times the quietest, and Saturday and Sunday ran 23% and 30% below the weekday average.

The Vultax whale tracker watches exchange trade feeds and raises an alert when a print is large for its pair and venue. Every alert is stored with the exchange, pair, side, price, quantity and notional at the moment it fired, stamped to the second in UTC. This page reports that log, unfiltered, for the 78 days from 18 June to 3 September 2026: 115,006 alerts from six exchanges, on 60 days with data. The 18 days without data are collection gaps and are reported as gaps rather than filled.

Three detectors contribute. A single large trade fires when one print clears the notional threshold for its pair, which sat at roughly $29,000 at the low end during this window. A burst fires when several large trades land inside a 45-second window. A momentum alert fires on three or more consecutive large trades on the same side. One trade can trigger more than one detector, so alerts outnumber trades: 59,803 of the alerts are single large trades, 34,948 are bursts, and 20,255 are momentum alerts.

Hour and weekday figures below are shares or averages of all alerts over the window, on days with data.

Download chart figures:CSVJSON with sourcesSVG chart

• Venues — Binance, HTX, Coinbase, Gate.io, Kraken and KuCoin, the six where the trade feed was complete enough to run detection
• Pairs — BTC accounts for 81% of alerts (BTC/USDT and BTC-USD together), ETH/USDT for 12%, SOL for 5.5%
• Notional — every figure is the print's own price times quantity; alerts that share a trade are double-counted in the $11.38B all-alert total, so venue comparisons below use single large trades only

Counting single large trades only, Binance produced the most prints and HTX produced the most money. HTX's median large trade was $65,000 and its 90th percentile $364,000, more than four times Binance's $84,000.

Put differently, HTX accounted for a third of the large trades in the sample and 56% of the large-trade notional. The gap widened over the summer: HTX large-trade notional went from $0.58B in June to $0.84B in July and $1.39B in August, while Binance went from $0.24B to $0.38B to $0.44B.

A large print is not, by itself, evidence of genuine demand. The academic literature on crypto volume, summarised in our wash-trading study, finds that reported activity on some venues is dominated by trades with no economic purpose, and a venue where prints are structurally larger deserves the same question. That is why the terminal shows a venue-quality score next to every print instead of ranking venues by size. Nothing here is a claim about why HTX prints are larger; it is a measurement of the fact that they are.

Download table figures:CSVJSON with sources

Most alerts are not enormous. 91,246 of the 115,006 (79%) were under $100,000. Another 15,572 fell between $100,000 and $250,000, 5,098 between $250,000 and $500,000, and 2,482 between $500,000 and $1M. Only 604 alerts were between $1M and $5M, and four exceeded $5M. Those 608 prints above $1M carried $0.94B, about 8% of all alert notional, from half a percent of the alerts.

The five largest prints in the window were a $7.06M BTC-USD sell on Coinbase on 27 June, a $5.45M ETH/USDT sell on HTX on 19 August, a $5.26M BTC/USDT sell on HTX on 13 July, a $4.69M BTC-USD buy on Coinbase on 20 August, and a $4.24M BTC-USD sell on Coinbase on 27 June. Four of the five were sells.

51.1% of all alerts were buys. No venue strayed far from that: KuCoin was lowest at 47.5% buys and HTX highest at 52.6%. Across the last three weeks of the window the daily buy share ranged from 49.4% to 56.5%. At the level of raw counts, large flow was near-balanced on most days. By hour of day the band was just as narrow, from 49.2% buys at 15:00 UTC to 53.4% at 06:00 UTC: the U.S.-morning surge described below is a surge in both directions.

What separates a day is not the count of large prints but how they cluster. The burst detector fired 34,948 times and the momentum detector 20,255 times, and those clusters, not isolated prints, are what a whale-flow read is built on. A single $400,000 buy is a data point. Six of them inside a minute on two venues is a position being built.

For scale, a 25% sample of every BTC trade on Binance and Coinbase over the 24 hours to 3 September 2026 puts the median Binance BTC/USDT print at five dollars, the 90th percentile at $383 and the 99th at $6,575. On Coinbase the median was $25, the 90th percentile $1,699 and the 99th $11,413. Trades of $100,000 or more made up 4.4% of sampled notional on Binance and 2.8% on Coinbase.

A $50,000 alert therefore sits more than 100 times above the median Binance trade and around the 99.5th percentile of the tape. Whale alerts are, by construction, the far tail, which is why a handful of them carries so much of the notional.

Large prints follow the sun into North America. From 23:00 to 06:00 UTC each hour carries between 3.3% and 3.8% of the day's alerts. Activity lifts through the European morning, reaches 5.05% at 12:00 UTC, and holds above 6% for three consecutive hours from 13:00 to 15:00 UTC, the window that contains the U.S. equity open at 13:30 UTC and the overlap of the European afternoon. The busiest hour, 15:00 UTC, carried 6.14% of all alerts; the quietest, 23:00 UTC, carried 3.25%. That is a 1.9-to-1 ratio between the loudest and quietest hours of the day.

The shape matches what the academic record shows for bitcoin volume generally. Eross, McGroarty, Urquhart and Wang found bitcoin trading activity peaks during U.S. market hours and troughs in the Asian night, and Baur, Cahill, Godfrey and Liu documented time-of-day and day-of-week effects in bitcoin volume that persist even though the market never closes. What this log adds is that the pattern holds for the far tail of the tape, not just for total volume.

Download chart figures:CSVJSON with sourcesSVG chart

Venues differ in how much of their large flow lands in the U.S. morning versus overnight. KuCoin, Kraken and Gate.io are the most U.S.-morning heavy: 29.4%, 28.9% and 27.5% of their alerts fell between 13:00 and 16:00 UTC. Binance is the most overnight heavy, with 26.7% of its alerts between 00:00 and 06:00 UTC, followed by Gate.io at 25.3% and HTX at 25.1%. KuCoin's overnight share was the lowest at 17.2%.

The practical consequence is that the same $300,000 print carries different context on different venues at the same hour. At 04:00 UTC it sits in Binance's busy period and in KuCoin's quiet one; a print that is one of dozens is ordinary, and one that is nearly alone is worth a second look. The whale tracker shows the recent count per venue for that reason.

Download table figures:CSVJSON with sources

Weekdays averaged 2,048 alerts per day. Friday was the busiest at 2,324 per day, then Monday at 2,235 and Wednesday at 2,208; Tuesday and Thursday sat lower at 1,686 and 1,789. Saturday averaged 1,570 alerts and Sunday 1,424, 23% and 30% below the weekday average. Crypto trades on weekends, but its large prints thin out.

Download chart figures:CSVJSON with sourcesSVG chart

Thresholds set once and applied around the clock will fire constantly at 15:00 UTC and rarely at 23:00 UTC. A threshold that tracks the hour's typical rate makes an alert mean the same thing at any time of day. The other use is reading a print in context: a large trade in a venue's quiet window has fewer peers and is harder to hide inside ordinary flow, which is exactly the kind of context the terminal attaches to each alert.

The strip at the top of this page is not a screenshot of the day this was written. Every ten minutes Vultax re-reads its own whale-event table for the last 24 hours, the public trade tapes of five venues for prints of $100,000 or more, Binance funding, Bybit open interest and Deribit's DVOL and rewrites the figures; the Vi IQ beneath them scores the same six domains the terminal scores for a pair, computed for BTC on the major spot venues, with any domain that cannot be computed shown as unavailable rather than filled in. Read the numbers as a live check on the argument above, and the revision notes at the end for what has changed since publication.

The first three figures come from Vultax's own detection across the exchange streams it watches: how many large-trade events fired in the last 24 hours, their combined notional, and the largest single one. The fourth is independent: the count of prints of $100,000 or more on the public tapes of Coinbase, Binance, Kraken, OKX and Bybit in the last thirty minutes. When the two disagree it is usually because the largest prints are on venues the public tapes do not cover; the pack behind the page breaks the day's events down by exchange.

Most 'whale' coverage in 2026 is about on-chain movements, and it is measuring something different from this article. CoinDesk reported in July 2026 that large holders accumulated about 270,000 BTC in two weeks while U.S. spot ETFs saw a record $4 billion of outflows, and CryptoQuant's exchange whale ratio, the share of exchange inflows coming from the ten largest deposits, reached a ten-month high in January with Glassnode measuring spot volume at its lowest since November 2023. Those are wallet-to-exchange transfers and holdings, and much of the largest flow is ETF custody moving between custodians.

This article counts prints: trades that executed on an exchange's tape above a threshold. A transfer onto an exchange is a possible sale; a print is a sale, with a side, a venue and a time. The two series can move in opposite directions for weeks, as they did in the summer of 2026, and the hour-by-hour pattern described here belongs to the tape, not to the chain.

Alerts count detector events, not unique trades, and thresholds differ by pair and venue, so a venue with more small pairs under coverage will show more alerts per dollar. Detection ran on six venues where the trade feed was complete; the terminal covers more venues for order books than for trades. Eighteen of the 78 days have no data and are excluded rather than interpolated. All notional is measured at the print price and is not adjusted for the same capital cycling through a venue. The raw alert log is retained, so every figure here can be recomputed. Hour-of-day shares are pooled over 60 days of data and six venues, so a single venue's curve is noisier than the aggregate, and the clock describes when large prints were logged, not whether trading at any hour is better or worse.

To watch large prints as they happen, the BTC venue desk in the Vultax terminal shows whale trades alongside each venue's book, the cross-exchange gap and liquidity.

• Window — 2026-06-18 to 2026-09-03 00:26 UTC, 60 days with data
• Threshold — pair- and venue-specific; the smallest alert in the window was about $29,000
• Dataset — daily alerts, large trades, bursts, momentum alerts, buy share and notional by exchange are available as a CSV on this page