Positioning from Bybit, OKX and Gate on 20 coins — Bybit as the SHARE of accounts long (a percentage), OKX and Gate as longs ÷ shorts (a ratio) — including the Gate cuts almost nobody records. Our own hourly capture sits on top of each venue's OWN daily archive, back to 2025-05-03. Each venue counts its own customers its own way, so the lines are shown side by side and never added.
66.4% BTC · Bybit · share of accounts long on Oct-22 2026, the last day in this build, measured and published by askthetape.com.
We read the positioning figures published by Bybit, OKX and Gate. They publish how many accounts are long against short, and separately what the biggest accounts hold. The interesting part is that the three exchanges disagree, often sharply, because they have different customers. We keep all three hourly on the same coins, so the disagreement itself becomes a chart. Gate publishes six different cuts of this and almost nobody records them. Historical archive fetched 2026-10-06 from the venue's public endpoint / the public CFTC endpoint; source units and gaps preserved; browser-only.
| leg | `venue_ls_1h` — own hourly capture of each venue's public positioning endpoint, the last hour of the UTC day · before our first day of capture, each venue's OWN daily archive: Bybit 2025-05-03 → · OKX 2026-03-20 → · Gate 2026-03-20 → |
|---|---|
| cadence | hourly, daily close · the archive is that venue's own one-day bar |
| since | 2025-05-03 (the whole page) · our own capture: Bybit, share of accounts long 2026-06-13 · OKX, all accounts 2026-06-13 · OKX, top accounts 2026-06-13 · OKX, top positions 2026-06-13 · Gate, accounts 2026-06-13 · Gate, takers 2026-06-13 · Gate, top accounts 2026-06-13 · Gate, top positions by size 2026-06-13 |
| 🩸 the archive is measured against our own capture, never trusted | 16,820 archived days were spliced in across 160 legs, pulled 2026-09-14. Two things were checked before a single point went on the chart, on the days where archive and capture overlap. THE LABEL: every bar was tried on three alignments — the label's own UTC day, the day before, the day after — and the winner is printed per leg below, because the venues do NOT agree: Gate's and OKX's bars carry the label's own day, Bybit's 00:00 bar carries the day BEFORE it (the same convention its open-interest archive uses, measured 2026-09-12), and OKX changed its mind mid-history: on 2026-08-14 its rubik endpoint moved every period one day forward, so this page judges and splices OKX with TWO rules, one per era, instead of one rule that would be wrong on a quarter of the overlapping days. THE SIZE, recomputed on every build from the two files, with the WORST COIN of each leg named rather than hidden inside the pool: Bybit, share of accounts long: 1,840 common full days, r 1.00, slope 1.00, median gap 0.12 points (0.2 % of a typical 72.4 %, worst coin TIA 0.26 points (0.4 %)), label «label day + 1» · OKX, all accounts: 1,860 common full days, r 0.97, slope 1.00, median gap 0.0500 (2.5 % of a typical 1.97, worst coin ENA 0.0800 (5.1 %)), label «label day until 2026-08-13 / label day - 1 from 2026-08-14» · OKX, top accounts: 1,860 common full days, r 0.97, slope 1.01, median gap 0.0436 (4.1 % of a typical 1.05, worst coin ENA 0.0500 (6.3 %)), label «label day until 2026-08-13 / label day - 1 from 2026-08-14» · OKX, top positions: 1,860 common full days, r 0.97, slope 1.00, median gap 0.0080 (0.9 % of a typical 0.89, worst coin ENA 0.0176 (1.9 %)), label «label day until 2026-08-13 / label day - 1 from 2026-08-14» · Gate, accounts: 1,860 common full days, r 1.00, slope 1.00, median gap 0.0000 (0.0 % of a typical 1.42, worst coin BTC 0.0000 (0.0 %)), label «label day» · Gate, takers: 1,860 common full days, r 0.71, slope 0.96, median gap 0.0000 (0.0 % of a typical 0.98, worst coin XRP 0.0000 (0.0 %)), label «label day» · Gate, top accounts: 1,860 common full days, r 1.00, slope 1.00, median gap 0.0000 (0.0 % of a typical 0.53, worst coin BTC 0.0000 (0.0 %)), label «label day» · Gate, top positions by size: 1,860 common full days, r 1.00, slope 1.00, median gap 0.0000 (0.0 % of a typical 0.97, worst coin BTC 0.0000 (0.0 %)), label «label day». Gate's four cuts are the same daily bar the venue publishes — our poller reads it at the close, so they agree by construction, not by luck; Bybit and OKX sit within 4.1 % of a typical day on the pool and are not expected to be exact — our number is the last hour of the day, theirs is the venue's own daily bar. |
| 🩸 the taker line is measured with the venue's OWN rule, not ours | the fourth Gate cut is a RATIO OF SUMS, not a snapshot: the exchange divides the whole day's long taker size by the whole day's short taker size. Read as the last hour of the day — the rule every other line here uses — our number missed the venue's own daily bar by 25.3 % of median. Rebuilt as our own sums over the same UTC day it lands on the venue's number EXACTLY on 1,287 of 1,760 common full days (median relative gap 0.0 %, p90 1.01 %), so this page now draws it that way and its archive is spliced like the other three. On this build the pool reads r 0.71, slope 0.96, median gap 0.0000 on 1,860 common days. The line is noisier than the rest by nature — a day of aggressive flow swings far more than a day of account counts — which is why its r is the lowest on the page while its median gap is zero. |
| 🩸 survivorship | the coin list is TODAY's — the 20 coins we capture now, carried back through the archive as if they had always been the list; a market that mattered on these venues in 2025 and is not in our capture today is absent from the whole history. And the first day of each archive is where the venue's WINDOW ends, not where its data begins: Bybit answers 500 daily bars, OKX 180, Gate refuses anything older than 180 days («from time exceeds 180-day limit», checked 2026-09-14). Deeper history is not published free by any of the three. |
| 🩸 the day the archive hands over | the dashed vertical on each view is the first day of OUR capture for that leg (Bybit, share of accounts long 2026-06-13 · OKX, all accounts 2026-06-13 · OKX, top accounts 2026-06-13 · OKX, top positions 2026-06-13 · Gate, accounts 2026-06-13 · Gate, takers 2026-06-13 · Gate, top accounts 2026-06-13 · Gate, top positions by size 2026-06-13): left of it is the venue's own daily bar, right of it is our hourly poll. Our first day is itself a PART day — we started mid-day, 14 to 18 hours of 24 — so where the archive covers that day, it is drawn from the archive instead: neither side of the seam shows an incomplete day. |
| policy | each number is that venue's own definition — Bybit publishes a percentage of accounts (50 % = evenly split), OKX and Gate publish ratios (1 = evenly split) — so they sit side by side, on their own axes, and are never added together or called by one name. Gate's two taker-size fields are in the venue's own units and are left off the chart rather than labelled as dollars; OKX's taker_buy and taker_sell fields are left off for the same reason, and both omissions are stated here rather than only one. · export: from 2026-06-13 (rows before that day are the exchange's own archive and stay here) |
| hourly window | prod venue_ls_1h, our own hourly capture of each venue's handle · since 2026-05-13 10:00Z by data, 3,924 hours with data inside a 3,924-hour span → 2026-10-23 21:00Z (the Tape and Desk windows draw up to 400 days of that, never more than we hold) · captured 2026-10-23T21:21:56Z |
| 🩸 one line to a view · [22.09 VIEWS1] | the OKX and Gate views now OPEN on a single line — that venue's ratio over ALL of its accounts, the widest cut it publishes — and the narrower cuts (top accounts, top positions, takers) wait in the legend behind one click, exactly as the 19 venue lines do elsewhere on this site. 20 OKX views and 20 Gate views were changed this way; no series, no number and no archive was touched, and a link that carries ?h= still shows every line. The reason is a measurement, not taste: three and four lines of the same colour family on one axis answer no question in five seconds, and the cuts are not comparable to one another anyway — «top accounts» is a different population from «all accounts», not a refinement of it. |
| archival source | Historical archive fetched 2026-10-06 from the venue's public endpoint / the public CFTC endpoint; source units and gaps preserved; browser-only. |
| historical views | Browser-only historical views retain the source dates and units. |
| this build | data to Oct-22 2026 |
| policy | no backfill · read it here · export: from 2026-06-13 (rows before that day are the exchange's own archive and stay here) |
The chart itself is drawn in the browser: turn JavaScript on to hover a day, pick a window and draw your line. The numbers above come from the last build, Oct-23 2026 21:59Z.
What the tape says today · what's new · all 111 metrics