How many bitcoin sit in open perpetual positions on Binance and Bybit, one side of the book, back to 2020. Bold line = the 30-day average.
95,870 BTC both venues · Binance USD-M on Oct-22 2026, the last day in this build, measured and published by askthetape.com.
Open interest is how much is still riding on perpetual futures at this moment — not how much traded, but how much stayed open. On Oct-22 2026 that was 95,870 BTC on Binance USD-M and 30,819 BTC on Bybit: 126,690 bitcoin between them, with Bybit holding 24.3% of it. Each exchange publishes its own history, but only so far back and in its own dialect — so this line is JOINED, and the join is written down rather than hidden. Before 2026-03-31 it is the venue's own daily series; from that day on it is our own capture of the same venue's ticker, the last snapshot of each UTC day. The two halves overlap by 165 days and the table below prints the factor, the correlation and the worst day of that overlap for each venue. 🩸 ONE SIDE of the book everywhere. Bybit reports both sides of every contract, which would print its book at double size next to Binance; it is halved over the whole history, once, and the build refuses to run if that stops being true. 🩸 This is COINS, not dollars. A dollar chart of the same thing rises when the price rises even if not one new contract was opened; in coins the line only moves when positions do. Bitcoin open on these two venues peaked at 166,787 BTC on one of them (Oct-26 2022) and the whole six years are on one axis here. No forecast is drawn.
| binance · provenance | the venue's own daily series before 2026-03-31 (Binance's public daily archives of USD-M futures metrics, 5-minute buckets read at the close of the UTC day); our snapshots of the same venue's ticker from 2026-03-31 on (the last snapshot of the UTC day); normalized ×1.0000 on 165 overlapping days (2026-03-31 → 2026-09-11, r 0.9973, p90 0.19%). The factor is not copied from anywhere: it is re-measured from the overlap file on every build and came out ×1.0000 on this one. USD-M is Binance's dollar-margined book: the contracts there are margined in dollars, not in the coin |
|---|---|
| bybit · provenance | the venue's own daily series before 2026-03-31 (Bybit v5 market/open-interest, intervalTime=1d, singleOpenInterest); our snapshots of the same venue's ticker from 2026-03-31 on (the last snapshot of the UTC day); normalized ×1.0000 on 164 overlapping days (2026-03-31 → 2026-09-10, r 0.9999, p90 0.10%). The factor is not copied from anywhere: it is re-measured from the overlap file on every build and came out ×1.0000 on this one. Their bar is stamped with the START of its period, so the day label is shifted one day back before the two halves are compared |
| bybit · sides | Bybit publishes BOTH sides of every contract; this series is ONE side. The halving is applied to the whole history, so there is no seam to date — and it is applied ONCE. The build refuses to run if that stops being true: it checks our BITCOIN Bybit row for 2026-09-10 (the same check guards every coin, because all of them come out of one file), which must read 28,254.222 — the single side of that day, against 56,508.444 in the exchange's own head field the same minute, checked 2026-09-12. Every other venue on this page and on this site already publishes one side and is left alone. |
| seam · binance, measured on this build | on the first overlapping day (2026-03-31) the venue's history reads 89,212.452 and our capture 89,201.712 — 0.012% apart. There is no step at the join: what jumps across it is the market |
| seam · bybit, measured on this build | on the first overlapping day (2026-03-31) the venue's history reads 24,148.837 and our capture 24,142.351 — 0.027% apart. There is no step at the join: what jumps across it is the market |
| 🩸 what the join is not | the two halves are NOT two independent sources. Both numbers come from the same exchange: its own published daily history before the join, our own snapshots of its live ticker after. What the overlap measures is therefore consistency of two readings of one venue, not agreement of two witnesses — and that is why the factor, the correlation and the worst day are printed instead of a verdict word. |
| how this differs from «How much is bet, on every venue» | that page counts USD of open bets across 18 venues, one principal instrument each, with the venue set stepping up over time; this page counts COINS on two fixed venues (Binance, Bybit) back to 2020 — no composition steps, so the sum and the share view are comparable across the whole history. Same one-side convention on both pipelines: on bitcoin, Bybit ÷ Binance agrees between them to a median of 1.0001 on 2,202 shared days (measured 2026-09-12). |
| gaps | Binance line: no missing days between 2020-09-01 and 2026-10-22 · Bybit line: no missing days between 2020-08-04 and 2026-10-22 |
| cadence | daily, at the close of the UTC day; our own half is the last of roughly 288 snapshots a day. The running UTC day is cut out of the line and shown hollow |
| since | Binance 2020-09-01 (2,243 days) · Bybit 2020-08-04 (2,271 days) — the first day each venue's own archive returns |
| 🩸 where it is thinner | two venues, not the market. Bitcoin perpetuals also trade on OKX, Bitget, Hyperliquid and a dozen smaller books. We show the wider venue set on «How much is bet, on every venue», with each venue starting on its own available date. This page trades breadth for depth: two venues, six years, one protocol. The two halves are also not the same instrument in one respect — the exchange's archive is its own daily aggregation and our half is a snapshot at a moment, so the overlap residual above is the honest size of that difference, not zero. |
| policy | no backfill · read it here · export: from 2026-03-31 (rows before that day are the exchange's own archive and stay here) |
| 🩸 price and open interest rising together · what is counted · [22.09 VIEWS1] | both legs are read for the SAME UTC day from the two files this page already runs on, and open interest is summed in COINS. An event is a day whose close is above the previous day's close and whose combined open interest is above the previous day's — both knowable at that close, nothing is picked with hindsight. Window 2020-09-01 → 2026-10-22, 2,243 days on which both venues report. Rose together: 618 days (2 episodes still inside their 7-day window and drawn hollow); next 7 days median -0.2136 %, mean +1.0094 %, higher 299 of 616. Fell together: 568 days; median +0.5455 %, mean +0.9915 %, higher 309 of 567. THE BASE of the same tape and the same horizon, from ANY of its 2,236 days: median +0.4316 %, higher 52.91 %. Every median is computed from the raw closes at full precision and rounded once, when printed. |
| 🩸 the placebo, and why the p is not the point · [22.09 VIEWS1] | the same count of days was drawn at random from the same tape 2,000 times, and the median of the real days was compared with the medians of the shuffles: rose-together sits below 99.5 % of them (two-sided p = 0.010), fell-together below 29.6 % (p = 0.622). That is a placebo test, not a proof: the 7 days windows OVERLAP, so neighbouring days are not separate draws. Keeping only events whose windows do not overlap leaves 227 of 616 rose-together days (median -0.30 %) and 233 of 567 fell-together days (median +0.50 %) — read those counts, not the p. Taken stricter still, days on which price AND open interest BOTH closed at a 30-day high: 42 days, median -1.10 %, higher 17 of 42. The count a reader may have seen elsewhere («69 days since 2020») is not this page's: on this carrier the plain rule gives 618 days, and the 30-day-high rule 42 days. |
| this build | data to Oct-22 2026 |
| policy | no backfill · read it here · export: from 2026-03-31 (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 21:24Z.
What the tape says today · what's new · all 111 metrics