The spread between Coinbase and Binance — the US bid against the global bid, measured daily, computed in-house from two public tapes.
Other languages: العربية · Français · Español · Türkçe
Current run: 1 consecutive positive days (from 2026-11-03)
A run at least this long last happened 2026-09-04–2026-09-04 (1 days).
Daily premium, percent, last 90 complete days. Green bars: Coinbase closed above Binance. Red: below.
| Latest complete day | 2026-11-03 |
|---|---|
| Premium | +0.24% |
| Days with a negative premium | 41.0% |
| Longest negative run | −112 (2026-05-06 → 2026-08-25) |
| Longest positive run | +67 (2020-10-02 → 2020-12-07) |
| Deepest negative | -5.35% (2018-11-14) |
| Highest positive | +13.35% (2017-12-23) |
| Complete days on record | 3366 |
| Record starts | 2017-08-17 |
The numbers on this page come from the platform's own series — the same record the app's charts read. The full history is also a public JSON: /api/v1/market/coinbase-premium. The series tops up every six hours; this page serves a five-minute cache of it.
The Coinbase Premium Index answers one question: is the American bid for bitcoin stronger or weaker than the global one? Coinbase is the venue where US dollar flows — retail on-ramps and institutional desks — do most of their buying. Binance is the world's largest exchange by volume, priced in USDT, dominated by global flow. When bitcoin trades above its Binance price on Coinbase, dollars are chasing coins harder in the US than everywhere else. When it trades below, the opposite is true.
The spread is watched because it separates two stories that look identical in a price chart. A rising price with a positive premium says US demand is leading the move. A rising price with a negative premium says the move is being carried by the rest of the world while the US stands aside — a different crowd, with a different appetite for risk, and historically a more fragile one.
That is why the metric earned a place on every serious bitcoin dashboard. It is not a prediction of price; it is a reading of who is doing the buying, updated once a day and hard to fake, because faking it would require moving real size on one exchange and not the other.
The definition fits in one line: (Coinbase BTC-USD daily close − Binance BTC-USDT daily close) ÷ Binance close. Both tapes are public, and the series here is computed in-house over the whole record — back to August 2017, where the Binance tape begins — not licensed from an index vendor. The platform syncs the two tapes onto one calendar and stores the spread day by day.
Three honesty rules are built into the series. First, the USDT basis is part of the number: Binance prices in USDT, so when USDT itself drifts from one dollar, the premium moves with no change in US demand — the same property the commercial reference indexes carry, stated here rather than hidden. Second, a missing day is a hole, not a zero: a day where either tape is absent is excluded from every count, because a spread against a stale close would manufacture a premium out of a calendar mismatch. Third, today's candle is running, not evidence: the UTC day closes at midnight, and until then its premium is drawn flagged and never enters a streak count.
The consequence is that this page's streak can differ from a reference index built on intraday data — both are honest answers to slightly different questions, so the question this series answers is written on the page, and every streak is computed from this series, never quoted from anyone else's.
A run is a stretch of complete days with the same sign; a zero day breaks one. The page shows the current run — how many consecutive days the premium has held one side, and since when — next to the longest negative and longest positive runs in the record, so a current stretch can be compared against history instead of being read in isolation.
The run length matters more than the sign. A one-day negative premium is noise; a six-week negative run during a rising market is a measurable fact about who is not buying. When the current run exceeds every same-sign run before it, the page says so plainly instead of dressing the record up.
Read the streak with the price, never instead of it. A negative run into falling prices describes capitulation; the same run into flat or rising prices describes an absence of US bid underneath a market someone else is carrying. Same number, different story.
A negative premium says dollars were not the marginal buyer on those days. It does not say Americans sold — selling and under-bidding look identical in this spread. It does not say the price must fall: there have been long negative runs that resolved upward once global demand slowed enough for the US bid to matter again.
The positive side has the same limits. A positive premium during US hours can be a desk leaning in — or a USDT basis wobble that made the Binance leg look cheaper without anyone in the US doing anything. This is why the series is read as a run, not as a day, and why the basis is declared on the page instead of edited out.
Treat the index as one input among the platform's other readings — exchange reserves, stablecoin flows, holder structure — rather than a signal on its own. It is at its most informative at the extremes: record positive stretches near local tops, and record negative stretches during heavy drawdowns, have both preceded turns in the past. Neither relationship is a law.
The spread needs a global baseline, and Binance is the deepest bitcoin market in the world, quoted in USDT around the clock. Against that baseline, Coinbase's dollar price becomes a readable signal: whatever premium remains after matching the two calendars is, to first order, the US bid itself.
Other pairings are possible — Coinbase against an aggregate of venues, or against other dollar exchanges — and they give similar-shaped series with different amplitudes. The Coinbase-versus-Binance pairing is the one the reference indexes standardized on, which makes this series comparable to the readings most traders already know.
What the pairing cannot do is separate US retail from US institutions; both buy on the same venue. A reading of who is buying inside the US requires the platform's entity and whale tracking, which works from on-chain evidence rather than from price spreads.
This page is the document for the number the app's market charts already draw. The chart and table above are server-rendered from the platform's own series — the same record the dashboard reads — and the full history is also served as public JSON for anyone who wants the raw days.
To put the premium next to its usual companions: exchange reserves show how many coins sit on venues ready to move; stablecoin flows show dry powder entering or leaving; the premium shows whether the US is the one swinging the bat. Read together, they answer the question a price chart cannot: who is actually doing this.
The percentage spread between bitcoin's price on Coinbase (BTC-USD) and on Binance (BTC-USDT). A positive value means the US venue trades above the global one — US demand leading; negative, the reverse.
In-house, from the two public tapes: (Coinbase BTC-USD daily close − Binance BTC-USDT daily close) ÷ Binance close, stored per UTC day since August 2017. No index vendor is involved.
The reference computes from intraday data; this series uses daily closes, and counts streaks on complete days only. Both are honest answers to slightly different questions — this page shows its own numbers, never quoted from another index.
That dollars were not the marginal buyer those days — the global bid, priced in USDT, was carrying the market. It does not mean Americans were selling, and it does not predict price.
The Binance leg is priced in USDT. When USDT drifts from one dollar, the spread moves with no change in US demand — the USDT basis is part of the number, and the page says so rather than hiding it.
The UTC day closes at midnight. Until then, today's candle is running, not evidence: it is drawn flagged and excluded from every streak count.
The series tops up every six hours and the page serves a five-minute cache of it. Streaks are counted on complete UTC days only.
Yes — the whole record is served as public JSON on the platform's market API, the same series this page renders.