How to Tell Whether a Rally Was Bought With Cash or With Leverage — Six Checks, and the One Identity That Has to Multiply Out Exactly — Field Guide #37

Every large move in bitcoin arrives with a free explanation attached: it was leverage. Sometimes that is right. In August 2026 it was measurably wrong, and the measurement took about ninety seconds and two public API endpoints.

This guide is the procedure. Six checks, in the order that makes each one cheap. Run them and you will know whether a rally was financed with borrowed money or with cash — and, just as importantly, which of the two your favourite commentator has failed to check.

Field Guide #37 in the Bitcoin Investor’s Reading Room.

Why the question matters

A rally bought with cash and a rally bought with leverage look identical on a price chart and behave completely differently afterwards.

Leverage has to be serviced. Perpetual longs pay funding every eight hours. Dated-futures longs pay the basis. Both are financed positions with a running cost, and both can be closed by someone other than their owner — a liquidation engine does not ask permission. A market whose gains sit on borrowed money carries a stock of forced sellers who will all appear in the same five minutes.

Cash does not do that. A spot buyer who is wrong sits there being wrong. The position has no running cost, no margin call and no liquidation price. The same 20% gain is a fundamentally different object depending on which of the two paid for it, which is why “it’s all leverage” is a claim about future fragility and not merely a piece of colour.

Check 1 — Read open interest in coins, and multiply the two changes out

This is the check that does the most work and the one almost nobody performs, because exchanges publish open interest in two units and the financial press quotes the wrong one.

Binance’s openInterestHist endpoint returns both sumOpenInterest (coins) and sumOpenInterestValue (dollars) for every daily snapshot. Dollar open interest rises when the price rises even if not one new contract is opened, because it is simply coins multiplied by price. Quoting it as evidence of leverage growth is a circular argument dressed as data.

The procedure is a single line of arithmetic. Take the start and end of your window, compute the percentage change in coins and the percentage change in the implied price (dollars divided by coins), and multiply:

(1 + coin change) × (1 + price change) − 1 = dollar change

If that identity does not close to several decimal places, one of your three numbers is from a different snapshot and you should stop and find out which. If it does close, you have decomposed the move and you can say exactly how much of the “record open interest” headline is positions and how much is revaluation.

Worked example, August 2026. Binance BTCUSDT perpetual, own pull, daily snapshots:

1.23671 × 0.97976 = 1.21168. The identity closes exactly. The dollar figure rose $1.44 billion and the coin figure fell by 2,196 contracts’ worth of bitcoin. Anyone citing the dollar number as proof of a leverage build in August was citing the price of bitcoin.

The trap: a two-point decomposition hides the path between the two points. In this example open interest peaked at 111,988.29 coins on 15 August and bottomed at 105,531.28 on 24 August — a 5.77% range that the endpoints erase. Always print the intra-window high and low next to the decomposition. If the high is above the start, “positions fell” is an endpoint artifact rather than a fact.

Check 2 — Annualise the funding rate, and find the cap

A funding rate of “0.01%” means nothing until you know the interval and the cap.

Binance settles BTCUSDT funding three times a day, so annualised cost = rate × 3 × 365. At 0.0100% that is 10.95% a year. The mean of the last 90 settlements at the time of writing is 0.006602%, which annualises to 7.23%. That is the actual running cost a perpetual long has been paying, and it is a number you can compare to anything else in finance — which is the whole point of annualising.

Then find the cap, because it changes the meaning of every reading near it. On this contract the observed maximum across 500 settlements is exactly 0.010000%, hit 42 times, with zero prints above it. When a series stops dead at a round number that often, you are looking at an administrative clamp, and a clamped reading is a lower bound on demand, not a measurement of it. Ten settlements pinned to a cap could be mild enthusiasm or a stampede; the series cannot tell you which.

One diagnostic that can: check the floor. If the same series has printed below the negative of the cap — as this one has, twice, with a minimum of −0.012276% — then the clamp is asymmetric or soft, and the apparent ceiling is telling you something real about the distribution rather than merely about the exchange’s rulebook.

Check 3 — Annualise the dated basis, then subtract the bill

Dated futures give you the cleanest read on leverage demand, because unlike a perpetual they have an expiry, so the premium converts directly into an interest rate.

Three steps, and one rule.

Step one: take the futures mark and the index price from the same venue and the same second. Comparing a Binance future to a Coinbase spot print introduces a venue spread into a number you are about to multiply by fourteen.

Step two: premium = mark / index − 1. Step three: annualise by 365 divided by days to expiry — and use the actual settlement timestamp, not the calendar date. A contract expiring at 08:00 UTC on 25 September is 25.07 days away from 06:13 UTC on 31 August, not 25, and on a short-dated contract that rounding is worth several basis points.

The rule: a basis is not a return until you subtract the risk-free rate. The cash-and-carry trade — long spot, short the future — locks in the premium, and the money used could otherwise have sat in Treasury bills. The excess over bills is the real reward for lending into crypto leverage.

At 06:13 UTC on 31 August 2026: the September quarterly annualised at 5.349%, the three-month Treasury constant-maturity par yield on 28 August was 3.90%, so the excess is +145 basis points. Positive, real, and thin. A leveraged bitcoin bull market prints double-digit annualised basis; 1.45 points over bills is a market where somebody is being paid to lend, but not paid well.

Check 4 — Compare two tenors and look at the shape

One basis number is a level. Two are a curve, and the curve is where the information is.

Same morning, same venue, same method: September quarterly 5.349% annualised, December quarterly 5.084%. Four and a half months apart and 27 basis points different.

Read it like this. A steep upward curve — far months annualising well above near months — means leverage demand is being priced for duration; someone wants financed exposure and is willing to pay increasing amounts to hold it. A flat curve means the marginal buyer is a carry trader indifferent between tenors, arbitraging a small spread. An inverted curve, with the front annualising above the back as here, is usually a short-dated squeeze or an event premium rather than a structural leverage bid.

The August 2026 curve is flat to marginally inverted. Combined with Check 1, that is a coherent picture: no new positions, and no term premium being paid for the ones that exist.

Check 5 — Split the cohorts, and read the spread not the level

Exchanges publish long/short ratios for different groups of traders. Binance publishes three for BTCUSDT: top traders by position, top traders by account, and all accounts.

The level of any one of them is close to useless — it drifts, it is venue-specific, and it has no natural neutral point. The spread between the large-account series and the all-account series is the usable statistic, because both series share the same venue, the same biases and the same construction, so most of the noise cancels.

On 31 August 2026: top traders by position 2.0586, all accounts 1.0812, spread 0.9774, against a ten-day range of 0.8949 to 1.2552. Size money roughly two-to-one long, the crowd barely off flat, and the gap narrowing over three sessions rather than widening.

The trap, and this desk has fallen into it: these are daily buckets. A ratio dated 31 August is a snapshot taken at a fixed hour, not a reading of what happened during the session. Do not describe it with an intraday verb. If you find yourself writing “the crowd flipped short as the price fell,” check whether your data has any idea when the price fell. It does not.

Check 6 — Find the cash leg

Checks 1 to 5 can only tell you that leverage did not pay for a rally. They cannot tell you what did. If open interest is flat and price is up 24%, the money came from somewhere, and a guide that stops at the derivatives is only half a guide.

Three places to look, in descending order of how well documented they are:

Spot ETF flows. For US-listed funds these are published daily, per fund, and are as close to a clean cash-leg measurement as this market offers. August 2026: $3,322.4 million of net inflow over twenty sessions, of which $2,883.7 million — 86.8% — was a single fund, IBIT (own recompute of the Farside Investors daily table). That is real cash buying real coins, and it is roughly the size the price move requires.

Corporate treasuries. Public companies disclose purchases in filings. This is high-quality evidence when it exists and absent when it does not — the largest such buyer disclosed no purchases at all between 22 June and 23 August 2026, so it contributed nothing to this particular month.

Exchange balances and stablecoin supply. Weaker evidence, more assumptions, more attribution risk. Useful as corroboration, not as a lead.

When the derivatives say “not leverage” and the cash leg says “$3.3 billion of ETF inflow,” the two halves agree and you can state a conclusion. When the derivatives say “not leverage” and you cannot find a cash leg, you have not found evidence of a mysterious buyer; you have found that your window is too short or your data too thin. Say so.

The one-page version

What the six checks cannot do

Three limits, stated so you do not over-claim with them.

They are venue-bounded. Binance is the largest perpetual venue but it is not the market. Open interest can migrate between exchanges, and a decomposition on one venue reads a fall in positions where there may have been a relocation. Where a conclusion matters, run Check 1 on at least two venues before publishing it.

They measure financing, not conviction. A cash-financed rally is structurally sturdier than a leveraged one; it is not thereby correct. Spot buyers sell too, and a market that went up on cash can go down on cash.

They are backward-looking by construction. Every input is a settled snapshot. The checks describe how the last move was paid for. The most common way to misuse them is to treat “leverage did not build into this rally” as “leverage will not build,” which is a forecast the data does not contain.

Used within those limits, the six checks turn the most common unexamined sentence in crypto commentary into something with a number attached. That is the entire ambition of this series: not to predict the market, but to make the claims about it checkable.

Disclaimer: This is an educational field guide, not investment advice. Bitcoin Mastery is not a financial adviser. Derivatives are leveraged instruments and you can lose more than your initial outlay; spot cryptocurrency is volatile and you can lose the whole of your capital. Every figure used as an example names its source and the time it was taken so you can reproduce it; do your own research and consider taking independent professional advice before acting.