Investigative
When spot buys and leverage sells
This research is for members
The full briefing is available to Insider subscribers and to clients enrolled on an Edge programme.
The Insider is £400 a year, or £40 a month — monthly outlooks on publication, quarterly research, and regulatory briefings as the UK regime develops.
Subscribe To The InsiderBTCUSDT · Binance spot and perpetual · 30-minute record · For members
Whether the two markets disagree tells you almost nothing about which way price goes next; what little it tells you, spot was already saying on its own.
The proposition
Bitcoin trades in two places at once: the spot market, where coins change hands, and the perpetual futures market, where leveraged positions do. Order-flow platforms draw a cumulative delta line for each — a running total of aggressive buying less aggressive selling — and when the two lines split, one rising while the other falls, the split is read as a turn. The belief: spot flow is real demand, leveraged flow is the crowd about to be squeezed, and when they diverge, follow spot.
It is a good story. It was checked.
What makes it testable
Both markets publish, for every half-hour bar since January 2020, how much of the volume was bought aggressively. From that, each bar has a delta and each day a running total, on each venue. A divergence is defined before any result is seen: over six hours, spot’s total rises while the perpetual’s falls, each by more than half its own typical six-hour swing, or the mirror. That fired 1,144 times in the bullish direction and 1,380 in the bearish, across six and three-quarter years.
From each firing, two prices race: one a fixed distance above, one the same distance below. Whichever trades first settles it. The distance is set at three, six and twelve times the recent half-hour range, so the result can be read at three horizons.
The result is compared against three things. Random bars from the same years. Random bars that had just made the same move price made before the firing, so that a bounce after a drop is not credited to the delta. And the same signal built from spot alone, with the perpetual ignored — because if spot on its own does the work, the disagreement between the markets was never the point.
What the split predicted
| direction | distance | followed spot | random bars | spot alone |
|---|---|---|---|---|
| bullish split | 3× | 50.4% | 48.5% | 49.9% |
| bullish split | 6× | 52.1% | 48.3% | 51.5% |
| bullish split | 12× | 56.5% | 52.4% | 54.1% |
| bearish split | 3× | 48.5% | 52.3% | 48.5% |
| bearish split | 6× | 47.2% | 51.2% | 48.5% |
| bearish split | 12× | 42.4% | 47.9% | 44.3% |
The bullish split does carry something: about four points over random bars at the two longer horizons, and at that sample size the gap is not luck. But spot alone carries three of those four points. The disagreement between the markets — the thing the panel is drawn to show — is worth roughly one point.
The bearish split carries the opposite of what it claims. When spot sells and leverage buys, price goes up more often than random bars would, at every horizon. Read together, the two arms are not a divergence signal. They are the same fact twice: spot flow leaned upward in a record that rose from 7,000 to 87,000, and any signal that includes “spot is buying” inherits that lean.
The year decides
| bullish split, 6× | 2020 | 2021 | 2022 | 2023 | 2024 | 2025 | 2026 |
|---|---|---|---|---|---|---|---|
| followed spot | 68.2% | 46.1% | 38.2% | 57.3% | 52.2% | 53.0% | 50.0% |
The signal scored 68 in 2020 and 38 in 2022. It did not learn something about turns and forget it; 2020 rose and 2022 fell, and the signal is a way of saying “spot was buying”, which is truer in rising years. A signal that tracks the year is the year.
What this is worth
Before the test, the expectation on record was that the split would show nothing beyond what single-venue delta already showed — and single-venue delta had already been measured at nothing. The result is slightly better than that and not usefully so: a point of edge, present in one direction, absent and inverted in the other, riding on the market’s long drift.
The practical consequence is for the screen, not the trade. A multi-market delta panel is a way of watching drift with more lines. The recording of aggressive flow has other uses — execution, and the placing of resting orders against visible interest — but as a reading of where price goes next, it goes in the pile with the footprint and the single line: consulted, described, not counted.
The usual honesty applies: one definition of divergence was tested, at one window and one threshold, fixed before the numbers; the perpetual record begins in 2020 and the spot record earlier, so the sample is the overlap; and the placebo panels were drawn two hundred times rather than ten thousand, which is more than enough at these margins and would not rescue a one-point effect at any size.
The view from the edge
Spot-versus-perp divergence carries no directional information of its own. The small edge it appears to have is spot buying in a rising market, which is drift, and the mirror case points the wrong way.
Levels are areas of interest, not trade instructions.
Alderlux provides education, research and succession planning services in relation to Bitcoin and digital assets. We do not provide regulated financial advice, investment recommendations, legal or tax advice. Alderlux is not authorised by the Financial Conduct Authority and does not provide cryptoasset custody, dealing or arranging services. Bitcoin, cryptocurrencies and digital assets are speculative and volatile assets. Prices may rise or fall significantly within short periods of time. Investors may experience substantial losses, including the loss of their entire investment. Past performance is not indicative of future results.