Hash Ribbons
Descriptive, oversold as prediction
When the 30-day average of Bitcoin’s hash rate falls below the 60-day average and then recovers, miners are said to have capitulated; confirmed by a short-term price crossover, the recovery is read as a buy signal.
- Return 12 months after each Hash Ribbons signal (%)
- Median after a price-crossover-only signal (106.3%)
- Median after a random buy day (98.2%)
Our own calculation on Coin Metrics Community Data (BTC reference rate and hash rate, daily, through 14 August 2026). The published rule produced 34 signals in total; the 29 shown here are those with a full year of subsequent data. We read the rule literally: capitulation ends when the 30-day hash-rate average crosses back above the 60-day, and the buy prints on the first day from then on which the 10-day price average is at or above the 20-day. The baselines are medians across all 5507 possible buy days and across the 136 signals produced by the price crossover with the hash-rate condition removed.
How it is built
Two moving averages of network hash rate, and the dates on which one crosses the other. Unlike the price-only constructions elsewhere in this study, the input is a physical quantity: how much computation is securing the network.
Our verdict
Descriptively useful, oversold as a timing tool. It is the only construction in this study built on a physical quantity rather than on price alone, and the mechanism behind it is real — this page said so when the numbers were kinder, and says it still. But when we removed the hash-rate condition, the remaining price rule did better than the full one. What is sold as a mining signal performs like a momentum signal carrying a mining story, and in the last two years the story has cost it.
The record
Every line below is our own calculation on the public price series — reproducible where it is our own calculation, and linked to its primary source where it is quotation.
| When | What the chart says | What the data shows | Source |
|---|---|---|---|
| 2019-10-30 | Hash Ribbons has around twenty signals since 2011, roughly 85 per cent of them correct, with a maximum drawdown near 15 per cent — the figures repeated across the coverage. | The original says something else. Edwards’ own article counts “9 historic buy signals”, an average gain to the next cycle peak of “over 5000%”, and an average maximum drawdown of “just 11%”. Two of the three numbers in circulation are not his. The publication date is wrong too: the secondary sources say August 2019, the article is dated 30 October 2019. A chart whose headline statistics drift as they are retold is a fair test of how carefully it is being read. | Charles Edwards, “Hash Ribbons & Bitcoin Bottoms”, Capriole Investments, 30 October 2019 (edited 13 April 2023) · archived |
| 2011 → 2026 | The published rule combines two conditions: the 30-day hash-rate average recovering above the 60-day average, confirmed by the 10-day price average crossing above the 20-day. | Recomputed on the full daily series, that rule produces 34 signals, of which 29 are old enough to measure a full year afterwards. It is a small sample from which to draw conclusions. | Moonkelp’s own analysis |
| 2011 → 2026 | Buying on the signal beats buying without one. | It no longer does. The median twelve-month return after a signal is 86.1 per cent, against 98.2 per cent for a random buy day in the same series; the hit rate is 69 against 73 per cent. Through May 2026 the signal was still modestly ahead — the 2025 cohort is what changed it. | Moonkelp’s own analysis |
| 2025 | A recovering hash rate marks the end of forced selling and the start of the next leg up. | Six signals fired between February and July 2025, at prices from $87,323 to $123,443. Every one of them is down twelve months later, by 19 to 45 per cent. This is the first sustained run of failures in the indicator’s history, and it happened in the cycle in which Bitcoin made its all-time high. | Moonkelp’s own analysis |
| 2011 → 2026 | The hash-rate component is what makes the signal work — it is the entire premise. | We removed it and kept only the price crossover. That alone produced 136 signals with a median twelve-month return of 106.3 per cent — better than the full rule and better than a random day. Adding the mining condition to the price rule did not improve it; it subtracted from it. | Moonkelp’s own analysis |
Why the method does not hold
01
The mechanism is real, and this page says so first: when price falls below the cost of production, the least efficient miners switch off and their forced selling ends. That is an economic event, not a shape on a chart.
02
The problem is attribution, and it has got worse rather than better. The price crossover contained inside the rule beats a random buy day on its own; the full rule, with the hash-rate condition added, does not. The component the indicator is named after is not carrying the result — on our recomputation it is now costing it.
03
Miner economics have changed as the industry industrialised and hedged, so the behaviour the signal keys on is not the behaviour it was calibrated to.
The strongest case for it
This one still has the strongest case in the study, and the mechanism deserves the credit: when price falls below the cost of production, marginal miners switch off, and the forced selling that comes with them stops. That is a real economic event with a real effect on supply, and no other chart here is built on one. The defence would add that twelve months is an arbitrary horizon and that 29 observations cannot settle anything — both fair. What the defence cannot do is explain why the mining half of the rule subtracts from the price half rather than adding to it.
What to look at instead
Read it as one input on stress, alongside gauges that describe valuation directly rather than inferring a moment.
Questions
Do Hash Ribbons actually work?
Less well than it used to, and less because of mining than because of price. On data through August 2026 the published rule returned a median 86.1 per cent over twelve months, against 98.2 per cent for a random buy day — it now trails the baseline. Stripping out the hash-rate condition and keeping only the price crossover gave 106.3 per cent, better than both. The component the indicator is named after is not what makes it work.
Why do the backtests look so good?
Because Bitcoin rose enormously over the test period. Buying on a random day in the same series was followed by a higher price a year later 73 per cent of the time, so “profitable most of the time” mostly describes the asset. Any signal has to be measured against that baseline, not against zero — and against that baseline this one no longer clears.
So is it useless?
No — and this is the one page in this study that says so. Miner capitulation is a real economic event: when price falls below the cost of production, marginal miners switch off and their forced selling ends. As context on stress that is worth watching. What our test does not support is the leap from that to a buy date, and the six signals of 2025 — every one of them down a year later, by 19 to 45 per cent — are the clearest illustration yet of the difference.
Sources & method
Every figure on this page comes from one public price series and arithmetic anyone can repeat. Where a claim would need a source we could not verify, it is not on this page.
- Coin Metrics Community Network Data — BTC reference rate and hash rate, daily — the price and hash-rate series every figure on this page is computed from
- Moonkelp’s own analysis — the calculation itself — method described above, reproducible from the series
- Charles Edwards, “Hash Ribbons & Bitcoin Bottoms”, Capriole Investments, 30 October 2019 (edited 13 April 2023) · archived — the indicator’s original publication — the source of the definition and of the numbers usually quoted second-hand
The other charts
Perspectives, not investment advice. This page criticises a method, not the people who publish it — and it says so where a construction does something real. How Moonkelp works
This study is reviewed quarterly; corrections normally ship within days. Corrections:contact@moonkelp.com