Volatility rises — and fades. We tell you ahead of time.
You cannot guess where the price will go: across 2088 tests our own model scored 51.20%, while "always up" scored 50.04%. That is a coin flip, so we do not predict price. Volatility, however, is measurable. When it spikes into the top 15% of a coin's annual norm, a signal fires. From 2019 to 2026 there were 15,025 such signals, and after 87% of them volatility fell. Counted by independent days — 81% of 1,734. Both figures are honest; only the way they are counted differs.
Why two figures? Within a single day, signals from adjacent hours almost repeat each other. So we also count strictly: one day, one test. There are 1,734 such days, and they confirmed in 81% of cases. 87% is the broad count (all 15,025 signals); 81% is the strict one. Both figures are honest.
What exactly we measure
You cannot guess where the price will go. Across 2088 non-overlapping windows our own model called the direction correctly in 51.20% of cases, while "always up" gave 50.04%. That is a coin flip, so we do not sell price. But another quantity is measurable — volatility. That is what we test.
Of 15,025 tested signals, after 87% volatility fell. The strict count by independent days is 81% (1,734 tests).
Median drop in volatility after a signal. Not a promise of profit, but a change in a measured quantity.
Volatility must land in the top 15% of a coin's own history for the year. This happens rarely — so signals are few.
Volatility is measured over 5 hours and checked 3 hours after the signal. The measurement canon is never changed without recalculating the whole history.
87% is the broad count (all 15,025 signals on hourly data, 2019–2026). Adjacent hours of one day almost repeat each other, so each day is additionally counted once: 1,734 tests, 81% confirmation. We show both figures openly — as a count of events, not a promise of profit.
How to earn from falling volatility
How it works is laid out above. Below is the practice: what exactly generates income, how to open a position and when to close it.
While volatility is high, options are expensive. Selling brings a premium. When volatility fades, options get cheaper. The income is the difference in premiums.
The coin's volatility moved into the top 15% of its history for the year. Across 15,025 tests (87%) it faded after the signal; by independent days — 81%.
Median drop in realized volatility after a signal. This is a measured quantity, not a promise of profit.
Via our link (no identity verification, KYC):
Other venues without identity verification (KYC):
Major exchanges (verification required):
Risk. This is not passive income and not a recommendation to buy or sell. Selling volatility is a short position without a stop order: on a sharp price move the loss can exceed the premium received. In every third three-day window the price moves more than 10% from the entry level. The historical 87% characterizes the behavior of volatility, and does not guarantee a profit on every trade.
Evidence
Here are all closed signals — both confirmed and not. The result is computed from actual exchange candles over a one-hour horizon; the administrator does not interfere. Losses stay in the showcase alongside wins.
| Instrument | Price, $ | 1 h | 24 h | 7 d | 7d trend |
|---|
Signals and their actual results
The headline of every signal is visible to all — that is the showcase. The numbers (entry, target, stop, write-up) open after you pay for a subscription. Closed signals stay in the showcase together with their actual result, so the history can be verified rather than taken on faith. An unconfirmed signal cannot be deleted: it is closed on actual candles and stays in the statistics.
Why you cannot trust percentages in this niche
Advertising promises one thing, independent review shows another. Below are numbers that are not ours: third-party measurements of the signals market, cited without any benefit to us.
Across 5,870 closed and verified signals from 145 providers. A half is not "almost always" — it is a random outcome.
The average provider in the industry takes a net loss, even while being right more often than not.
The gap between the best and the industry average is 2.5 p.p. Provider choice matters, but it is not 90% accuracy.
That many p.p. lower is what independent review usually finds versus a claimed "90%+ accuracy". Our gap is zero: the numbers and the method are open.
Sources: signaltrack.io (measurement across 5,870 closed signals), cryptosignals.guide/accuracy-tracker (updated 30.07.2026).
- ✕We do not sell price direction. Direction was called correctly in 51.20% of cases across 2088 non-overlapping windows. Selling that would be misleading.
- ✕We do not treat a hit rate as a metric without accounting for risk/reward. 78% of hits with small targets and large stops are unprofitable; 55% with a 2:1 ratio are profitable. Look at the average outcome, not the percentage.
- ✕We do not rig the entry level. A classic trick: "entry from 3%" and the result is measured from the most favorable edge. Our entry is fixed by the exchange price at the moment of publication, and it cannot be rewritten after the fact.
- ✕We do not publish reviews or ratings. We have none, and we will not publish fabricated ones. There is a verifiable history — that is what we offer to inspect.
Subscription
A subscription opens the numbers for each signal — entry, target, stop and the write-up — and new signals as they appear. The closed history, actual results and statistics remain open to everyone regardless of subscription: without a verifiable history the price of a subscription would be a promise, and we do not trade in promises.
Two signal lines. “BTC reversion” — when a coin has moved away from Bitcoin more than usual: an intraday trade back toward the market. “Volatility” — when a volatility spike lands in the top 15% of the past year: on average it fades afterwards. Each subscription opens one line; need both — subscribe to both.
Payment is in cryptocurrency, the invoice is created automatically: the address and amount are shown in your account. No manual transfer needed.
Frequently asked questions
We answer plainly, including the unpleasant parts.
Is this a price prediction?
No, we do not sell price direction. It is called correctly about half the time: our own model across 2088 non-overlapping windows scored 51.20%. We measure a different quantity — realized volatility.
What does 87% mean?
Of 15,025 tested signals (hourly data, 2019–2026), after 87% volatility fell. Adjacent hours of one day almost repeat each other, so we additionally count each day once — 1,734 independent tests, and there confirmation is 81%. We show both figures openly.
This is a count of events, not a promise of profit. A confidence interval for the broad count was not calculated — adjacent events partially overlap.
What happens if a signal is not confirmed?
It stays in the shared history and enters the statistics. By independent days roughly every fifth signal is not confirmed — that is baked into the 81% figure. It cannot be deleted, and that is exactly why the statistics are verifiable.
How is a signal's result computed?
The entry price is fixed on the exchange at the moment of publication, and is not taken from a draft and not rigged. The result is computed from actual candles over the window from publication to the end of the horizon. The recount runs automatically, so the history cannot stay empty and does not depend on our wishes.
The showcase is empty. Is something broken?
More likely normal. The signal condition is deliberately strict: volatility must land in the top 15% of a coin's own history for the year. With volatility below the threshold such moments are almost nonexistent, and we leave the showcase empty rather than publish signals just for the sake of volume. If there are no signals for a long time, that is a reason to check the exchange data.
Is this investment advice?
No. We show measured statistics, and we do not advise what to buy and sell. Cryptocurrencies are a volatile asset, and a loss on an individual signal is a normal part of the method, not a sign of error.
How a signal is computed
Realized volatility is calculated from logarithmic returns: a 5-hour window is taken, the standard deviation is found and multiplied by the square root of the number of hours in a year. This yields a comparable quantity for any coin — instruments are measured on a single scale.
The quantity is then compared with the coin's own history for the year. If it lands in the top 15% (above the 85th percentile), the coin has not shown that level of volatility for almost the whole year — and that is the signal. After 3 hours the actual candles are fixed: did volatility fall or not.
The measurement canon is fixed: window 5 hours, horizon 3 hours, threshold 0.85, history depth 1 year. It is not changed without recalculating the whole history — otherwise one method would be measured on old statistics while a new one was published.
Adjacent hours of one day almost repeat each other, so for an independent estimate each day is counted once: 1,734 tests give 81% confirmations.
On the coin page a linear-regression corridor over the log of price is additionally drawn. This describes movement that has already happened, not a prediction: the model's range is shown in full, including the cases where it is wrong.
Limitations
This is not investment advice. We do not promise profit and we do not sell a hit rate. The measured 87% (81% by independent days) is a property of volatility, not a guarantee of income: on an individual signal a loss is entirely possible, and roughly every fifth signal is not confirmed.
Check the history, not the promises. Cryptocurrencies are a volatile asset: a ±20% swing in a week is normal, not a model error. And do not carry these figures over to other venues and other exchanges — our history is collected on Bybit, and we have not proven equivalence with other sources.