Which crypto markets break out next? Five signals tested against 124 pump onsets, the survivors on 153 more
Published by vyx
The sharpest clue of the five we tested is a market already beating BTC while it sits under its 24-hour high: it lit before 66% of 124 pump onsets, where chance caught 35–49%, and on 153 onsets it never saw, each time it switched on a pump followed within 6 hours 2.5× as often as after a random hour. Volume spikes and rising open interest caught no more pumps than chance.
See which crypto markets Quiet leader is lit on right now — the rule from this study, live on every Hyperliquid crypto market, beside the ones breaking out this hour.
The question
Every trader wants to know which market moves next. On a venue with hundreds of perpetuals, a breakout scanner that lit up before most pumps would be worth a lot — but "it lit up before the pump" is the easiest claim in trading to fool yourself with. A signal that is on a third of the time is on before a third of all pumps by luck alone. So we graded five popular ideas the honest way: against real pump onsets, beside what chance catches, and then on a stretch of history they had never seen.
onset: close > prior(highest(high, 24), 1) && quoteVolume > 2 * median(quoteVolume, 48) confirm: fwd(highest(high, 6), 6) / close - 1 > 0.04
A pump onset is the first hourly candle that closes above the last 24 hours' high on at least twice its usual volume — something a trader can see when it happens — and only counts if price then runs at least 4% higher within the next 6 hours. 124 onsets matched on 26 of 28 crypto markets between Sep 3 – Oct 14, 2026.
Five ideas
Each idea is a condition a scanner could watch on every market, every hour. A signal "caught" a pump when it was lit at any point in the 6 hours before the onset.
- Volume waking up — Hourly volume 1.5× its two-day median, price still under the 24h high.
- Near the 24h high — Price within 1.5% under its 24h high.
- Pressing the high — Within 2% of the 24h high and volume waking up.
- Open interest building — Open interest up 3% in 6 hours while price moved less than 2%.
- Quiet leader — Up 1.5% more than BTC over 6 hours, still under its 24h high.
Bars: the share of 124 pump onsets each idea was lit before (up to 6 hours). Gray band: what the same idea catches by chance — its lit pattern moved onto another market at the same hours, or onto other days at the same time of day, 99 seeded draws each; the stricter of the two is shown. A bar inside its band is luck.
The numbers
| Pumps it lit before | Chance (mean) | Chance (95% range) | |
|---|---|---|---|
| Volume waking up | 75% | 78% | 71%–84% |
| Near the 24h high | 73% | 62% | 56%–69% |
| Pressing the high | 55% | 51% | 44%–57% |
| Open interest building | 19% | 22% | 15%–29% |
| Quiet leader | 66% | 49% | 38%–60% |
Busy signals catch pumps by luck
"Volume waking up" was lit before 75% of the pumps — the highest catch rate of the five — and it means nothing: the same pattern on another market at the same hours caught 78%. Volume rises across the whole board at once, so a volume alert is on before most pumps simply because it is on often (29% of all market-hours) and on many markets at once. Rising open interest with flat price did the opposite: it was rarely lit at all and caught 19%, again no better than chance. Adding volume to "near the high" made it worse, not better (55% caught, inside its chance band).
At the hours pumps began: how often each idea was lit on the market that pumped, against the markets that did not move at the same hours. A signal that lights the whole board scores the same on both.
The numbers
| Markets that pumped | Quiet markets, same hours | |
|---|---|---|
| Volume waking up | 75% | 76% |
| Near the 24h high | 73% | 60% |
| Pressing the high | 55% | 47% |
| Open interest building | 19% | 20% |
| Quiet leader | 66% | 31% |
This is where the two survivors separate from the rest. "Near the 24h high" was lit on 73% of the markets that pumped — but also on 60% of the ones that didn't. "Quiet leader" was lit on 66% of the pumping markets and only 31% of the quiet ones: it points at the right market, not just the right hour.
Precision against a random hour: the share of times a signal switched on that a pump onset followed within 6 hours, divided by the same share for a random market-hour (2.9%).
The numbers
| Pump rate after it lights, vs a random hour | |
|---|---|
| Volume waking up | 1.0× |
| Near the 24h high | 1.9× |
| Pressing the high | 1.5× |
| Open interest building | 1.2× |
| Quiet leader | 2.1× |
On history it never saw
Testing five ideas and keeping the winners is exactly how a backtest lies, so both survivors were graded again on the Jul 23 – Sep 3, 2026 window — 153 different pump onsets on the same markets. Both held: "Near the 24h high" at 2.1× a random hour and "Quiet leader" at 2.5×, each beating both chance checks again. Near the 24h high catches more pumps (88% against 60%) but is lit on 35.2% of market-hours; Quiet leader is lit on 8.6% — about a quarter as often — and each time it switches on, a pump is likelier to follow (2.5× against 2.1×).
Pump rate after the signal lights, against a random hour, on the window the ideas were chosen on and on the earlier window they never saw.
The numbers
| Tested · Sep 3 – Oct 14, 2026 | Unseen · Jul 23 – Sep 3, 2026 | |
|---|---|---|
| Near the 24h high | 1.9× | 2.1× |
| Quiet leader | 2.1× | 2.5× |
ARB's hourly close around its breakout, as % from the breakout candle's close. Shaded hours: Quiet leader lit — 4 hours on and off through the day before, most recently 3 hours before the breakout candle. Within the next 6 hours price traded 31.3% above the breakout close.
The numbers
| Hour | % from the breakout close | Quiet leader lit |
|---|---|---|
| -18 | -5.4% | |
| -17 | -5.95% | |
| -16 | -5.38% | |
| -15 | -5.25% | |
| -14 | -5.73% | |
| -13 | -6.48% | |
| -12 | -6.28% | |
| -11 | -4.16% | yes |
| -10 | -4.35% | |
| -9 | -5.38% | |
| -8 | -2.57% | yes |
| -7 | -4.05% | yes |
| -6 | -5.23% | |
| -5 | -3.45% | |
| -4 | -4.44% | |
| -3 | -3.47% | yes |
| -2 | -4.85% | |
| -1 | -2.67% | |
| 0 | 0% | |
| +1 | +2.93% | |
| +2 | +7.27% | |
| +3 | +16.14% | |
| +4 | +23.62% | |
| +5 | +28.4% | |
| +6 | +27.16% | yes |
| +7 | +38.7% | |
| +8 | +38.5% | yes |
Of the 82 pumps Quiet leader lit before (66% of all onsets in the tested window), how many hours ahead it first lit. 28 were lit the full 6 hours ahead — the market had been leading for a while.
The numbers
| Hours before the pump began | Pumps |
|---|---|
| 1h | 15 |
| 2h | 10 |
| 3h | 12 |
| 4h | 5 |
| 5h | 12 |
| 6h | 28 |
sum(residBps, 6) > 150 close < prior(highest(high, 24), 1)
residBps is the market's move each hour after taking out what BTC did (its beta to BTC), in basis points; summed over 6 hours, above 150 means it beat BTC by 1.5%. The second line keeps it on markets that have not broken out yet. Both must hold.
Even the best idea is followed by a pump only 6–8.4% of the times it lights — far better than a random hour, and still mostly quiet. It tells you which markets deserve a look; it does not tell you to buy. This study measured events, not trades: no entries, exits or costs.
Run this study yourself
Every number here comes from vyx's Proof graded against an event — the same check any trader can run on their own signals, from the desk chat or a connected AI. Describe the event (a visible onset plus a forward confirm, a saved screen, or your own list of alerts), name the signal, and the answer comes back with what it caught, how early, how often it was wrong, and what chance catches beside it.
- In the desk chat: "Grade Quiet leader against pump onsets — a close above the 24h high on double volume, 4% higher within 6 hours — on 1h crypto."
- Over the connector:
vyx_validate_signalwithtarget: {onset, confirm},interval: "1h",classes: ["crypto"],lead: 6— addtoto grade an older window.
How this was done
The same method as every study here — how vyx tests a signal.
- Venue and data
- Hyperliquid perpetuals, hourly candles from vyx's own store (price, volume, open interest, and the residual against BTC computed per market).
- Markets
- The 28 liquid crypto perpetuals on vyx's board.
- Windows
- Tested: Sep 3 – Oct 14, 2026 (1,000 hourly candles). Unseen: Jul 23 – Sep 3, 2026 (the 1,000 before).
- Caught
- Lit at any point in the 6 hours before an onset (early); lit at the onset counts too, later does not.
- Chance
- Two placebos, 99 seeded draws each: the signal's lit pattern from another market at the same hours, and the signal's own pattern on other days at the same time of day. Both keep how markets move together and the daily rhythm.
- Verdict
- Validated only when both the catch rate and the precision beat both placebos under a Benjamini–Hochberg false-discovery gate (q = 0.10), with at least 30 events.
- Reproduce
- The study re-runs on request: every number on this page is regenerated from the same grade on the same pinned windows, and the desk linked above runs the very rules that were graded.
What it does not show
- Five ideas were tried. The unseen window is the guard against keeping a lucky winner; both survivors passed it, the other three were never re-tested.
- The markets are the 28 liquid crypto perpetuals vyx counts as tradable today — a market delisted or illiquid since is not in either window.
- One stretch of market history (Jul 23 – Oct 14, 2026). A different regime could rank these ideas differently.
- Events, not trades: no entry, exit, sizing or fees were modelled, and nothing here says which direction a market will move after the pump.
Questions
Can you predict which crypto will pump next?
Not with certainty, but some clues beat chance. In this study, each time a market started outperforming BTC by 1.5% over 6 hours while still under its 24-hour high, a pump onset followed within 6 hours 2.1–2.5× as often as after a random market-hour on Hyperliquid — still only 6–8.4% of the time.
Does high volume predict a crypto breakout?
Not on its own. Hourly volume 1.5× its usual level was lit before 75% of pump onsets, but the same pattern taken from another market at the same hours caught 78%: volume rises across the whole market at once, so it says little about which coin breaks out.
What is a pump onset?
Here: the first hourly candle that closes above the previous 24 hours' high on at least twice its usual volume, counted only if price then trades at least 4% higher within the next 6 hours.
Why compare a trading signal to chance?
Because a signal that is switched on often is on before many events by luck. Comparing it to the same pattern moved onto another market, or onto other days, shows how much of its hit rate is real — a 75% catch rate means nothing if chance catches 78%.
What does it mean for a coin to beat BTC?
Its own move after removing the part explained by Bitcoin. vyx computes it per market as the residual return against BTC (residBps); a positive sum over a few hours means the coin is rising on its own, not just with the market.
How can I test my own signal like this?
On vyx, ask the desk chat or a connected AI to grade the signal against an event you define — a visible onset plus a forward confirm, or your own list of alerts. The Proof returns what it caught, how early, how often it was wrong, and what chance catches beside it.
Watch it on today’s board
See which crypto markets Quiet leader is lit on right now — the rule from this study, live on every Hyperliquid crypto market, beside the ones breaking out this hour.
Open the deskFree, no install. Or grade your own signal the same way — how vyx tests a signal.