The median signal trade in our index closes 7.20 hours after publication, and the same corpus scored a different way gives 37. Both numbers are real, and the gap between them is the answer to a question most people never ask.

How long does a signal trade actually last?

Under our one-take rule, the median gap between a signal being posted and the trade closing is 7.20 hours, measured across 1,935 replayed signals in our index as of 17 August 2026. A quarter of them are done inside 1.85 hours. A quarter run past 21.75 hours.

The distribution is short at the front and long at the back.

Closed within Share of the 1,935
1 hour 17.5%
4 hours 38.6%
24 hours 77.4%
longer than 72 hours 9.3%

Nearly four trades in ten are over within four hours of the message, and roughly eight in ten within a day of it. The average is 28.53 hours, four times the median, because a handful of trades sit open for weeks. The longest in the set ran 1,002 hours, which is 42 days.

Why does the same corpus also say 37 hours?

Because "closed" is a decision we make, and we make it two different ways.

The 7.20-hour figure comes from our one-take rule: enter at the price the channel published, take the first 2% move, exit at the channel's own stop or at a 10% fallback if it published none, and allow 30 days. It is deliberately blunt, so that channels with five targets and channels with one are measured on the same clock.

The second rule replays each signal the way it was written, walking the full target ladder the channel published. The clock stops at the last rung, at the stop, or at the end of the window. On that rule, 1,485 outcomes carry a closing time and the median is 36.60 hours, with quartiles at 9.43 and 156.08 hours.

One-take rule As published
Exit at first 2% move, the stop, or a 10% fallback the last rung of the published ladder, or the stop
Outcomes with a closing time 1,935 1,485
Median 7.20 h 36.60 h
Quartiles 1.85 h / 21.75 h 9.43 h / 156.08 h
Closed inside 24 h 77.4% 42.6%
Open longer than 72 h 9.3% 40.4%

Same signals, same index, same window. One rule says a signal trade is a same-day object. The other says a third of them are multi-day positions, with the longest running 2,716 hours. Anyone quoting a single duration for crypto signals is quoting the rule they chose, and usually without saying which one it was.

Which of the two numbers describes your trade?

Whichever one matches how you would actually close the position. If you bank a fixed move and leave, the first column is your world. If you sit through the whole ladder because the advertised return lives at the far end of it, the second column is.

The gap is not a measurement error. It is the cost of the ladder expressed in time: nearly 30 extra hours at the median, and a jump from 9.3% to 40.4% in the share of trades that outlive three days. Whether that time buys anything is a separate question, and the answer depends on how much of the position survives to the last rung.

Do winning trades close faster than losing ones?

Under the one-take rule, yes, and by roughly a factor of two.

One-take outcome Count Median time to close Closed inside 1 h Open past 72 h
Reached the 2% take 1,305 5.77 h 21.1% 7.2%
Stopped out 578 9.43 h 10.9% 8.5%
Neither, within 30 days 52 100.75 h 0.0% 69.2%

A win takes 5.77 hours at the median and a loss takes 9.43. One winning trade in five is over inside the hour, against one losing trade in nine.

The third row is a different kind of object from the two above it. A trade that reaches neither the take nor the stop runs 100.75 hours at the median — seventeen times a winner — and more than two in three of them are still open after three days. Not one closed inside the first hour. On this rule, "still open" is not a neutral state a trade passes through; it is where the slowest 3% of the corpus ends up.

So a position that keeps not closing is, on this rule, statistically the worse position to be holding.

That sentence stops being true the moment the exit rule changes. Replayed as published, the profitable trades take 77.18 hours at the median across 608 outcomes, and the ones that finished flat or negative take 22.86 hours across 877. The ranking inverts completely.

Both results are artefacts of the exit rules, and they are worth understanding rather than memorising. A one-take win only needs the first 2% to print, which happens fast when it happens at all; a one-take loss needs price to travel further, in the wrong direction, to a stop that usually sits further away. Under the ladder, a win has to collect every rung, which takes days, while a loss needs one touch of the stop. The clock is measuring the exit rule at least as much as it is measuring the market.

Does the clock start when the message arrives or when you fill?

Barely matters, and that is itself the finding.

For 1,380 outcomes we have a timestamp for the moment the entry actually filled. Measured from that timestamp instead of from publication, the median is 8.12 hours against 7.20. The remaining 555 of the 1,935 carry no fill time we can use.

The two clocks land within about an hour of each other, so waiting for the entry adds little to the elapsed time — and note that the two figures are not computed over the same rows, since only the trades that filled have a fill time at all. What the numbers on this page measure is the trade, not the queue in front of it. Which order type turns a published price into an actual position is a separate problem, covered in which order type turns a published entry into a position.

What about the outcomes with no closing time at all?

They are the largest group in the dataset, and no duration on this page describes them.

Of 3,927 outcomes in the index, 1,958 are cancelled, which is 49.9%. Cancelled means the move ran into our 2% target zone while the entry the channel published was still sitting unfilled behind it, so there is no position whose life we could measure. Another 31 never reached their entry at all, and 2 have no candle data. Under the as-published rule the shortfall is larger still: 2,035 outcomes are unresolvable and a further 378 cancelled.

Half the index, then, has no answer to the question in the headline. What that does to a channel's advertised record is the subject of half the trades never happen.

It also sets the frame for the hit rate we publish. Among the 1,935 outcomes that do have a closing time, 1,305 reached the take, which is 67.4%. That percentage rests on the half of the corpus that produced a tradeable, timeable trade.

Does the answer change from channel to channel?

By a factor of six, on the same rule and in the same window.

Channel group Closes with a time Median
4C Trading Signals 22 2.28 h
Free Pump Crypto Signals 420 2.67 h
CryptoNinjas Trading 117 5.14 h
The three-channel AI clone feed 296 to 297 each 10.06 h
Rocket Wallet VIP Signals 23 15.20 h

Two cautions belong next to that table, and both are large enough to move the corpus median on their own.

A single feed supplies 46.0% of the closes. Three channels in the index publish an identical stream, 334 of 335 signals in common at 100% overlap, and between them they account for 890 of the 1,935 timed closes. Any average taken across channels counts that one feed three times, and its 10.06-hour median is doing far more work in the 7.20-hour corpus figure than a list of 22 publishers suggests. How that pattern shows up from the outside is set out in clone signal channels.

A single channel supplies 48.4% of the index. Free Pump Crypto Signals posted 1,928 of the 3,988 signals we hold. It contributes only 420 timed closes, 21.7% of them, because most of what it posts is cancelled before it becomes a trade. Its 2.67-hour median is fast, and it is fast on a fifth of what it published.

What does a six-hour median mean for planning?

Not a holding time. Nothing here says how long to stay in a trade, and the two rules above disagree by a factor of four on what the same trades even did.

What both rules agree on is the shape. A signal position is usually resolved within hours to a couple of days, and the fraction that turns into a week-long hold is small on one rule and a third of the sample on the other. Under neither rule is the typical outcome a position you open and revisit next month.

That has an operational consequence worth stating plainly. On the one-take rule, 38.6% of trades are finished within four hours of the message, which means the exit is frequently decided during a window when nobody is watching the chart. Whatever handles that exit has to be resting in the market before the move, because at these durations the alternative is finding out afterwards. Positions taken from a feed you did not build are also worth sizing on that basis, which is worked through in position sizing when the trade idea is not yours.

The second consequence is about reading track records. A channel advertising a run of wins over a month is describing trades that mostly lasted hours. Its record is a count of short events, and short events accumulate fast enough that a month of them can look like a long history while resting on very little time in the market. How many of those events it takes before a percentage means anything is covered in how many trades a hit rate needs before it means anything.

What this does not prove

Half the index has no duration. Every figure here describes outcomes that produced a closing time. The 1,958 cancelled outcomes, 49.9% of the corpus, are absent from all of it, and they are not randomly distributed across channels.

The sample is one summer. Our index spans 2021 to 2026 on paper, and 90.4% of the signals in it were published during 2026. Treat these durations as a description of this window, not of the market across cycles.

Three of the numbers are one feed. The clone cluster contributes 46.0% of the timed closes, and one publisher contributes 48.4% of all signals held. Neither the corpus median nor the channel table should be read as a survey of independent publishers. Only 22 of the 47 accounts we track contribute any parsed signal at all.

The most recent signals have not finished. A signal gets 30 days to resolve, so the ones published in the days before this snapshot are still being replayed as candles arrive; 44 of the 52 that currently sit in the "neither" row have not yet used their full window. Their verdicts, and the durations attached to them, can still change — which is precisely why we recompute rather than freeze them.

The replay is generous about fills. A candle whose range touches the published entry counts as a fill, which flatters both the number of trades we time and, in a direction we cannot establish, the durations we record for them. The rest of what our engine assumes, and what it therefore cannot settle, is laid out in what a backtest can and cannot prove.

These are timestamps, and only timestamps. We measured when trades opened and closed. We did not measure what holding them cost in funding, fees or slippage, and none of those scale with time in the way the durations here might tempt you to assume.

We do not measure why. A slow trade may be a slow market, a wide stop, an illiquid pair or a signal published into nothing. Our replay records the clock and cannot tell you which.

The practical read

Ask any channel quoting a typical trade duration which exit rule the number came from. In our corpus that single choice moves the median from 7.20 hours to 36.60, moves the share of multi-day positions from 9.3% to 40.4%, and reverses the answer to whether winners close faster than losers. Two honest analysts working from the same signals can hand you both numbers.

Then check the denominator. A duration computed over 1,935 outcomes out of 3,927 describes the half of the feed that became a trade, and the other half is where a channel's record tends to hide.

Both figures for every channel we can score are on its page at Signal Providers. Nothing here is a recommendation to hold, close or take any trade.

Sources

  • Duration figures, both rules, the per-result and per-channel breakdowns: index snapshot taken 17 August 2026. Corpus: 3,988 signals, 3,927 outcomes, 47 accounts, 22 of them contributing parsed signals. One-take durations from SignalOutcome.one_take_exit_at; as-published durations from SignalOutcome.exit_at.
  • One-take replay rules and the 2% take, channel stop and 10% fallback: Methodology.
  • Clone overlap figures, 334 of 335 signals in common at 100% overlap across three channels: ChannelCloneLink records in the same snapshot.