Across 3,927 replayed signals, 1,958 were never a trade a follower could take. One channel accounts for 1,501 of them, which is the more useful finding: the rate varies far more between publishers than the headline suggests.

What we measured

Every signal in our index gets replayed against real exchange candles under one fixed rule set. Enter at the price the channel published, take profit at a 2% move, exit at the channel's own stop or at a 10% fallback, and allow 30 days to resolve. Fees are charged. The full rules are on the Methodology page.

Cancelled is one of the possible outcomes, and it is the subject of this piece. A signal is cancelled when price reached the 2% target zone before the published entry ever filled. The move the channel described happened; the trade it described did not, because the entry was never available on the way there.

That is different from an entry the market simply never reached, which we record separately as a no-fill. Cancelled means the call was overtaken by its own thesis.

Where do the unfillable calls concentrate?

Almost entirely in one channel. Of the 1,958 cancelled outcomes in our corpus as of 17 August 2026, a single publisher accounts for 1,501, which is 78% of everything it posts. Across the other twenty-one channels that publish parseable signals, the rate is 22.8%.

Group, as of August 2026 Outcomes Cancelled Share
The one high-volume channel 1,922 1,501 78.1%
The other twenty-one channels 2,005 457 22.8%
Whole corpus 3,927 1,958 49.9%

Read as a statement about the corpus, the 49% holds. Read as a statement about the channel you are about to pay, it does not — one in five calls being unfillable is the ordinary case here, and four in five is what a single publisher's feed produces.

That publisher posts enough to move the whole dataset on its own.

A second distortion sits in the same table. Three of the twenty-one remaining channels carry nearly identical numbers, at 296 to 297 scored verdicts each and hit rates within a sixth of a point of one another, because they are one signal feed republished under three brands. Any average taken across channels counts that feed three times.

What does the whole corpus look like?

Cancelled is the largest single outcome in the dataset, larger than wins and losses combined in several channels. Across 3,927 resolved outcomes from 3,988 parsed signals, as of August 2026:

Outcome Count Share
Cancelled: price ran past the target before the entry filled 1,958 49.9%
Reached the 2% take before the stop 1,305 33.2%
Stopped out first 578 14.7%
Neither within 30 days 52 1.3%
Entry never reached at all 31 0.8%
No candle data 2 0.1%
Carrying no verdict yet 1 0.0%

Fewer than half of the outcomes produce a verdict. A signal is scored only if it reached the take, hit the stop, or ran out the 30-day window, which comes to 1,935 of the 3,927. That shortfall is again concentrated: strip out the one high-volume channel and the remaining twenty-one score 1,515 of their 2,005 outcomes, or 75.6%. For that one channel, every rating we publish rests on 420 of its 1,922 outcomes.

Direction is not the driver: cancels split 998 short against 960 long, close to the split across all 3,927 resolved outcomes, which is 2,163 short to 1,764 long. Note also that three of the seventeen publishing channels are one feed under three brands, so any figure averaged across channels counts that publisher three times.

Why would an entry go unfilled?

Because a published price is an instruction, and an instruction is not a transaction. A limit order resting at the entry price fills only when three separate conditions hold — Binance documents all three (Why Wasn't My Limit Order Filled?):

  1. The market price reaches your limit price or better.
  2. There is sufficient liquidity at that price.
  3. There is sufficient time for the order to execute.

An entry never filled is what you get when any one of them fails.

The second and third conditions are where a broadcast signal fails. Every subscriber reads the same message and queues at the same price, and exchanges match in price-time priority. Coinbase's developer documentation describes a "continuous first-come, first-serve order book" in which "Orders are executed in price-time priority as received by the matching engine" (Coinbase Exchange matching engine). Whoever arrives first is filled, and the rest inherit whatever depth is left.

That leaves a follower two options, both of them worse than the published record implies. Wait at the entry and risk never trading it, or chase with a market order and accept a worse price than the one the channel will later grade itself against. What happens in that gap is covered in Following crypto trading signals: how it actually works.

Why is our 49% a floor rather than a ceiling?

Because our replay is generous about fills. When a candle's range touches the published entry, we treat the entry as filled, which assumes a real follower would have been served at that price. The exchange's own documentation says that assumption is optimistic:

"During periods of high volatility, your order may not be able to reach the end of the order book for execution, even if the market price reaches your limit price." — Binance, Why Wasn't My Limit Order Filled?

So an unknown number of the signals we scored as filled would not have filled for a real subscriber — particularly the ones published into fast moves, which is precisely when signal channels post. Every one of those belongs in the unfillable column.

The number in this article is a lower bound, then, and our hit rates are computed on a set of trades more favourable than the one a subscriber could have traded. Where the rest of our engine's assumptions sit is set out in What a backtest can and cannot prove.

How fast does a called move actually run?

Fast enough that the question of who reads the message first decides the outcome. The academic literature on coordinated crypto pumps has measured this repeatedly, and the numbers are in seconds and minutes rather than hours.

Xu and Livshits studied 412 pump events organised across more than 300 Telegram channels (The Anatomy of a Cryptocurrency Pump-and-Dump Scheme, USENIX Security 2019). In one case they document, price peaked 18 seconds after the announcement, the first buy order was placed and completed within one second of the message, and price fell back below its opening level three and a half minutes after the pump began. Across 355 pumps, Dhawan and Putnins measured a mean signal-to-peak return of 65.47%, reached on average in about eight minutes and at a median of 1.54 minutes (A new wolf in town? Pump-and-dump manipulation in cryptocurrency markets, working paper, later published in Review of Finance, 2023).

The head start is also priced. Studying 902 pump operations, La Morgia and colleagues found that ranked or paying members received the signal between one and ten seconds ahead of unranked ones, at roughly half a second to a second per level of hierarchy. Ranks sold for between 0.01 and 0.1 BTC (The Doge of Wall Street, ACM TOIT 2023).

Those studies describe deliberate manipulation schemes, which is the extreme end of the spectrum rather than the average signal channel. The mechanism transfers even where the intent does not: when a call moves the market it names, the published entry describes a price that existed before most readers could act on it.

Is that concentration unusual?

No, at least not in the part of this market that has been studied. Concentration is the normal shape there, so a dataset dominated by one publisher is closer to what the literature would predict than to an artefact of our sample.

Hamrick and colleagues examined the pump ecosystem across Telegram and Discord and reported "high levels of concentration in both exchanges employed for pumps and channels involved in running the pumps" (An examination of the cryptocurrency pump and dump ecosystem, Information Processing and Management, 2021). Three Telegram channels ran roughly 45% of the Telegram pumps they observed. That study covers deliberate manipulation schemes, which is the extreme end of this market and not a description of the channels in our index.

The same study found the size of the move depends heavily on what is being pumped, with median increases of 3.5% to 4.8% for coins in the top 75 by capitalisation against 19% to 23% for coins ranked beyond 500. Small illiquid pairs are where both the biggest advertised moves and the thinnest order books live, which is the same place unfillable entries come from.

What does an unfillable call do to a channel's record?

It flatters it twice. A call that never fills costs the follower nothing on paper and costs the channel nothing at all, while still being available to claim afterwards if the market moved in the named direction.

Consider the structure. The channel posts an entry below the current price for a long, and price rises without ever coming back to that level. The move was called correctly, so the post can be presented as a correct read — and no stop was ever risked, because no position was ever opened.

Repeat that often enough and a feed accumulates a record of correct calls, none of which a subscriber could have traded.

We exclude cancelled outcomes from the denominator of our published hit rates. That is the honest treatment and it has an uncomfortable consequence: the percentage then describes a minority of what the channel actually posted. For the channel in this dataset at 78% cancels, our own rating rests on 420 of its 1,922 outcomes, so the number on its page is computed from roughly a fifth of its output. A percentage carrying that little of the record settles very little on its own, which is the subject of Win rate: why 90% accuracy can still lose you money.

What this does not prove

Our corpus covers the channels we can parse, which is 22 of the 47 in the index as of 17 August 2026. Channels that publish screenshots, voice notes or free-form commentary contribute nothing to these counts, and they are not a random sample of what is out there. The 49% describes this dataset, not the market.

We also do not measure intent, and nothing here alleges misconduct by any channel. A high cancel rate fits several explanations at once: a publisher posting entries far from the market to look decisive, a publisher posting after the move, or a genuine strategy of resting limit orders that often go untouched. Our replay records what happened to each call and cannot tell you why the entry sat where it did.

The clone cluster is a further caution: three of the seventeen publishing channels carry the same feed under different brands, so treating the index as seventeen independent publishers overstates how much evidence it contains.

Finally, the figures move. Signals resolve continuously as the replay catches up with the market, and the counts here are a cut taken on August 7, 2026. Treat them as dated, and expect the shares to shift as more channels enter the index.

The practical read

Three questions come out of this data for anyone weighing a channel. What share of its calls ever became a fillable trade. How many scored verdicts sit behind the percentage it advertises. And whether the entries it publishes sit where the market can still reach them when the message arrives.

The first two are on every channel page we publish, at Signal Providers. The third takes ten minutes with the feed and a price chart, and it is the check most subscribers skip. Nothing here is financial advice.

Sources