Study · Oct 7, 2026
The track-record edge disappears live: we tracked 29 Telegram signal channels
We tracked 29 signal channels live: their track record made +0.12 R per trade; live, they lost −0.03 R. Why it happens and what to check before following one.
By Julián Cardona, founder of Verpips · Reviewed Oct 7, 2026 · How we measure
The short answer
The track-record edge disappears live. We measured the track record of 29 Telegram signal channels (what you see when you join the channel), then followed them live, signal by signal, from the second each one was posted. Their track record averaged +0.12 R per trade; live, they averaged −0.03 R. The gap, 0.15 R per trade, has a 95% interval of 0.02 to 0.29: it’s real, even if we don’t know its exact size. Risking 1% per trade, that’s the difference between making 12% every 100 trades and losing 3%.
In 20 of the 29 channels, the live result was worse than the track record. And the track record barely predicts live results: of the 21 channels that were profitable on their track record, only 11 were profitable live.
How we measured it
We measured each channel twice, with the same engine, the same rules and the same broker prices, tick by tick and including the spread:
- The track record: signals posted before we started following the channel, read afterward, exactly as they appear in the channel. It’s what anyone who joins sees: whatever was deleted is gone, and edited posts show their latest version. 99% of those signals were read more than a day after they were posted.
- Live: signals posted from then on, read the instant they went up and saved in their first version, even if the channel later edits or deletes them.
That’s 29 channels measured live for anywhere from a few days to 5 weeks (20 public and 9 private): 2,967 track-record trades and 527 live trades, almost all in gold. A channel counts if it has at least 10 measured trades in its track record and 5 live. Every channel carries the same weight, and the interval comes from resampling channels 10,000 times. Results are in R, multiples of the risk on each trade (what is R?).
We wrote the method down before looking at any results. And we don’t give an “X times better” figure: with the live result below zero, that ratio means nothing. The gap in R does. How we measure every channel.
Finding 1
The track record makes promises the live results don’t keep
R per trade: track record vs. live (average across channels)
- All 29 channelsRecord+0.12 RLive−0.03 R
- 20 public channelsRecord+0.20 RLive+0.01 R
- 9 private channelsRecord−0.04 RLive−0.12 R
| Track record | Live | |
|---|---|---|
| R per trade (average across channels) | +0.12 R (+0.06 to +0.19) | −0.03 R (−0.15 to +0.10) |
| R per trade (all trades pooled) | +0.10 R | −0.09 R |
| Channels in profit | 21 of 29 | 11 of 29 |
| Trades that end in profit | 66% | 51% |
95% interval in parentheses.
Channels vary a lot: the middle half get worse by between −0.01 R and +0.32 R per trade, with a median of +0.07 R. Some do better live, but they’re the minority. The ranking doesn’t hold either: the correlation between a channel’s rank on its track record and its rank live is 0.14, close to none. The best track record doesn’t point to the best channel live.
More in the glossary: What is backtesting? · What is expectancy?
Finding 2
Why it happens, cause by cause
In the 20 public channels we can measure almost everything: their track record made +0.20 R per trade and live they made +0.01 R, a gap of 0.19 R. Here’s how it breaks down, as far as it can be measured:
| Cause | How much it explains | How we know |
|---|---|---|
| Picking the channel for its good track record | 0.07 to 0.11 R (35–60%) | We split 89 track records in two. The ones above +0.10 R in the first half (average +0.26 R) made +0.15 R in the second |
| Signals the channel deletes afterward | 0.04 R at least (almost 25%) | 10 of 146 signals seen live were deleted later; they averaged −0.61 R and 6 were stop-outs |
| Edits that change the signal after the result | none seen | None of the 527 live signals changed levels after closing while we were watching |
| A change of strategy | no sign of it | The median stop live is the same as in the track record; only 3 of 29 channels change it by more than 30% |
| The market in these weeks, and chance | the rest, 0.03 to 0.08 R | Within the interval |
The biggest cause isn’t a scam: it’s how channels get picked. Anyone shopping for a channel picks the one with the best track record, and a good track record is partly luck. When the luck runs out, results fall back toward the average. You can see it inside the track records themselves: from one half to the other, the 89 channels as a group drop only 0.02 R (from +0.12 R to +0.10 R); the ones you would have picked on their first half drop 0.11 (0.06 to 0.16). Nearly all of that drop comes from the picking.
The second cause is on the channel: deleting. Across the 29 channels, 19 of the 527 signals we saw live disappeared later (3.6%), in 13 different channels. Two out of three were stop-outs (13 of 19), and on average they lost −0.59 R. Whoever joins later never sees them, so the track record comes out clean. It’s a floor: we only watch for deletions for a while, and a channel can delete months later.
An “edited” label sounds worse than it is. In the private channels, 189 of 381 signals carried the “edited” label, but in 168 the text hadn’t changed by a single letter. Only 9 changed levels, in 2 channels, and always before the trade closed: fixing a typo isn’t rewriting the result.
For the private channels we can’t measure how much picking them matters, so this table doesn’t apply to them. Their gap on its own (−0.04 R track record vs. −0.12 R live, over a median of 4.4 weeks) points the same way, but its interval, −0.03 R to +0.20 R, includes zero.
More in the glossary: What is survivorship bias?
Finding 3
What they announce isn’t what happens
There’s a third layer: the channel’s own posts. We matched every “target hit” and every “stop hit” they posted with what actually happened to that signal.
| Track record | Live | |
|---|---|---|
| Trades that win, according to their posts | 81% | 78% |
| Trades that win, measured against price | 66% | 51% |
| Stops measured / acknowledged by the channel | 886 / 264 (30%) | 202 / 46 (23%) |
| Targets measured / announced | 1,905 / 997 (52%) | 253 / 157 (62%) |
| “Target hit” posted on a signal that ended at the stop | 131 | 38 |
Channels announce more than half of their targets and acknowledge one stop in four. In the public channels it’s even starker: of 365 stops in their track records they acknowledged 10, and of the 47 we measured live, none. To rule out a flaw in our reader, we also counted in the raw text, with no interpretation, across the 63 public channels we watch (the 20 in this study come from them): posts about targets being hit outnumber posts about stops 13 to 1, and 19 of the 63 never posted a single stop.
Our study of gold signals already counted how many targets and stops get announced, and our TP2 and TP3 study how many are actually hit. The new number is a different one: of the stops that did happen, how many get acknowledged.
More in the glossary: What is win rate? · What is a stop loss?
What this means for you: how to evaluate a channel
- Don’t judge a channel by its track record. In this sample it promised +0.12 R and delivered −0.03 R. Use it as a filter (a bad track record rules a channel out), not as a promise.
- Measure it live, and for long enough. One trade has a standard deviation of 1.14 R. To know a channel’s result within ±0.2 R you need about 124 live trades: about 12 weeks at a typical pace of 10 signals a week. For ±0.1 R, about 497: almost a year. And your own live results arrive late on top of that: how much you lose copying signals late.
- Save every signal when it’s posted. A timestamped screenshot is enough. 13 of 29 channels deleted at least one signal within a few weeks, mostly stop-outs: 13 of the 19 deleted.
- Count the stops, not the announcements. Compare the stops price actually hit with the ones the channel acknowledged. If it acknowledges fewer than half, its win rate is marketing.
- Look at what an edit changes, not at the label. What matters is a change to the entry, stop or target after the result. The “edited” label on its own almost never is.
- The brighter the track record, the harder it falls. Channels above +0.10 R in the first half of their track record lost almost half of that edge in the second half.
The eight checks in how to tell if a signal channel is legit still apply; this adds the one that counts most: measure it live before you believe it.
What we can’t conclude
- The live period is short. One to eight days in the public channels; up to 5 weeks in the private ones. Every live period falls in the same weeks of the market: if those weeks were bad for gold, they pull every channel down at once, which is why the real interval is wider than the one we calculated.
- With a bigger sample per channel, the number holds but the interval widens. Requiring 10 live trades per channel leaves 12: the gap is similar (+0.16 R), but its interval, −0.02 R to +0.39 R, now includes zero.
- The sample was chosen. The public channels were followed because they had a good track record; the private ones, by our own choice. It isn’t a random sample of Telegram.
- Deletions and edits are floors. We only see them while we’re watching each post; the track record may have lost more before we got there.
- Which channel is cheating. This is an aggregate: picking on luck isn’t the channel’s fault, and neither is a corrected typo.
- These are past results: they don’t predict future results and aren’t an investment recommendation.
Measured on Oct 7, 2026.
Frequently asked questions
- Are trading signal channels fake?
- They don’t have to be for their track record to mislead you. Most of the gap comes from picking the channel that did best in the past, which was also lucky. But the channel plays its part: in our measurements it acknowledges one stop in four, and 13 of 29 channels deleted at least one signal while we were watching.
- How long should you follow a channel before trusting it?
- About 124 trades measured live, around 12 weeks at the usual pace, to know its result within ±0.2 R per trade. With fewer, a lucky streak and a good channel look the same.
- Why does a channel’s track record look better than its real results?
- Three reasons, from biggest to smallest: you picked it because its track record was good (and that was partly luck), it deletes losing signals you’ll never see, and it announces its targets more than its stops.
- How do I know if a channel deletes or edits signals?
- Save every signal when it’s posted and compare it a few days later. We keep every version of every post from the channels we measure: how we measure.
This is information, not investment advice. Trading on leverage can lose you more than you deposit: risk disclosure.
Hypothetical or simulated performance results have certain limitations. Unlike an actual performance record, simulated results do not represent actual trading. No representation is being made that any account will or is likely to achieve profits or losses similar to those shown.
Channel by channel
These are medians. What about the channel you follow?
Paste its link or find it in the directory: we measure every signal against the real price, with its stops, its worst day and what you lose by copying late.