Transparency report
Verpips transparency report: October 2026
By Julián Cardona, founder of Verpips · Reviewed Oct 7, 2026 · How we measure
First edition · Twice a year: the next one in April 2027 · Figures pulled from our database on Oct 7, 2026
We ask signal channels to show everything they post, including what goes wrong. This is the same thing, turned around: how much we have measured, what we have corrected, who pays us, what data we keep and where our method can fail. Every six months, with the same sections, so each edition can be compared with the last.
Our independence, in one sentence
Verpips does not charge channels or brokers to be listed or for their position. No channel can pay to appear, to move up the rankings or to have its report taken down, and we have no affiliate links and earn no commission on accounts opened.
Today we have no affiliations at all. If we ever did, we would disclose it on the report of the channel or broker involved, and in this report.
What we have measured
14,982
signals measured against real market prices
Each one counts once, even if we measure it under several rules. Ambiguous, duplicate and in-house signals are left out.
363,456
Telegram messages read
Signals, results, updates and ads from the channels we read, each message counted once.
2,548
signal channels reviewed
The ones our crawler found and read, plus the fully audited ones, with no channel counted twice.
16
channels fully audited, signal by signal
Their whole history rebuilt against price, tick by tick and with the spread.
0
channels with the verified badge
The badge requires at least 30 trades, the whole interval above zero and no signals deleted while we watched.
0
reports claimed by the channel owner
Owners prove the channel is theirs by posting a code in it.
0
owner replies published next to our measurement
Only an owner who has claimed the report can reply.
0
corrections to a published study or figure
See “What we have corrected” below.
September 2024
the oldest signal we have measured
When we audit a channel we also read its history.
September 2026
when we started measuring live
Every signal from the moment it is posted, with its edits and deletions.
These are this edition’s figures and they do not change. Today’s figures, which update on their own, are in Verpips by the numbers, with how each one is counted.
What we have corrected
In this first edition there is no record of any correction to a published study or figure. Until now we did not keep a public log of corrections: from this edition on, each one is listed here with its date, what changed and where.
If you spot an error, let us know. If you own a channel, you can claim its report, and your reply is published next to our measurement.
We audited our own audits
In September 2026 we checked whether we could trust the way we read messages. From 14 audits we drew at random, with a seed fixed before looking, 210 signals, 136 messages we had not read as signals and 70 trades with all their follow-up messages. We labeled them blind, without seeing what the system had understood, and saved the labels before comparing. A person checked 20 labels picked at random by hand: 20 were right.
The result: we corrected how messages were read in 12 of the 14 audits, and none changed its conclusion. Channels that were winning kept winning, and those we could not tell apart from luck stayed that way. Re-running all 14 with that day’s code, without touching the readings, gave exactly the same result in 12.
How accurate our signal reader is
We will publish it in the next edition. Today we do not have a clean measurement, and we would rather not give a number that is not one.
The blind sample from September was used precisely to fix the reader, so measuring it now on those same messages would give an inflated number. For the next edition we will draw a fresh random sample of messages posted after those fixes, label it blind before comparing, and publish the accuracy per field (whether it is a signal, direction, entry, stop and targets), each with its interval.
What we already do with every change to the reader is a regression check: we re-read every message from our audits, and the change is rejected if it stops reading, or reads differently, a single message that was already being read. That shows the reader is not getting worse. It does not show how accurate it is.
How we are funded
We live off subscriptions from the people who use the service, and nothing else: not from channels, not from brokers, not from ads. That way the only thing that pays off for us is measuring well. There is more on this in About us.
What data we keep
- Site visitors: we count visits without cookies, using a fingerprint that changes every day and does not store your IP address. Visits are deleted after 90 days.
- Account holders: your email and how you sign in, what you audit and follow, your claims and replies, and your Telegram session if you connect it, encrypted and read-only.
- Channels: their messages and the results we measure from them.
We do not sell data or use it for advertising. All of it, including how long we keep it and how to have it deleted, is in our privacy policy.
Known limitations of our method
What we know our method does not see well, or has not solved yet. Each one links to the step of “How we measure” that explains it.
- We read everything they post →
We cannot see what was deleted before we started tracking a channel. An audit of its history reads what is still posted: if the channel deleted its losing signals earlier, it looks better than it is. That is why the verdict comes once we track it live.
- We read everything they post →
A message can be read two ways. Ambiguous follow-ups (a bare “done”, or “close half” with two trades open) are left out. One issue from the September review is still open: in one channel, 8 stop changes posted without the new value are read as the stop being removed, which can record losses larger than 1 R that may not have happened.
- We replay every signal against the market →
Prices depend on the data provider. With other prices, a very tight signal may hit the stop or the target when ours does not, or the other way around. Every report says which feed it was measured with.
- We replay every signal against the market →
Some signals have no outcome we can decide. If our prices cannot tell which level was hit first, we mark the signal as ambiguous and do not count it as measured.
- We tell you how much could be luck →
We do not measure “Telegram”; we measure the channels we read. Most of them trade gold, and they were found by our crawler or submitted by a user: they are not a random sample.