As of 6 October 2026, fourteen of the 60 channels that posted signals in the last 30 days were feeds with no single trader behind them, and they posted 64% of all signals. What automated and aggregated feeds are, why they get the same score on a separate list, and why something in them always hits.
What is an automated or aggregated signal feed?
It is a channel where no single trader writes the calls. Automated crypto signal channels post what an indicator or a bot produces, and aggregated ones collect signals from several sources and repost them.
The automated kind usually starts on a chart. Someone sets a condition on an indicator, and when the condition triggers, the charting platform sends the alert text to a program that posts it to the channel. TradingView's help centre describes the first link in that chain: "A TradingView webhook notifies your external app when an alert is triggered" (About webhooks). The message subscribers read is a template filled in by that condition, and nobody has to be watching the chart when it goes out. The same page warns that webhooks "may occasionally fail to reach the specified URL", so a missing or late signal is an ordinary failure for this kind of feed.
The aggregated kind reads other channels and passes their calls on. Telegram's Bot API has a one-call method for the reposting step: copyMessage "is analogous to the method forwardMessage, but the copied message doesn't have a link to the original message" (Telegram Bot API). From the repost alone, a subscriber cannot tell whose call it was or when it first appeared. The model is old enough that regulators have described it. ESMA's briefing on copy trading describes a firm that "collects/gathers/pools trade signals provided by copied traders and then allows clients to subscribe to receive the signals, but this firm does not execute the orders" (ESMA35-42-1428, March 2023).
A collection that reposts bot-generated signals from several places counts as aggregated, since what reaches the subscriber is the collection. Neither kind is the same thing as a bot that places signals on a follower's own account. That bot sits on the follower's side and is covered in auto-trading crypto signals with a bot.
Why is a feed's record different from a trader's?
Because it measures no single person's judgement. The numbers describe a tool's output, or a blend of several authors, as the feed posted them.
An automated feed's record belongs to a rule set. Someone chose that rule, quite possibly after trying others, and can change it without telling anyone, so the record from before and after a change sits on one page. An aggregated feed's record is a mixture whose ingredients are not listed, and the mix can shift from one week to the next as sources are added or dropped.
The words "bot" and "AI" in a channel's description are claims about how the calls are made. ASIC's 2026 release on investment scams describes the marketing version bluntly: "Scammers claiming their trading bots use AI to generate passive income and unachievable returns. It is very unlikely that these trading bots exist" (ASIC 26-063MR). A real automated feed is judged like every other channel: by what it posted, when, and what the market did next.
How much of the signal flow comes from feeds?
Most of it, from a small number of channels. As of 6 October 2026, 14 of the 60 channels in our index that posted signals in the last 30 days carried a feed label, 10 aggregated and 4 automated. Together they posted 6,246 of the 9,722 signals in that period, about 64%.
Inside that group the volume is very uneven:
| Posted in the last 30 days, as of 6 October 2026 | Signals | Per day |
|---|---|---|
| The busiest feed | 3,708 | about 124 |
| The second busiest feed | 634 | about 21 |
| The median feed | about 185 | about 6 |
| The quietest feed | 6 | under 1 |
| The median channel run by people | 22 | under 1 |
| The busiest channel run by people | 991 | about 33 |
Two things follow from the table. "A hundred signals a day" describes one feed, and the typical feed posts about six. And volume does not identify a feed: the busiest channel run by people posted more than every feed except one.
Ten of the fourteen feeds are new to the index: they were added this month with their earlier history imported from the channel, so part of those 30 days was read after the fact. Imported history is what a channel had left in its feed by the day it was added, and live reading of a feed starts on that day.
How does ChainRated decide a channel is a feed?
A person sets the label, using only what the channel says about itself: its name, its own description, the headings of its posts. How often it posts never decides it.
The page then shows those words and the date they were noted. An aggregated feed's card says the channel "collects signals from several sources and reposts them" and quotes the channel's own wording. It adds that "the numbers below are the combined record of this feed as posted here, not the calls of one trader". An automated feed's card opens "By its own description, this channel posts signals generated by an indicator or bot". On all fourteen feeds the noted date is 4 October 2026.
The label describes who produces the feed. Pooling other people's calls is a business model regulators have already described, and posting an indicator's output is a legitimate way to run a signal channel.
Some cards carry one more line on what the record leaves out, for instance that a feed also posts calls on tokenised stocks or commodities and only its crypto signals are tracked. The same call posted twice back to back is counted once, so a feed that repeats itself does not double its record. A clone, one channel republished under a second name, is a separate case with its own treatment, set out in clone signal channels.
Why are feeds scored the same way but ranked apart?
So that a reader comparing channels compares like with like. Every channel's score comes from the same formula, and the label changes only the list a feed appears on.
The methodology puts it in one line: "Aggregated and automated feeds are scored the same way and ranked apart." Each score is built from the same three axes: Return at 40%, Risk at 35% and Honesty at 25%. They are computed the same way for every channel, and the feed label moves none of them. No feed loses a point for being a feed.
What the label changes is placement. Feeds have their own tab in the catalogue, Aggregated & automated, which held fourteen channels on 6 October 2026, numbered among themselves. They are left out of the Verified tab and the top of the home page, and on the All tab they sit in a separate block at the end without a place number. The leaderboard counts them instead of ranking them. On 6 October its footer noted 8 feeds as "scored but ranked separately"; the other six feeds did not yet have enough scored history for a ranked score.
A person's record and the output of a blend or an indicator answer different questions, and a single ranking would invite the reader to treat them as one. Feeds are ranked against each other, on the same numbers everyone else gets.
Why does something in a feed always hit?
Because a near target, a far stop and a large number of calls make hits close to certain. A busy feed always has a winner to show, whatever its calls are worth.
Take the test every channel on ChainRated goes through, which the site reports as one-take accuracy: a single take at +2%, against the channel's own stop or, where it published none, a fallback stop at 10%. For a price that drifts nowhere, the chance of touching a level a% above the entry before one b% below it is b/(a+b), the classic gambler's-ruin result (Sigman, Columbia IEOR 4700 notes). With a +2% take and a 10% stop that is 10 in 12, about 83%, before any skill enters the picture. A feed posting a hundred such calls a day would announce about 83 hits a day by chance alone.
That is an illustration and describes no channel. Real prices drift and gap, trades cost fees, and no feed is a random walk. What the arithmetic shows is how cheap a high hit rate is at a close target, a point what Signal Verified means makes with our own figures.
A results post shows calls picked from everything a feed posted that day. Researchers studying backtests describe the same habit in another setting: "The investment manager only publicizes the model that works but says nothing about all the failed attempts" (Bailey, Borwein, López de Prado and Zhu, Notices of the AMS, May 2014). The useful question is how the whole list did, with every call counted.
What does an 87% hit rate on a feed with no stops look like?
On 6 October 2026, the automated feed with the most parsed signals in our index had 4,015 of them, and not one carried a stop. Its one-take accuracy read 87%: of 2,491 scored signals, 2,169 reached +2% before the stop, 301 hit the stop first and 21 did neither.
All 2,491 of those verdicts were scored against our 10% fallback stop, because the feed never published one of its own. That puts 87% close to the 83% a driftless price would give at the same +2% take and 10% stop, with no skill at all, so most of what the number records is how far away the stop was. Why the fallback flatters channels that publish no stop is covered in why half of all crypto signals come without a stop-loss.
The rest of the card shows what the 87% leaves out. With no stops there is no as-published record to replay, so the return shows a dash and every row of the replay table reads "too incomplete to follow as published". The risk axis, where stop coverage is the largest part, reads 0.0. The score is 28.1, the lowest of the fourteen feeds, and the feed is not on the leaderboard yet.
The tab lists feeds by one-take accuracy until the sort is switched to Score, and on 6 October this feed was first on it. The highest hit rate on the tab and the lowest score on it belonged to the same channel, both on one page.
Can a feed be right often and still lose as posted?
Yes, and on 6 October most feeds did. One-take accuracy asks how often a channel is right; the as-published replay asks what following it would have returned, and feeds give very different answers to the two.
The aggregated feeds differ from the automated feed with no stops. On the cards we checked, 99 to 100% of their signals carry a stop, and their accuracy, where shown, ran from 55% to 82%. Replaying their own targets and stops as published gave a different picture. Figures are as of 6 October 2026, and the last column adds up the results of all closed trades, measured in percent of one trade's stake:
| Feed | Closed | Win rate | Average per trade | Sum of trades |
|---|---|---|---|---|
| Feed A | 647 | 41% | −0.1% | −34.7% |
| Feed B | 798 | 35% | −0.7% | −520.4% |
| Feed C | 125 | 27% | −0.2% | −23.6% |
Across the 13 feeds whose page shows an average result per trade, 4 were positive and 9 negative, and several of those averages rest on few trades. One positive feed, at +6.6% a trade, has 5 closed trades behind it. Two automated feeds sat slightly above zero, at +0.6% and +0.9% a trade. Samples that size settle little in either direction, which is the subject of how many trades a hit rate needs.
Entries that never fill distort a feed's record further. On one aggregated feed, 849 of 1,057 parsed signals "were never entered (the price passed the first target before the entry filled)". How that happens, and what it does to a record, is in half the trades never happen.
How do you choose from a feed that posts dozens of signals a day?
Choose the feed and a rule for which calls you take, then stop choosing call by call. The feed's record covers everything it posted. Yours will cover only what your rule picks.
- Read three things on the feed's page first. The share of signals that came with a stop, shown as "Stop-loss published: with a stop X% (N of M parsed signals)". How many of the scored verdicts used our fallback stop instead of the feed's own. And the as-published record next to the one-take accuracy. A high accuracy built on the fallback stop, with no as-published record beside it, is mostly a measure of the stop.
- Write the rule before the next signal arrives. Which pairs, which direction, how many positions open at once (longs opened in the same hour count as one, since crypto pairs tend to move together), and what you skip. A feed posting dozens of calls a day will always offer one that suits the moment, and a rule fixed in advance is the selection that moment cannot bend.
- Keep a record of every call your rule picked, losers included. Deciding afterwards which calls you "would have taken" repeats the selection a results post makes. Your subset needs its own sample before its hit rate means anything.
- Size for your account. A feed does not know your balance, your leverage limits or what you already hold. How to size a position is in position sizing for signal followers.
What this does not prove
The label reflects what a channel says about itself on the date shown. A channel that runs a bot and never says so is not labelled.
Every figure here is a reading taken on 6 October 2026, and most of the fourteen feeds were added this month with imported history, so the numbers will move as live reading accumulates. Some feed records are small, and their averages say very little yet.
Nothing in this article is a verdict on any feed's intent, and none of it forecasts what a feed will post or return next.
The practical read
A feed is a stream of calls with no trader behind it, scored on the same scale as every other channel and listed on its own tab. Before following one, read its page for the share of signals with a stop and the as-published record, and treat its results post as a selection from everything it posted. If you follow it, follow a rule you wrote in advance, and judge that rule by your own list of every call it picked.