Methodology — the One-Take Test

The test

For every signal a channel publishes, we replay what would have happened to a trade that follows one fixed rule set:

• Enter at the channel’s published entry price, with the whole position, no leverage.
• Some channels never publish a price — they say “enter at market”. There the entry is the opening price of the first minute after the message went out, read from the same candles the replay runs on. Not the minute the message landed in: that minute started before the message existed, and nobody could have traded on a signal that had not been posted yet. Where a channel does name a price, that price is used, even if the same channel usually enters at market.
• One take-profit at a +2% price move from entry (long: entry × 1.02; short: entry × 0.98). The entire position closes there.
• The stop-loss is the channel’s own published stop. If the channel published no stop, a stop on the wrong side of entry, or a percent stop wider than 50%, we use a fallback stop 10% away from entry — and we disclose on the channel’s page how many of its verdicts rest on the fallback.
• The trade gets 30 days to resolve, replayed against real 1-minute candles (highs and lows, not closing prices).
• The candles are Binance futures, and every channel is scored on that one source. A coin Binance never listed stays without a verdict rather than being replayed somewhere else: minute highs and lows differ between exchanges, and a stop that fires on one venue can survive on another — mixing sources would make two channels’ numbers incomparable. Each verdict records the exchange it was replayed on.
• Exchange taker/maker fees are charged in the replay.

Accuracy is then: of the signals where the entry filled and the market gave an answer, the percentage that reached +2% before the stop.

The possible outcomes

• Reached +2% before the stop — counts for the channel.
• Hit the stop first — counts against the channel.
• Neither within 30 days — counts against the channel: a signal that goes nowhere for a month is an answer, not missing data.
• Entry never reached, price ran to the take before the entry filled, or we have no candle data — excluded from the percentage entirely. "We could not tell" is a different statement from "it lost", and mixing them would corrupt the number.
• On a channel that enters at market: the price had already passed the channel’s own first target by the time the entry would have filled. There was no trade to take — a fill above the first target is not an instant win, it is a signal that arrived too late to act on. Excluded from the percentage, and counted with the unfilled entries on the honesty axis.

Why one fixed take instead of the channel’s own targets

Channels publish wildly different target ladders. A channel whose first target sits 0.5% from entry "hits TP1" constantly; a channel aiming for 5% rarely does. Numbers built on the channels’ own targets are therefore incomparable between channels — and trivially gameable by placing the first target closer. One fixed +2% take, full position, asks every channel the same question: when you said "go", did the market actually move 2% your way before it hit your stop?

Why +2% specifically

A first take-profit in the real world is typically set 2–3% from entry. We take the near edge of that range, because the question this metric answers is not "how much could you have made" but "how often does a signal actually go the way the channel said it would, right after it was published".

That is also a real way to trade, not a synthetic benchmark. A conservative trader closes the whole position at the first take: they give up the upside of a runner, and in exchange they cut their risk sharply. The one-take number is exactly what that trader would have experienced following the channel — nothing modelled, nothing assumed about how you would have managed the rest of the position.

Minimum sample size

A hit rate shown on the listings rests on at least 10 scored verdicts. Below that we show a dash: 3 signals at 100% is an anecdote wearing a percentage.

The risk profile: what the signals asked of you

The hit rate says how often a channel was right. It says nothing about what being right felt like — and two channels at 65% can be completely different trades to sit in. So each verified channel’s page also carries three numbers, computed from the same one-take replay the hit rate comes from, never from a different test:

• Median wait to +2% — how long after the post the winners reached the take. Measured from publication, because that is the moment a reader had, not from the moment the entry filled.
• Median dip before the target, with the deepest of those trades beside it — how far price went against a winning signal before it turned. Zero is a real value here: some trades never trade below their entry.
• Median wait to the stop — how quickly the losers announced themselves.

Medians, not averages: one trade that took a month would otherwise speak for a channel whose signals usually resolve in an hour. Each of the three appears only once at least 10 replayed signals of that kind carry the number — the same threshold the hit rate uses — and shows a dash with the reason until then.

None of this enters the channel score. It describes a channel; it does not rank one.

What this metric does not tell you

• It ignores the channel’s suggested leverage and position sizing — deliberately, because those are follower choices.
• A +2% move covers fees at moderate leverage but says nothing about whether the channel’s own suggested exits were good.
• Wide stops make the test easier to pass and cost more when they fail — read the hit rate together with the stop distances and the as-published record on the channel’s page.
• Past hit rate is not a prediction. Nothing on this page is financial advice.

Feed composition and the "Mostly promo" label

For channels we scrape, we also measure what the feed consists of: how many posts we read over the trailing 30 days, and how many of them were actual trade signals under the channel’s proven parsing mask. Both numbers are shown on the channel’s page.

• The label "Mostly promo" appears when fewer than 5% of at least 100 observed posts are trade signals — a feed where the signals are the bait and the promotion is the product.
• A typical honest signal channel publishes one signal per several posts (entries, follow-ups, recaps), which lands well above the threshold.
• Channels we cannot parse yet carry no label either way: we do not measure what we cannot verify.

Channels that publish no entries at all

A channel with no track record is not automatically a channel we have skipped. Some publish no entries in the first place — only outcomes after the fact, only commentary, or only invitations to a paid group. For those, "Unverified" alone is misleading: it suggests we have not got round to them, when in fact there is nothing on their feed that could ever be replayed.

So we count the feed. We read a large sample of the channel’s posts (at least 200) and sort every one of them into exactly one category: an entry, an outcome report, a promotion, or something else. If a single post looks like an entry — a price to enter at, with a target or a stop — the channel gets no label at all, because then it is a channel that needs a parsing mask, not a verdict.

• "Results only" — at least a quarter of the posts report how trades ended, and none open one. A feed of winners with no way to check when they were entered.
• "Commentary only" — market talk, charts and calls in prose, with no tradable entry to replay.
• "Promo funnel" — at least 40% of the posts are invitations to register, join or pay. The feed is the advert.

The opposite case gets the opposite label. Some channels publish entries we parse without difficulty — sometimes with a full ladder of targets — and never, in any post, the stop that would end a losing trade. Such a signal has an opening and no published ending: it cannot be replayed as published, and whoever scores it afterwards chooses when to stop counting. The label "Entries without exits" states exactly that fact.

• It requires at least 20 parsed signals and, among them, exactly zero that state a stop — as a price or a percentage. One signal with a stop and the label is refused: that channel is judged by its track record.
• The evidence shown on the channel’s page is the count of parsed signals and the period they cover, straight from our database.
• The label is removed automatically the day the channel publishes a signal with a stop.

Channels we cannot verify — our limit, not theirs

There is a third reason a channel shows a dash, and it says nothing against the channel. Some feeds publish complete, readable entries — instrument, direction, entry, targets, stop — and we still cannot produce a record for them. When that happens the honest thing is to name our own limit rather than leave the reader guessing whether we judged the channel or simply never looked.

• "Market we do not cover" — the calls are for gold or currency pairs, and our candle history is crypto. Replaying a gold trade against a crypto-priced proxy is guesswork wearing the costume of verification: we checked one such proxy against a channel’s own gold calls and it drifted further in a night than the distance from the channel’s entry to its stop, which is enough to flip a winning trade into a losing one on paper. So we publish nothing instead.
• "Too few entries to measure" — fewer than 10 parsed entries so far. Any accuracy figure built on a handful of trades is noise with a percent sign, and it would be read as a verdict.

The channel score (0–100)

The one-take hit rate above answers "how often is this channel right". The channel score answers a different question — "is this channel worth your attention" — and it is the number the leaderboard ranks on. It is a weighted average of three axes, each computed from data on this site, each shown broken down on the channel’s own page.

• Return, 40% — risk-adjusted, never raw. What the baseline replay returned, divided by the deepest fall it went through along the way. A channel that made +300% through an 80% drawdown scores below one that made +80% through 15%: the second is survivable, the first is a coin flip that happened to land.
• Risk, 35% — how the calls were built. Stop coverage is 60% of the axis (what share of scored signals carried the channel’s own stop rather than our fallback), median published leverage is 25%, and the replay’s worst drawdown is 15%. Two measured behaviours can then mark this axis down, and never up — see below.
• Honesty, 25% — edits and deletions after publication are 40% of the axis, entries the market never reached are 25%, being a repost of another channel’s feed is 20%, and how complete the calls are — symbol, direction, entry, stop, target — is 15%.

Honesty carries the smallest weight on purpose. It measures hygiene, not skill: a channel that never edits a word while being consistently wrong should not outrank an honest channel that is right. A quarter is enough that manipulation cannot be outrun by good numbers, and small enough that "clean but useless" does not lead the table.

Where the risk axis is marked down for behaviour

Everything above scores what a channel publishes: a stop written into the message, the leverage it asked for. That is worth scoring, and it is not the same thing as what the trade then did to the person following it. One channel here publishes a stop with every single signal — and its losing trades sat 41 hours in the red before that stop was taken. Scoring the promise alone, we gave it 80 out of 100 for risk.

So two facts from the same replay can reduce the risk axis. How deep a winning trade went into the red before it came good: nothing up to 5%, the full reduction from 25%. And how long a losing trade stayed open before its stop was reached: nothing up to 12 hours, the full reduction from 48. Each takes at most a quarter off the axis, and both are medians over the channel’s replayed trades — the same numbers printed in "What the signals asked of you" on its page.

• The reduction lands on the risk axis, which is 35% of the composite — not on the composite directly.
• Past the far anchor it stops growing. A dip of 25% and a dip of 60% are marked down the same: beyond that point the difference between bad and worse is not something to rank on.
• Both need 10 replayed trades of the relevant kind — 10 wins for the dip, 10 stop-outs for the wait. Below that there is no median, and no reduction: a channel is not charged for a figure we could not compute. A channel too young to have those trades is therefore untouched here, which is the honest reading of a sample we do not have, not a verdict in its favour.

What each sub-metric is measured against

Every sub-metric is turned into a 0–100 figure against a fixed anchor, published here so that a channel’s score can be recomputed by hand from the numbers on its page:

• Worst dip on a winning trade: up to 5% costs nothing, 25% or deeper is the full quarter off the risk axis.
• Wait to the stop on a losing trade: up to 12 hours costs nothing, 48 hours or longer is the full quarter off the risk axis.
• Return per unit of drawdown: 3.0 or better scores 100, and the scale runs down to 0 linearly. A channel whose replay is below water scores 0 on this axis and is ordered by the other two — losing records are not ranked against each other.
• Leverage: a median up to 3x costs nothing, 25x scores 0, linear in between.
• Drawdown: 10% costs nothing, 100% scores 0.
• Edits and deletions: 0 scores 100, and a channel that changed or removed 10% of its recent posts is at the floor. Measured over the same trailing 90 days we re-read.
• Unfilled entries: up to 10% costs nothing, 50% scores 0.
• Stop coverage and call completeness are already percentages, and are used directly.

What happens to a sub-metric we cannot measure

It is dropped, and its weight is redistributed across the sub-metrics that remain. It is never counted as a zero. A channel that has never stated its leverage is not thereby a reckless channel, and a channel we have not re-read yet has not thereby been caught editing: scoring our own gaps as somebody’s failure would make the score a measure of our coverage.

The one place this rule is deliberately suspended is a channel with no replay at all. Such a channel gets no score whatsoever, rather than a score assembled from the axes that survive — otherwise a channel we have never measured arrives at the top of the leaderboard on the strength of a single fact, "not a repost", worth 100.

The owner’s trading account is shown, never added in

Some channels are run by traders who have connected an exchange account to us and published its record. Where that is the case, the channel’s page shows that account below its own results — the same percentages and the same two dates that stand on the track’s own page.

It does not enter the channel score, and it never will. The score is built from this channel’s replayed signals and from nothing else; an account and a feed are two different claims, and a trader who trades well is not thereby publishing calls that work for the people reading them. Reputation does not add up here, and nothing carries over from a person to a channel or back.

Recalculation

When we improve parsing or fix a resolver bug, historical verdicts are recomputed under the same published rules for every channel at once. We never hand-adjust an individual channel’s numbers.