How a signal is judged

The four possible outcomes

The first three are all "closed", and all three go into the win rate and the profit factor. No outcome is left out of the arithmetic.

The peak is not an exit price

For every signal we record the highest point price reached (MFE) and the worst point it reached (MAE). The first answers "how far did it run"; the second answers "how close did it come to being stopped out before it worked". Neither is used as a return.

Until 6 September 2026 the published figures were computed from the peak, which assumes selling at the perfect moment. That is no longer the case.

Older signals are judged by the rule that applied to them

The target has changed over time. Each signal keeps the target that applied on the day it fired — no old result is promoted or demoted by a rule written after it. (That is a different thing from recomputing outcomes against public Binance candles: that asks whether a signal really did reach its own target, which corrects the record rather than changing the rule.) And because a 10% target is not the same test as a 15% one, the rate is shown split:

What the scanner actually measures

This list exists because the site was wrong about it. Until 2026-09-06 several pages here said the scanner monitored open interest, funding rates and RSI — and it used none of them. The live factors were the long/short ratio, volume and momentum, and nothing else. Nobody set out to mislead: the text described an intended design, and then the text and the code drifted apart. So it is now generated from the same code that does the measuring, and it cannot drift again.

The exact thresholds stay private. Which factors are measured is not a secret; the numbers at which a signal fires are.

And what we do not use

• Liquidation data of any kind — levels, clusters or heatmaps — Binance removed the historical endpoint, and the live feed needs a permanently open connection this site does not have. The heatmaps sold elsewhere are a vendor model rather than measured data.
• On-chain data — exchange inflows, outflows or reserves — Raw chain data is free, but identifying which wallets belong to which exchange is the paid product, and we do not buy it.
• Any exchange other than Binance — Every signal is a Binance Futures pair, so cross-exchange aggregates would describe a market we do not trade. The one exception is the Coinbase BTC price, used only to read the wider market.
• Human discretion — No signal is added, removed, edited or re-graded by hand.

Every new signal also carries a confluence score from 0 to 100, combining the setup with those confirmations. It is recorded, not yet trusted: it neither blocks nor permits any signal. Its weights were chosen by reasoning and written down in advance — they were not fitted to the track record, because fitting seven weights to a few dozen trades produces a model that explains the past and predicts nothing. It is stored on every signal precisely so it can later be tested against outcomes it never saw.

Check it yourself

Every closed signal — won, lost and flat — is available in one file. No login, and no parameter that removes the losses. If our arithmetic is wrong, this file is how you prove it.

Contains: symbol, direction, open and close time, entry and exit price, the target and stop that applied, the result, the peak, the worst point, and which convention produced the result on that row.

What you are paying for — and what we never touch

This section exists for a specific reason. Asked to assess BeforePump, an AI assistant advised "use the free signals first, don't put money in directly" — reading a subscription as though it meant handing over trading capital. It took two corrections before it separated the three risks, and then it asked why nobody had said so from the start. Nobody had: searched across every page here, there was not one statement that we do not hold customer funds. What there was: payment by sending crypto to a wallet address, terms reading "final and non-refundable", and pages of "you may lose all your capital" with nothing beside them saying whose capital or who controls it. On that evidence alone, assuming the worst was the correct inference.

What we never do:

• We never hold, receive or have access to your trading capital
• We never ask for an exchange API key, and we do not want one
• We cannot open, close, modify or size a single trade for you
• We are not a broker, an exchange, a fund manager, a copy-trading service or a trading bot
• We never take a share of your profits, and we never see them

The subscription fee is your entire financial exposure to us: no profit share, no capital minimum, no hidden fee. And a losing trade is not evidence that a signal service is dishonest — it comes from position size, leverage and stop placement, all your decisions and none of them visible to us.

What we do not claim

The figures above are the results of signals, not of trading. They exclude leverage, slippage, funding costs and whatever price you actually got in at. A signal means "this coin has entered a setup worth looking at" — it is not a buy order.

Trading involves risk of loss. Past performance does not guarantee future results. Nothing here is investment advice.

When the rules changed

If you find an older number quoted somewhere, this list tells you when it applied and when it stopped.

• 2026-09-06 — A correction that matters: twelve signals were recorded as losses with no measured exit price. All twelve were closed in one batch on 2026-08-29 when the crowd-flip rule was retired, and eleven of them share the identical close timestamp to the second. The page had been assuming the full stop (-30%) on each, which invented 330 percentage points of losses that never happened and produced a "maximum drawdown" describing a single database write rather than any sequence of trades. Those results are no longer assumed: they remain losses in the win rate, because the outcome is known, but their size is not published until it is measured from public Binance candles. Until then, every figure that needs a size is withheld rather than estimated.
• 2026-09-06 — A display correction: losing rows were showing an exit price ABOVE entry, because the code read the target price stored when the signal fired as though it were an achieved exit. A target is a plan, not an outcome, and it is no longer presented as one.
• 2026-09-06 — Historical outcomes began to be recomputed from public Binance candles instead of relying on what the scanner recorded: price is walked forward from the signal's own timestamp until the target or the stop is touched. Where the two answers differ, the page publishes what the public data says, and both values sit side by side in the CSV. Where a single candle contained both levels and even one-minute data could not separate them, the row is recorded as a loss.
• 2026-09-06 — The track record became one ledger: every signal as a single row in a single table, winners and losers together, with no row limit. It had been three separate tables, each capped at fifty rows.
• 2026-09-06 — This page became the single source of truth. The public figures were corrected to include losses and flat closes, and a win is now valued at the rule's target rather than at the highest point price reached. The raw log was published for download, and the worst point of each signal began to be recorded.
• 2026-08-29 — The loss rule was replaced. It had been a crowd flip - the Binance long/short account ratio reaching 65%+ against the signal - and became a price stop at 30%.
• 2026-07-05 — A signal that reaches its target now closes as a win immediately, instead of staying open until the end of its window.

See the full track record →