Desk verdict Providers Briefings Five tests How we judge FAQ Compare services
Test

Grades that are measured

A conviction grade should be calculated, not vibed.

Most forex rooms attach confidence words to calls — “strong”, “high conviction” — that mean whatever the sender wants on the day. A measured grade is different: it marks where a call sits in its own model's return distribution, with a number behind it.

On the desk we back, every call carries a grade from A (highest) to D (lowest), and the threshold is set per model. The bar that earns an A is fixed against each model's own returns, which is why the same letter stands for a different absolute move on a different clock:

Grade-A thresholds are canonical and set per model. An A on one clock is not the same absolute move as an A on another.
Mean-reversion modelClockGrade-A bar (per trade)
Day Tradesame session, a 0 to 60 minute window0.70% average per trade
Multi Hourhalf a session out to roughly two sessions4.50% average per trade
Swing Traderoughly 7 to 28 days (the flagship)6.00% average per trade
Investinga long, higher-conviction horizonlong-horizon, no single bar

An A marks the top band of a model's own measured return distribution; a D is the lowest band still published. Because the bar is fixed per clock, an A on a same-session call (near 0.70% a trade) and an A on the flagship multi-week Swing call (near 6.00%) both certify “top band for this horizon” rather than one shared target stretched across holding times that are nothing alike. The scale stops at D: the E grade was retired from the live product in 2026 so the four steps keep a stable meaning.

Why per-model calibration matters to a forex trader

A single blanket pass-mark would quietly penalise the quick model and inflate the slow one. A same-session move near 0.70% and a multi-week move near 6.00% are not comparable in raw size, so judging both against a single absolute target would tell a reader nothing useful. Grading each call against its own model's spread means a B on a same-session call and a B on a flagship Swing call each say the same thing — “above-typical for this horizon” — which is exactly the cue a trader who cannot take every signal needs in order to know when to size up. For a trader who only wants the flagship Swing Trade model, the A and B calls in that row are the ones worth waiting for; for a trader spreading across clocks, the grade is what makes the rows comparable at all.

A grade only means something on top of a real record

The grade is a position in a distribution, so the distribution itself has to be honest first. A grade calibrated against a curated highlight reel is calibrated against a lie, however precise the letter looks — which is why this test rests on the denominator test rather than standing beside it. And a denominator is only evidence when the full count, losers included, is shown:

On its own, a percentage is a slogan rather than a fact. A “90% win” banner with no count beside it might rest on nine of ten hand-picked screenshots, or it might silently drop every losing week — and a reader has no way to tell which, which is precisely why the number is shown without its denominator.

Set that against the flagship figure on the desk we back: 74.4% across 78 Swing Trade signals in 2026. The 78 is the denominator — the full count of calls, losers left in, over a continuous run. With it in view the percentage turns into something a reader can probe: roughly 58 of those 78 calls closed in profit and the remainder did not, and the +225% return reads against a real drawdown instead of hanging in mid-air. A lower win rate with its count is almost always worth more than a higher one without, because the count is the one part of the claim a dishonest service cannot fake without telling an outright lie.

The question to carry into any sales page: out of how many, and are the losers still in there? If the count is missing, treat the rate as marketing rather than evidence.

Because the grade is one of the fields folded into the on-chain fingerprint (see locked before settlement), it is fixed before the outcome and cannot be revised once the trade closes. There is no E grade; it was removed from the live product in 2026 so the four-step scale keeps a stable meaning. Between the per-model calibration and the on-chain lock, the grade stops being a marketing dial a seller can turn up after a winner prints.

What failing this test looks like

A conviction label stops being evidence the instant it is a feeling instead of a figure — and a label that was never anchored fails twice, because it can also be quietly rewritten once the pair has moved.

  • Messaging-app channels (Telegram, Discord). The operator owns the feed, so a call can be added after the pair has already moved, edited where it sits, or deleted with no trace. That fails an outside witness and locked at issue outright — and usually the denominator too, because the losing posts simply never go up.
  • Copy-trading and PAMM rooms. A platform tracks participant results, which is more than a chat offers, but the calls are rarely witnessed by an outside organiser, rarely timestamped per signal and rarely graded — so they fail an outside witness, locked at issue and a measured grade even when a rough denominator exists.
  • Social-media callers. Posts can be quietly removed or selectively amplified, and the income often arrives through broker rebate links, so a caller tends to fail nearly every test together — an outside witness, a real denominator and clean incentives at once.
  • Signal-aggregator sites. They republish other people's forex calls without auditing them, so every gap in the original passes straight through unfixed. They fail an outside witness by inheritance.

This is why the guide frames itself as ranking a field rather than reviewing a single product: a grade tied to measured returns and locked before the outcome is the test most of the field cannot clear, which is exactly what makes clearing it worth paying for.