How we rate brokers

Where the scores actually land today

Published reviews

48

Carrying a score

32

Not rated yet

16

Range

2.5–3.7

Median

3.0

Highest possible today

3.7

No review on this site scores above 3.7 out of 5, 29 of the 32 scored reviews sit on exactly 3.0, and 16 published reviews carry no score at all. None of that says the brokers are equivalent, and none of it says they are poor. Those 29 reviews share a score because they share an evidence base: one authorisation confirmed at the register that issued it, and nothing else checked at a primary source yet.

Two of the three categories below can currently be scored for almost no broker, because the claims they rest on have not been checked at a primary source. That is why the scores sit so close together and why none of them is high. The alternative — spreading the roster out with estimates, or filling the gaps with the scores the source reviews carried — would produce a more useful-looking page and a less true one.

What this methodology can and cannot score

It scores three things, and only from documents read at their source:

• Regulatory standing — authorisations confirmed at the register that issued them.
• Cost transparency — headline dealing costs published by the named legal entity and read there.
• Disclosed terms — the loss, leverage and eligibility terms that entity publishes about itself.

Not scored, and why

Nobody at this site has opened an account with any broker on this roster. Every category below needs a dated record of someone actually using the service, so none of them is scored — not weighted lightly, not estimated, not scored.

• Order execution and slippage — needs a dated test artefact from a funded account. No account has been opened. A score here would be an assertion of first-hand experience the site does not have.
• Real spreads and total cost of a round turn — needs dated spread captures on a live account. A published spread table is a marketing figure until someone measures the fill. We can report what the broker publishes; we cannot score what it delivers.
• Withdrawal reliability and time to funds — needs a completed withdrawal attempt, dated. There is no way to desk-verify this. It is the single most requested category and the one we are least able to evidence.
• Support responsiveness — needs dated contact logs across channels. Reading a support page establishes that channels exist, not that anyone answers.
• Platform and mobile app usability — needs a dated walkthrough of the live platform. Screenshots from a broker's own marketing are not a usability finding.

Each is added as a scored category when dated test artefacts exist for enough of the roster to make the comparison meaningful. Adding a category changes every total, so it is a versioned change to this methodology, with a new effective date and a re-derivation of every score, not a quiet edit.

The categories and their weights

Regulatory standing 50% of the score

How much of the authorisation a review names has been confirmed at the register that issued it, and whether a named authority has published an adverse finding about the firm.

Why this weight. It carries half the weight because it is the only category with a primary source that does not depend on the broker's cooperation, and because it is the one thing a reader can independently repeat in five minutes.

Evidence required

A licence recorded as verified against the issuing regulator's own public register, carrying the licence number and the entity name the register itself prints. Warning and enforcement flags count only from the permitted warning-source set.

The bands

Cost transparency 25% of the score

Whether the broker publishes the two figures a reader compares first — the minimum deposit and the headline spread — for the specific legal entity that reader would contract with, and whether we have read them there.

Why this weight. A quarter of the weight, not more: what a broker publishes about its own costs is a disclosure, not a measurement. It says the number exists and is findable, not that it is what you will pay.

Evidence required

A minimum deposit or spread recorded as verified, attributed to a named legal entity, sourced to that entity's own live terms or account-opening page with a URL and a date, and inside its staleness window.

The bands

There is no zero band. Scoring zero would require a recorded negative read — someone looked at the entity's own terms and the figure is not published — and the fact model does not yet carry that finding for cost terms. Until it does, unchecked and undisclosed are indistinguishable in our data, and a score that cannot tell them apart should not be given.

Disclosed terms 25% of the score

Whether the named legal entity publishes the terms that decide what a reader actually gets: the percentage of its own retail clients that lose money, the maximum leverage it offers that entity's clients, and who it will and will not accept.

Why this weight. Equal to cost transparency, because a loss-percentage disclosure is the most decision-relevant number a leveraged-products provider publishes about itself, and because eligibility and leverage are the terms most often stated group-wide when they are entity-specific.

Evidence required

A retail-loss, leverage or eligibility figure recorded as verified, attributed to a named legal entity, read on that entity's own disclosure or terms page, and inside its staleness window.

The bands

Eligibility counts once whether it is recorded as client acceptance or as a registration; they are two ways of recording the same disclosure and counting both would let one document score twice.

How the categories become one number

Each category yields a band score out of 5. The overall score is the weighted mean of the categories that could be scored, using the weights below. A category with no primary-source evidence is NOT scored as zero — zero would assert a finding ('this broker is opaque') where the truth is an absence ('we have not checked'). Scoring an absence as zero is the same error as printing an unverified number.

Evidence thinness is already priced in, once, by the evidence ceiling in the ceiling policy: a record with no verified claim is capped at 4.0 and one with a single confirmed licence at 4.5. Charging for thin evidence a second time inside the category scores would push well-documented, long-authorised firms toward 1/5 for a gap in our own record — which the ceiling policy explicitly forbids describing as a quality ranking.

When there is nothing to score

A published review with no primary-source evidence at all has nothing to derive a score from. It is published without one rather than given a low one: a low score would be read as a finding about the broker when it is a gap in our record. Sixteen published reviews are in this state at launch, including several of the largest and longest-authorised firms on the roster.

Not rated yet. Nothing on this review's record has been confirmed at a primary source yet, so there is no evidence to derive a score from. This is a gap in our work, not a finding about the broker, and it is the reason the review carries no number instead of a low one. One register lookup. The authorisations these reviews name are recorded and waiting to be checked; confirming a single one moves the review into the regulatory-standing category and produces a score.

Two caps, both one-way

• The regulatory and evidence ceiling: a published score never rises above the ceiling the ceiling policy sets for that record. Binds now: No record. Every derived score already sits below its ceiling, because the categories that would lift a score are the ones we cannot yet evidence.
• The published score never rises above the score the translated source review carried. This site has evidence for capping the imported editorial judgement and none for improving on it. Verification outranks a source value for a FACT — a licence number, a loss percentage — and the record shows two cases where it did. An editorial score is not a fact, so where our thin evidence base would raise a translated score, we hold the lower number instead. Launch-scoped and reviewed at the same 90-day cadence as the evidence ceiling. It is removed when the evidence base is thick enough to stand on its own. Binds now: One record: InstaForex, derived 3.0, held at its source value of 2.5.

Weights are integer percentages and band scores are exact tenths, so the weighted mean is computed in integer arithmetic and rounded half-up to one decimal. No floating-point comparison decides a published score.

Two categories that did not survive

These reviews began as translations of Arabic-language reviews written for a different readership. Two of the things those reviews scored are not scored here, and removing a scored category changes every total, so both decisions are recorded rather than made quietly.

• Arabic-language support (scored in 54 of 54 source reviews) — dropped. The source rated brokers for an Arabic-speaking readership, where support in the reader's own language is a real differentiator. This site publishes in English for a general readership, and scoring one language's availability would carry the source's audience assumption into a roster that does not share it. Its weight is not redistributed to a like-for-like replacement, because there is no equivalent for a general audience — support quality of any kind needs a test artefact, and none exist. The weight went to the two disclosure categories, where evidence is at least possible without one.
• Islamic (swap-free) account (scored in 52 of 54 source reviews) — reported as a fact, not scored as a quality. Availability of a swap-free account type is a product fact, not a quality dimension: it tells a reader whether the account exists, nothing about its cost or terms — and swap-free accounts frequently carry an administration fee in place of the swap, which the source scored as a positive without checking. Scoring its presence as better would also embed an audience assumption this roster does not share. It moves to the summary card as a stated fact about the broker. At launch not one of these values has been checked against a broker's own terms, so under the record's own rule an unverified value does not render, and the fact is absent from the card until it is verified.

Every review keeps the score the source carried, recorded next to the derived one, so a change of score is auditable rather than invisible. The derived score is the one that renders.

The registers and warning lists, and how to search them

Every one of these is public and free. Where our own record cites one, the link below is the page we read. Where an authority's register defeated us, that is noted too — it is usually the reason a review on this site carries no score.

The patterns worth walking away from

None of these is a licence check, and none of them replaces one. They are the recurring shapes in the complaints that reach regulators.

• A number that cannot be found on any register, or a licence claimed from an authority that does not regulate the activity being sold.
• An entity name that changes between the website footer, the terms of business and the payment page — or paperwork that names a different company from the one you were sold.
• Pressure to deposit now: a bonus that expires, a 'position closing soon', an account manager who calls repeatedly. Authorised firms are constrained in how they may solicit; unauthorised ones are not.
• Any guaranteed or fixed return on a leveraged or traded product. There is no such thing, and offering one is a reliable marker of the firm you least want.
• A withdrawal that requires a further payment: a tax, a fee, a 'verification deposit', or a bonus turnover requirement invented after the fact.
• Being asked to install remote-access software, or to let anyone else trade the account for you.
• Recovery offers after a loss. Being contacted by someone offering to get your money back, for a fee, is a second fraud aimed at the victims of the first.