How we rank forex brokers

The six criteria and how each is scored

Each criterion is scored from 0 to 10 and rounded to one decimal place. The formulas below are the ones the site runs; nobody adjusts a score by hand afterwards.

• Costs, 25%. The all-in cost of one EUR/USD trade in pips: the lower of the commission-free account's published spread and the raw account's spread plus round-turn commission (2 × commission per lot per side ÷ 10). Score = 10 − 4 × pips, kept between 1 and 10. A commission quoted in GBP, EUR or AUD is counted as if it were USD.
• Regulation, 25%. Counts the broker's entities that are licensed by a Tier 1 regulator and accept retail clients: one gives 6, two give 7, three give 7.5, four or more give 8. With no such entity the score is 3 if a Tier 2 entity exists and 1.5 if there are only offshore or unlicensed ones. Then +0.5 if a Tier 1 entity confirms negative balance protection, +0.5 if a retail entity belongs to a compensation scheme, and −1 for a regulator's action or a sale of the client book.
• Platforms, 15%. Starts at 2, then +2 for the broker's own platform, +1 for MT4, +1.5 each for MT5, cTrader and TradingView, +0.5 for an API, up to 10. It rewards the range of recognised terminals, not how pleasant they are to use.
• Accounts, 15%. Starts at 3, then +1.5 for a swap-free Islamic account, +1 for a raw or ECN account, +1 for a professional account, +0.5 for a demo account (only if the broker's own page lists it). The minimum deposit adds 2 if it is 10 or less, 1.5 up to 100, 0.5 up to 250 and nothing above that or when it is not published. Deposits are compared at face value in their own currency.
• Research and education, 10%. Not scored; shown as "not measured" for every broker.
• Customer support, 10%. Not scored; shown as "not measured" for every broker.

Tier 1 in our data covers FCA, ASIC, CySEC, DFSA, MAS, JFSA, CIRO, BaFin, Central Bank of Ireland, CFTC/NFA, FINMA, KNF, Danish FSA, Finantsinspektsioon, SEC/CFTC. An offshore licence (for example FSA Seychelles or FSC Belize) scores below a Tier 2 one, whatever leverage it allows.

A missing figure counts as 5 out of 10, not as zero

A criterion with no published data shows 0, which on our pages means "not measured", not "poor". The overall rating is the weighted average of the four desk criteria, with their weights (80% in total) scaled up to 100%. A desk criterion without data enters that average at a neutral 5. Leaving it out would judge a broker that publishes nothing only on its strengths and could lift it above one that publishes an average spread.

Example: regulation 8, platforms 6.5, accounts 8 and no published EUR/USD spread give (5 × 0.25 + 6.5 × 0.15 + 8 × 0.25 + 8 × 0.15) ÷ 0.8 = 6.8. On 19 October 2026, 6 of 20 brokers did not publish an EUR/USD spread in pips, so their costs score reads "not measured".

Each broker also carries the share of the full weighting that is actually measured. Without our own tests the ceiling is 80%, and 14 of 20 brokers reach it. In a list sorted by a single criterion, a broker with no figure for that criterion goes to the end rather than being ranked as the worst.

Every list has its own filter and sort order

No page on this site names a "number one for everyone". Each ranked page stores its own rule, and its list is built from the same data on 20 brokers in five steps.

• Brokers withdrawn from our data are removed.
• A filter keeps only brokers that match facts in the data: accepting residents of a given country, holding an entity of a given regulator tier, offering given platforms or account types, or a minimum deposit at or below a limit. A deposit quoted in another currency fails a deposit filter, because we do not convert currencies.
• Equal scores are split by a fixed order derived from the site and the page address, so a tie is resolved the same way every time the page is rebuilt.
• The brokers are sorted by the page's criterion: the overall rating; one of the six criteria; lowest minimum deposit; lowest published EUR/USD spread; alphabetical order for directories; or a beginner blend of accounts (35%), research (25%), support (20%) and regulation (20%), with the lower deposit winning a tie. While research and support are unmeasured, the beginner blend is decided by accounts and regulation alone.
• A page may pin a broker to the top only if that broker passed the filter, and may cut the list to a set length. The rule format also allows putting partner brokers first; no ranked list on the site uses that option.

Some facts apply whatever the page rule says. A broker that refuses residents of a country, or carries an editorial flag for that market (a regulator stop order or a sale of its client book), does not appear on that country's list; FXCM, for example, is left off the UK list because its UK client book is being sold to Trade Nation. An entity that no longer takes retail clients does not count towards a regulator-tier filter. On a country page, the regulator tier is checked on the entity that serves residents of that country, not on the brand as a whole.

Above each list we print its rule in one sentence, the number of brokers that passed the filter and the date of the data. For example, the MetaTrader 4 list keeps only brokers that offer MT4 and sorts them by platforms score, while the list for new traders keeps brokers with a Tier 1 entity and sorts them by the beginner blend.

Where the data comes from and how old it is

Facts come from three kinds of source: each broker's legal and pricing pages for every entity, regulator registers, and the retail loss figure each entity publishes with its risk warning. The source address is stored with each fact. The current set was collected on 19 October 2026. There is no fixed re-collection schedule yet, so the date printed under each list is the date its data was collected.

83 of 152 regulated entities were checked on the regulator's register on 20 October 2026; the rest show the licence number as published by the broker, marked as such. Each regulator badge links to the register, so you can check the licence yourself before opening an account. Spreads are taken as published: some brokers quote a "from" figure, others a typical or an average one, and we do not adjust between them. That mix is one more reason the ratings are labelled preliminary.

What the rating does not cover

The desk score leaves out everything that can only be measured from inside an account: execution speed and slippage, overnight swap rates, withdrawal times and fees, inactivity fees, the reliability of iOS and Android apps, research quality and support response times. It also does not weigh how a broker treats clients in a dispute. If one of these matters most to you, read the broker's fee schedule and client agreement before you deposit. Once we test live accounts, each tested criterion will show its test date, and ratings built on tested data will drop the "preliminary" label.