How We Score Spread Betting Accounts

Written by Justin Grossbard

Our spread bet content is supported and we may receive payment when you visit a partner site. This page is information rather than financial advice.

How We Score, Step By Step

Each criterion is marked out of 10 for every account in our UK spread betting platform comparison, and the marks are published on that account’s own review. The names below are the names on those tables.

The six criteria are applied in the order below; each step names what is measured, the instrument, its cadence and the dataset.

How To Use These Checks To Pick A Broker

To choose a spread betting broker in the UK, check the FCA register, confirm the account is open, compare the published nine-pair spread at your stake, read the disclosure type, check guaranteed stop terms, check the platform and app.

1. Check the firm’s FCA register entry and the company you would contract with. The FCA regulation page lists all 14.
2. Confirm the spread betting account is still open to UK clients. The reviews directory marks brands with no account to open.
3. Compare the published nine-pair spread figure at your own stake. The numbers are on the statistics page and spread betting calculator.
4. Read what kind of spread figure the provider publishes before trusting the comparison. The disclosure table is on the statistics page.
5. Check whether a guaranteed stop is offered and on what terms. Each review covers this.
6. Check the platform and app carry the spread betting account and do what you need. The spread betting apps page and app test cover this.

What The Numbers On A Review Mean

Three scales appear on a review page and they measure different things. Reading a mark on the wrong one is easy to do and this section is the fix.

• A criterion mark is out of 10. It is one of the six above, for one account.
• The overall score is out of 100. It is our assessment of the whole account, and it is what our ranking is ordered by. Rank 1 is the highest score, and the order is strictly the score order with nothing else in it.
• The rating in the page’s structured data runs from 1 to 5. That is schema.org’s scale rather than ours: it is the same overall score expressed on the scale search engines read, and every review declares the scale alongside the value so it cannot be read as anything else.

The criteria are weighted rather than totalled. We tested whether any set of weights reproduces our published scores from these marks alone, and none does. That is worth being blunt about, because a table of marks beside a total invites you to add the marks up and check. The one figure on a review you can rebuild yourself is the spread average, from the nine rows printed above it.

We publish no crowd rating anywhere on this site. We collect no user reviews, so there is no average of anyone’s opinion to publish. A star rating built out of our own editorial score, dressed up as a customer verdict, is not something we will put in front of you.

What We Test, And What We Only Assess

Two datasets on this site are dated first-party tests: the monthly spread harvest and the 18 to 27 August 2026 execution timing below. Everything else on a review is either read from the provider’s own published material or formed from using the account, and the reviews say which. The table sets out every kind of claim we make and what sits behind it.

Two things in that table deserve a sentence of their own.

Our mobile app comparison used to be hands-on and undated, and that row has now flipped too. The order published on our spread betting apps page is still an ordering rather than a score, and we still publish no rating and no score out of anything for an app. What has changed is that the facts beside the ordering are now measured: 28 installs on two named handsets on two recorded dates, with the tap count, the alert result and the in-app guaranteed-stop outcome logged for every one. The section below sets out how.

We used to hold execution-speed figures and refuse to publish them, and that row has now flipped. Our broker data carried millisecond figures from an earlier exercise with no recorded date, device or method, and a number like that is either a test result or it is nothing. Rather than publish it we ran the test properly. The section below is the result: 210 timed orders on live funded accounts, with the date, the device, the connection and the measurement source recorded for every one. The undated figures were never published and are not published now. Nothing on this site quotes them.

Execution Speed Testing

We opened live funded spread betting accounts at all 14 FCA-regulated brokers we rank. Each account held between £250 and £500. All orders were placed from London over a fibre 500Mbps ethernet connection. We measured fill times using each platform’s own order log and trade history timestamps. The test covered 15 orders per broker, 210 orders in total. Instruments were EUR/USD and the FTSE 100. Order types were market open, market close, stop triggered and limit filled. Testing ran from 18 to 27 August 2026. Medians are computed directly from the raw rows. The full per-order dataset is published on our statistics page.

The p90 column beside each median is the linear-interpolated ninetieth percentile of that broker’s 15 fills, rounded to the nearest millisecond. With 15 readings there is no single ninetieth percentile, and the candidate methods disagree by enough on the slowest brokers to change which one you would quote, so the method is stated rather than left for a reader to guess. A median says what a normal fill looked like; the p90 says what a bad one looked like, and the gap between them is the figure worth reading.

What this test does not measure is worth stating too. It is a fill-time measurement on EUR/USD and the FTSE 100 from one London connection over ten days. It is not a measure of price quality, of how a broker behaves in a gap, or of what you would see from another country or another line. A faster median does not make a broker better, and the ranking on this site does not move on it.

App Testing

We ran the app tests on an iPhone 15 Pro running iOS 18.6 on 22 August 2026 and a Google Pixel 8 on Android 15 on 23 August 2026, using live funded accounts with each of the 14 brokers. That gave us 28 per-install data points. We recorded the exact app version numbers on both platforms. For every install, we measured the fewest taps needed to go from a logged-in state to a placed trade, tested whether price alerts triggered correctly, checked whether a guaranteed stop-loss order could be placed from within the app, and noted any crashes, hangs, interface glitches, or missing functions. Execution speed medians come from a separate dataset of 210 live timed orders placed across the same 14 brokers between 18 and 27 August 2026, with 15 orders per broker and the median published for each. The full per-install dataset, including every app version, tap count, alert result, GSLO outcome, and logged issues, is published for download alongside this page.

What this test does not measure is worth stating here too. It is two handsets, one of each operating system, on the builds that shipped in the third week of August 2026. It says nothing about how an app behaves on an older phone, on a weak connection, or after an update we have not seen, and a lower tap count is a convenience finding rather than a reason to open an account.

Where The Data Behind A Review Comes From

Every figure on a broker review is stored in one place with a source, a date and the quote it came from. A figure without those does not get published. Four datasets do most of the work.

The monthly spread harvest

We collect each provider’s own published spreads for the nine most traded currency pairs and store them in points, which is the unit a spread bet is priced in. The current harvest covers thirteen accounts and was run on 1 Aug 2026, and every page showing a spread from it states that date. The industry row a review compares against is the mean of those thirteen accounts. Each pair is compared against that pair’s own average rather than against the overall figure. An account can sit under the overall average while pricing above it on the pair you actually trade.

What kind of figure each provider publishes is on the record too, because that is a separate question from what the figure says. On 5 Sep 2026 we read sixteen brands on their own UK properties and stored, for each one, the kind of spread figure it publishes, which product that pricing covers and how many of the nine pairs carry one. They are not all averages, which is why the sentence above no longer says they are. The record is published brand by brand on our spread betting statistics page, with the source and the date beside every row.

Averages on this site are the arithmetic mean of the nine pairs printed beside them, so you can add up the table and get the number in the sentence. That is a guard with a history. For several months these were means of eight pairs published as nine-pair figures, and the check that would have caught it did not exist.

How We Compare Spread Betting Costs Between Brokers

To see how spread betting costs compare, we multiply each provider’s published spread figure (points) by your stake and monthly bet count. The industry figure for thirteen accounts is dated 1 Aug 2026.

At £5 a point and 20 bets a month, the industry figure of 1.7 points costs £8.50 per bet and £170.00 a month. The nine-pair spread figures across the thirteen accounts range from 1.29 to 2.26 points, so monthly costs run from £129.00 to £226.00. These are the figures as we transcribed them on 1 Aug 2026, and they are not all published averages: the disclosure read of 5 Sep 2026 records that, of the thirteen, five publish a minimum only, two quote live with no static figure, four publish a fixed, target or typical spread and two publish an average. The named rows are on the statistics page.

Authorisation, checked on the register

For every provider we track we store the FCA-authorised entity name and Firm Reference Number, read from the firm’s own UK property on a stated date. All of them are published in one table on our UK spread betting regulation guide. The brand on the website is frequently not the company on the register: five of the firms we track will not return the company you expect. That page names each one and says what you will find instead.

Whether the account still exists

We check whether each brand still offers a UK spread betting account at all, on the brand’s own UK site, and we store the answer with its date. That is why our broker reviews directory carries six brands with no account to open alongside the fourteen that have one. A review of a product that was withdrawn should say so in its first line rather than sell you a button.

Everything a provider publishes about its own account

Minimum stakes, minimum deposits, inactivity fees, guaranteed stop terms, platform lists and market counts all come from the provider’s own UK pages. Specifically, from the spread betting side of those pages. Spread betting and CFD trading are different products at the same firm, with different instrument counts, different minimums and sometimes different companies behind them. A figure from the CFD side of a provider’s site does not go on a spread betting page here, even when it is the only figure the provider publishes. That rule is the reason some reviews on this site publish fewer numbers than the sites they compete with.

How We Cite Other People’s Ratings

We publish Trustpilot and TradingView scores exactly as those platforms display them, with the date we captured them, and we never recalculate an average from a star breakdown. Where a provider operates more than one Trustpilot profile we cite the one attached to its primary global domain, because Trustpilot organises profiles by domain rather than by country. We do not cite regional profiles with fewer than a hundred reviews. Where a profile carries a paid Trustpilot subscription, which permits automated review invitations, we say so beside the score.

App store ratings are storefront-specific, so an Apple App Store or Google Play rating only appears on a UK page here if the capture came from the GB storefront. A rating taken from the US storefront is a rating of a different audience’s experience.

What Happens When We Get Something Wrong

Corrections are published on the page that carried the error rather than filed somewhere quieter. Every page shows an Updated stamp and a “What Changed?” note naming the actual change. The stamp moves when a claim changes for a reader: a figure corrected, a statement added, a statement removed. It does not move for a layout change, and it never moves to make a page look fresher than it is.

Some corrections are subtractions. Figures have come off this site over the last month. Reading the source showed where they really belonged: to a different provider, to a different product, or to no published source at all. A review with fewer numbers on it than it had last month is usually a review that got more accurate. The note on the page will tell you which figure went and why.

If you think something here is wrong, our contact page reaches us, and a specific claim with the page it is on gets checked against its source.

Money, And Why It Does Not Move A Score

Some of the links on this site are affiliate links, and a page carrying one says so at the top. That is how the monthly re-checking gets paid for.

What it does not buy is a position. The ranking is strictly ordered by the overall score, the score is set on the criteria above, and no provider can be moved up it. The shortlist of alternatives on a review is the top of that same ranking with the page’s own subject removed. A provider we hold no commercial arrangement with still appears there, in its own position, with no button beside it. Providers we do have an arrangement with get the same treatment when the data is unflattering, and several reviews here lead with the thing their subject is worst at.

Our Methodology: FAQs