IBKR and IG displayed conflicting CFD-loss rates across concurrent public pages

Public United Kingdom pages checked on 8 November 2026 gave more than one answer to the same basic question. Interactive Brokers U.K.’s CFD contract-specification page displayed a 58.9% retail loss rate, its main CFD product page displayed 57.9%, and a still-accessible risk-disclosure statement dated 10 April 2026 displayed 59.7%. IG’s dedicated risk-disclosure page displayed 70%, while its UK CFD calculator displayed 69%. The observations show why a provider-level figure needs a page, legal entity and calculation date.[2][3][10][5][6]

The finding does not prove that either firm calculated its underlying percentage incorrectly. Both discrepancies can be explained by update timing: the Financial Conduct Authority requires the calculation every three months over the preceding twelve months, so an older document can legitimately contain the previous quarter’s number. The control problem is at the publication layer. A customer or comparison service can reach two public pages on the same day and extract two different figures without being told on every page which calculation date applies.[1][2][3][5][6]

Two providers, four public numbers

Interactive Brokers’ dated statement identifies Interactive Brokers (U.K.) Limited and says 59.7% of retail investor accounts lost money. The two live product surfaces checked on 8 November displayed 58.9% and 57.9%. The most plausible explanation is asynchronous quarterly updating, but the pages do not place a common start and end date beside each percentage. The public record therefore supports a disclosure-control finding, not a claim that any one figure was deliberately false.[2][3][10]

IG’s difference is less readily explained from the face of the pages. Its UK risk-disclosure notice displayed 70%, while the CFD calculator’s key-figures panel displayed 69%. Both pages were publicly accessible during the same check. One may have been updated before the other, or the pages may draw from different content systems. The records inspected do not establish which number was newer, and a one-point difference is not evidence of a material change in customer outcomes. It is evidence that page location affects the number a reader sees.[5][6]

The FCA requires an up-to-date figure, but pages can move at different speeds

COBS 22.5 requires the percentage to be up to date, recalculated every three months and based on the twelve months preceding the calculation. It also requires firms to retain the account records used for five years. The rule standardises the calculation and the warning sentence. It does not mechanically synchronise every webpage, cached component, downloadable disclosure and marketing template the moment a new quarter is adopted.[1]

That distinction matters for compliance monitoring and journalism. A crawler may land on a high-traffic product page, a help article, a footer, a calculator or a PDF. If those surfaces refresh on different schedules, the crawler can report a stale number as current or treat two numbers from one legal entity as two separate observations. A defensible dataset therefore needs the URL, access time, legal entity, document date where available and a fingerprint or archived copy—not merely the broker name and percentage.[1][2][3][5][6]

The wider sample ranged from 58.9% to 78%

The other pages checked supplied context rather than a clean league table. CMC Markets displayed 69%, Pepperstone 73.6%, XTB 78% and Trading 212 77%. Across the current live-page observations, the displayed range ran from 57.9% to 78%. That arithmetic is reproducible, but duplicated providers and potentially different calculation windows make a simple mean or median misleading. The correct first step is to reconcile each legal entity to one operative period.[4][7][8][9][10]

Even after that reconciliation, the lowest percentage would not prove that customers lost the least money. The FCA metric counts the proportion of eligible retail accounts whose sum of realised and unrealised net profit on the relevant products was below zero after costs. It gives one losing account to a customer down £1 and one losing account to a customer down £100,000. The numerator records signs, not magnitudes.[1]

A binary measure discards the size of the outcome

Consider two simplified brokers with 100 active retail CFD accounts. At Broker A, 60 accounts lose £1 each and 40 accounts gain £1.50 each. Its warning is 60%, although the accounts collectively break even in the illustration. At Broker B, 75 accounts lose £10 each and 25 accounts gain £50 each. Its warning is 75%, although the accounts collectively gain £500. The mandated percentage makes Broker A look lower-risk on frequency while the aggregate-money result points the other way.[1]

The example is deliberately artificial. Real books include open positions, variable funding, financing charges, spreads and customers who trade at different times. Its purpose is to isolate the denominator. The warning answers how many eligible accounts finished below zero. It does not disclose the median loss, total losses, total gains, return on funded capital, drawdown or the distribution of results around zero.[1]

Customer mix can change the percentage without a broker-control change

A mass-market app attracting first-time traders can produce a different account distribution from a broker serving experienced, high-balance clients. Product mix also matters: index CFDs, individual-share CFDs, rolling spot foreign exchange and restricted options can behave differently. The public warning combines the firm’s eligible account book. It does not standardise experience, leverage use, holding period, instrument mix or account size across providers.[1]

Movement around zero can dominate the headline. Ten small profitable accounts turning slightly negative raises the loss rate even if one large losing account recovers most of its deficit. Conversely, many tiny losses moving just above zero can lower the percentage while a few large accounts lose more. Costs are included, so a small gross trading profit can still become a losing account after fees and commissions. That is appropriate for a consumer warning, but it prevents the figure from isolating execution quality or pricing.[1]

Accounts are not necessarily unique people

The prescribed wording refers to retail investor accounts. It should not be rewritten as the percentage of customers who lost. One individual may have more than one eligible account, depending on the firm’s account structure, and firms can differ in their use of currency or product sub-accounts. The retained calculation records would be needed to test how each provider handled multiple accounts, openings, closures and migrations during a particular period.[1]

This is another reason a one-decimal-place comparison can look more precise than the underlying comparability. The FCA method standardises the sign test and exclusions, but it cannot make every customer base and account architecture identical.[1]

What an auditable warning would add

A stronger public disclosure would place the calculation date and twelve-month end date beside the percentage, identify the exact legal entity, and expose one canonical machine-readable value used across all pages. Older PDFs could remain available as an archive, but their historical status would be explicit. That would let a customer distinguish a genuine quarterly change from a content-management lag and would make automated monitoring less error-prone.[1][2][3][5][6]

For outcome analysis, firms or regulators could also publish the median percentage return, median monetary loss, distribution by account size, ratio of total losses to gains and proportion of accounts closed for margin. Those additions would create privacy and comparability challenges, but they would answer questions the current binary warning cannot. A lower loss-account rate accompanied by larger median losses would carry a different meaning from a lower rate with uniformly smaller losses.[1]

The finding and its limit

Nothing in the pages inspected establishes that Interactive Brokers or IG breached the FCA rule. The differences are consistent with ordinary quarterly updates reaching web surfaces at different times, and Interactive Brokers’ 59.7% document is expressly dated. Nor does a one-point mismatch materially alter the central consumer message that leveraged CFDs are high risk. The finding is narrower: public distribution controls can leave more than one percentage accessible for the same provider, so a responsible comparison must reconcile page, entity and calculation date before ranking anything.[1][2][3][5][6]

The same caution applies after reconciliation. The warning is valuable because it shows how common losing accounts are. It is not a measure of loss severity, average return or broker quality. On 8 November 2026, the first obstacle to a clean league table appeared before any statistical adjustment: two providers did not present one uniform number across the public pages checked.[1][2][3][5][6]

Report an error or submit an update