Forex Client Sentiment: What Each Provider Actually Counts

A broker publishes a percentage beside a currency pair and calls it client sentiment. On the same page, often in the same footer, sits a second percentage: the share of that firm’s retail accounts that lost money. One of those two numbers has a written specification behind it. The other has none.

One Firm’s Book Is Not the Market

Spot foreign exchange has no central venue and no consolidated tape. Nobody collects a public register of who is long and who is short. What a broker can see is its own client base, and that is what it publishes.

This matters more than it first appears. The firm is not sampling the market and reporting an estimate. It is counting its own customers, and the answer is exact for that group and silent about everyone else. Two brokers with different client bases, different regional licences and different minimum deposits will hold different books on the same pair, and both readings can be correct at once.

For currencies there is exactly one positioning dataset compiled by a public authority rather than a firm, and its scope is futures and options traded on an exchange, at or above defined reporting levels. Our page on the one positioning census published for currencies works through what that report does and does not cover, and it is a different instrument from the spot pair on your chart.

The Base Is Not the Same From One Provider to the Next

Every one of these tools reports a percentage. A percentage needs a numerator and a denominator, and the denominator is where the providers part company.

Three bases are in circulation. A reading can count open positions, so a client running four small long trades contributes four times as much as a client running one.

It can count client accounts holding any position in the pair, which makes every customer equal regardless of size. Or it can count volume, which weights the reading toward whoever trades in the largest size. The three answers can point in opposite directions on the same book at the same instant, because a handful of large accounts can be short while a crowd of small ones is long.

The divergence is not a rare edge case. Scaling into a position, which is ordinary practice, splits one view across several tickets and inflates a position count without adding a single client. Partial closes do the same in reverse. A book where most customers hold one small long and a few hold several large shorts will read majority long on an account base and majority short on a volume base, from the same rows in the same database.

None of the pages that rank for this term states which of the three its own figure uses. The tools display a bar, a ratio and a set of thresholds, and leave the base unstated. That is not a small omission, because someone holding two providers side by side is weighing quantities of different kinds, without being told so.

The dealing model sits underneath all of this too. A firm that internalises client flow and a firm that passes it to a liquidity provider both hold a client book, but what that book represents to the firm is not the same in each case, and our guide to how a broker handles client orders sets out the difference.

Why Two Feeds Disagree on the Same Pair at the Same Moment

Open two sentiment tools on the same currency pair and the numbers will rarely match. There are four separate reasons, and none of them means one provider is wrong.

The client bases differ. A broker whose customers are concentrated in one region carries that region’s habits in its book, and a broker with a high minimum deposit carries a different mix of account sizes.

The bases differ, as the previous section sets out. The refresh intervals differ, so two readings taken at the same wall-clock minute may summarise different windows. And the instrument set differs: a figure covering a spot pair and a figure covering a CFD on the same pair are drawn from separate position registers even at the same firm.

What Blending Does When the Inputs Are Not Comparable

Some sites do not publish their own book at all. They collect the figures several brokers publish and present a combined ratio, on the reasoning that a wider sample beats a narrow one.

That reasoning holds when the inputs measure the same quantity. Here they do not. Averaging a position-weighted figure from one firm with an account-weighted figure from another produces a number that has no definition: it is not the share of positions, not the share of accounts, and not the share of volume across the combined group. The arithmetic runs cleanly and the result means nothing in particular.

The thresholds layered on top inherit the same problem. A blended tool that treats a reading above a fixed level as a crowded position is applying one cut-off to a quantity whose composition changes whenever a contributing source adds or drops a feed. The cut-off stays where it is; what it is cutting has moved.

The Other Percentage on the Page, and Why It Is Specified

The contrast worth seeing is on the same screen. UK-authorised CFD providers carry a standardised risk warning that states the share of retail investor accounts losing money with that provider, and the Financial Conduct Authority wrote the specification out in full in policy statement PS19/18.

The percentage must be firm-specific rather than an industry figure. It must be recalculated every three months. Each calculation must cover the twelve months preceding it. And a single retail client account counts as losing money when its realised and unrealised net profits on the restricted products add up to less than zero. Population, period, recalculation frequency and the test applied to each account are all written down.

That specification came out of a measured concern. When the European Securities and Markets Authority agreed its CFD measures in March 2018, it reported that analyses by national regulators across EU jurisdictions found that 74 to 89 percent of retail accounts typically lose money, and put the average loss per client somewhere between 1,600 and 29,000 euros.

Set that beside a sentiment percentage published two paragraphs higher on the same page, with no stated base, no stated period and no stated recalculation rule. Figures of the second kind circulate widely in this corner of the web, including a claim that 95 to 99 percent of traders lose money, which no regulator has published and which sits far above the range the national analyses produced.

Retail Share of Turnover Is Not Disclosed Anywhere Official

Suppose a sentiment reading is precise, its base is stated and it refreshes every minute. One question still decides whether it can tell you anything about price: how much of the market does the measured group represent?

The widest answer available comes from the BIS Triennial Central Bank Survey. For the April 2025 round, authorities in 52 jurisdictions gathered returns from over 1,100 reporting dealers, and the result puts average daily OTC foreign exchange turnover at 9.6 trillion dollars, of which spot accounted for 3 trillion dollars a day, or 31 percent of the total.

That total was 28 percent higher than the 7.5 trillion dollars recorded three years earlier, so the denominator any retail book sits inside is growing.

The survey also breaks turnover down by counterparty. Trading between reporting dealers was 46 percent, trading with other financial institutions 50 percent, and trading with non-financial customers 5 percent.

There is no retail-client category in that breakdown, and the 5 percent line is corporates rather than individual traders. The most comprehensive measurement of this market does not size retail positioning, which means no official source states what share a broker’s client book represents.

The Contrarian Reading, Stated as the Claim It Is

The usual instruction attached to these tools is to take the opposite side once the reading passes some level. Presented as a method, it is doing a lot of unstated work.

Three conditions have to hold for it to make sense. The measured group has to be large enough for its positioning to affect price.

Those positions have to be closed in a way that moves price rather than being netted internally or hedged away by the firm. And the reading has to arrive before the effect rather than after it, which a percentage of currently open positions cannot guarantee, since it describes the present state of a book rather than the flow that is about to hit the market.

The first condition cannot be checked against any published figure, as the previous section shows. The second depends on the broker’s own hedging arrangements, which are not disclosed. The third is a property of the data type.

None of the pages presenting the contrarian reading as a method addresses any of the three, and one of them carries no publication date and no modified date on an article whose most recent data reference is 2023. Exchange-traded gauges are specified more tightly, and our page on an exchange-published positioning gauge shows what a documented sentiment measure looks like when the exchange defines it.

Deciding Whether This Data Belongs in Your Process

The reading is a fact about one firm’s customers, and it is a real fact. Used as a description of who that broker’s clients are, it holds up. Used as a description of the market, it asserts something the data cannot carry.

A reader who wants a specified, auditable number should prefer one that comes with a written method. A reader who finds these feeds interesting should ask the provider what its percentage counts and how often it recomputes, and treat an unanswered question as part of the answer. Where the underlying idea is that hidden positioning can be read off a screen, our page on what a retail forex chart contains takes the same question further.