How Retail Forex Brokers Hedge: A-Book, B-Book, Internalisation and What Listed Brokers Disclose

A retail forex broker that acts as market maker is the counterparty to its clients’ trades, and it manages the resulting risk in three ways: it offsets clients’ opposing positions against each other (internalisation), it hedges some of the remaining exposure with wholesale liquidity providers (often called “A-booking”), and it holds the rest as its own market risk within set limits (often called “B-booking”). Most large brokers run a mix, and the mix determines whether the broker’s profit depends only on spreads and fees or also on whether its clients win or lose.

In short: listed brokers describe this plainly in their annual reports. GAIN Capital’s last 10-K said each client trade may be “naturally hedged against an offsetting trade from another customer, hedged through an offsetting trade with one of our liquidity providers or may become part of our net exposure portfolio”. IG says “the vast majority of trades offset as customers take opposing positions” and that it hedges residual risk when it nears Board-approved limits. Plus500 says “not all net OTC client exposures are hedged” and reports gains or losses on clients’ positions as a separate revenue line. Internalising matched flow saves hedging costs at no risk; holding the unmatched residual adds risk and makes the broker’s result depend on client performance. Our worked model shows how quickly that dependence grows.

The wholesale side of this, where brokers’ hedges end up, is in how the FX market is structured. The conflict-of-interest angle for retail clients is in forex broker conflicts of interest. This piece is about the mechanics.

The misconception: “my order goes to the interbank market”

Signal groups and broker marketing often imply that a retail order is passed straight to banks, or that “B-book” means a broker is rigged against you. Both are oversimplifications.

A retail FX or CFD trade is normally a bilateral contract with the broker. Even a broker that hedges every trade is still your counterparty; it simply opens an offsetting trade with someone else. That is why so-called agency or “no dealing desk” brokers describe themselves in filings as riskless principals: FXCM’s 10-K called its role “a credit intermediary, or riskless principal”. Our conflicts-of-interest piece uses “agent” in this economic sense.
The BIS Triennial Survey counted only $242 billion a day of retail-driven turnover in April 2025, 2.5% of global FX, as seen by the reporting dealers. Because brokers net client trades first, much retail trading never reaches a bank at all.
Holding client risk is not illegal and not, on its own, abusive. Regulators focus on how the resulting conflict is managed, how prices are formed, and whether the broker tells clients the truth about its model.

The three things a broker can do with a client trade

GAIN Capital Holdings, which ran FOREX.com and City Index before StoneX acquired it in 2020, gave the clearest public description in its Form 10-K for 2019:

“When a retail customer executes a trade with us, the trade may be naturally hedged against an offsetting trade from another customer, hedged through an offsetting trade with one of our liquidity providers or may become part of our net exposure portfolio. For naturally hedged trades, we receive the entire bid/offer spread we offer our customers on the two offsetting transactions. For trades hedged with our liquidity providers, we earn the difference between the retail bid/offer spread we offer our customers and the wholesale bid/offer spread we receive from our liquidity providers.”

Trades held in the net exposure portfolio, the 10-K continues, are managed within risk limits set by the Board’s Risk Committee, and “we receive the net gains or losses generated through managing our net exposure”. That gives three buckets:

Bucket	Industry label	Broker earns	Broker’s market risk
Natural hedge (client vs client)	Internalisation	Full spread on both trades	None, if simultaneous
External hedge with a liquidity provider	A-book, STP	Retail spread minus wholesale spread	None, but pays hedging cost
Held in net exposure	B-book, market making	Spread plus or minus the result of the position	Yes

The labels “A-book” and “B-book” are industry slang. You will not find them in regulators’ rulebooks or in most annual reports. Listed firms describe the same mechanics as internalisation, hedging and market risk limits.

What listed brokers disclose

Annual reports from listed retail brokers are the best public evidence for how the model works, because they are audited and because investors want to know how much of revenue depends on client outcomes. Their descriptions, in their own words:

IG Group (annual report, year to May 31, 2025)

IG describes two steps. Internalisation: “IG is the counterparty to OTC trades executed on our platform. We centralise the market risk resulting from customer dealing activity worldwide and offset positions to determine our net real-time exposure. As a result of our scale, the vast majority of trades offset as customers take opposing positions.” External hedging: residual risk is managed “within Board-approved tolerance levels. If exposure approaches our limit, we begin hedging passively to reduce our risk. In the event that we reach this limit, we hedge aggressively to eliminate additional exposure.” IG states that its revenue “is driven by spread, commission and overnight funding charges, it is not driven by client losses”; that is the company’s own characterisation, not an independent finding.

CMC Markets (annual report, year to March 31, 2025)

CMC says trading revenue depends on turnover and on client income retention, “which reflects the proportion of this income that we convert into revenue”, and that retention depends on “risk management gains and losses, driven by our exposure to client positions, hedging activity and associated costs”. In the fourth quarter of that year CMC adopted “a revised market risk appetite, increasing our overall risk tolerance”: it continues “to benefit from natural hedging, with external hedging now applied more selectively”, which it expected to lower hedging costs, “though it may lead to increased earnings volatility”. A footnote adds that, given “significant use of internal hedging, the net cost of hedging cannot be attributed to individual clients”. Trading net revenue was £248.9 million, 73% of net operating income.

Plus500 (annual report and preliminary results, year to December 31, 2025)

Plus500 is the most explicit about client outcomes because it splits revenue into Customer Income (spreads and overnight charges, plus commissions on its futures business) and Customer Trading Performance, defined as “gains/losses on customers’ trading positions”. Its financial risk note says “not all net OTC client exposures are hedged and the Group may have a substantial net OTC position in any of the financial markets in which it offers products”. It sets maximum long and short exposure limits per instrument; “if exposures exceed these limits, the Group either hedges or new OTC client positions are being offered in a smaller size and partially could be rejected”, and it observes “the ‘natural’ hedge arising from the Group’s global OTC clients”. The report describes hedging as “targeted” and “on a limited basis”.

GAIN Capital (Form 10-K for 2019)

Beyond the three-bucket description, GAIN said its net exposure “is evaluated and rebalanced continuously, in real-time throughout the trading day”, and listed among its strategies “lessening the variability between customer transaction revenue… and reported results”, “implementing artificial intelligence driven hedging programs” and “decreasing the cost of hedging”.

Worked example 1: how much of revenue depends on clients’ results

From Plus500’s preliminary results for 2025, computed in Python:

	2025	2024
Customer Income	$739.4m	$667.6m
Customer Trading Performance	($9.8m)	$44.0m
Trading income (sum)	$729.6m	$711.6m
Customer Trading Performance as % of Customer Income	-1.3%	+6.6%

In 2025 clients’ positions made money against Plus500 in aggregate, costing it 1.3% of customer income. In 2024 they lost money, adding 6.6%. The company says it expects Customer Trading Performance to be “broadly neutral over time”. The swing between the two years, $53.8 million, is the visible signature of holding client risk: a market maker that hedged everything would show zero on this line every year.

How the limits work

All four disclosures share a structure that professional risk managers will recognise:

Aggregate first. Client positions are netted across the whole client base, often globally, to get one net exposure per instrument.
Limits per instrument and per correlated group. Plus500 sets limits by asset class, size, liquidity and volatility, and for groups of instruments whose prices are positively correlated. CMC and IG describe Board-approved tolerances.
Graduated response. IG hedges passively as exposure approaches the limit and aggressively at the limit. Plus500 hedges, shrinks new position sizes or rejects new positions.
Value-at-risk and stress testing. CMC runs a one-day, 99% VaR on its residual exposure, and notes that the model does not assume any intraday hedging.

Worked example 2: A-book, internalise-and-hedge, hybrid and full B-book

To see how the choice changes a broker’s economics, we simulated a stylised EUR/USD book in Python. Every parameter is an illustrative assumption, not any firm’s data.

Clients trade €1,000 million a day one way, and the broker charges 0.5 pips per side relative to mid ($100 per pip per €1 million). Spread income: $50,000 a day.
Hedging with a liquidity provider costs 0.1 pips per side.
70% of client volume nets against opposite client flow; the unmatched residual is €300 million.
Over the period the residual is held, EUR/USD moves with a standard deviation of 25 pips, plus a possible average drift in clients’ favour (their “edge”).
The hybrid broker holds up to €100 million and hedges the rest.

Expected daily P&L (from the formulas, confirmed by 200,000 simulated days) and risk:

Model	Clients’ edge -3 pips	Edge 0	Edge +3 pips	Daily s.d.	Worst 1% of days (edge 0)
A-book: hedge every trade	$40,000	$40,000	$40,000	$0	$40,000
Internalise matched flow, hedge residual	$47,000	$47,000	$47,000	$0	$47,000
Hybrid: hold up to €100m	$78,000	$48,000	$18,000	$250,000	about -$530,000
Full B-book: hold all €300m	$140,000	$50,000	-$40,000	$750,000	about -$1.7m

What the arithmetic shows:

Internalising matched flow is pure saving. Netting €700 million of opposite client trades saves 0.1 pips on that volume, $7,000 a day, at no market risk. This is why every scale broker internalises, and why the BIS sees the same behaviour at bank dealers.
Holding the residual turns the client’s edge into the broker’s P&L. In the full B-book, every pip of average client edge on €300 million moves the broker’s expected result by $30,000 a day. Break-even client edge is 1.7 pips; above that, the broker loses on average.
Risk dwarfs income. The full B-book’s daily standard deviation is 15 times its spread income, which is why brokers use limits and why Plus500’s Customer Trading Performance swings from year to year.
The incentive to sort clients follows. A broker can improve the hybrid’s result by holding the flow of clients who on average lose and hedging the flow of clients who on average win. Nothing in the arithmetic requires the broker to do anything to clients’ prices; it only has to choose which risk to keep.

That last point is the heart of the conflict. A broker that keeps unhedged risk has a financial interest in the outcome of its clients’ trades. Whether that becomes harmful depends on whether the broker also influences those outcomes, through pricing, slippage, rejections or stop-outs.

Where the hedges go

The external hedges become the “retail-driven” flow in the BIS survey. The BIS reporting guidelines tell dealers that when retail investors trade through platforms acting as “retail aggregators” (they name Oanda, FXCM, Saxo, Gaitame.com and Gain Capital as examples), the dealer’s counterparty is a wholesale financial firm and the trade is reported as retail-driven turnover with “other financial institutions”. The brokers’ liquidity providers are banks and non-bank market makers, often reached through a prime broker, and they in turn internalise and hedge on the interdealer venues.

What regulators have found

The FCA on CFD providers (2014)

The FCA’s thematic review TR14/13, Best execution and payment for order flow (July 2014) covered 36 firms across five business models, including contract for difference providers. On CFD providers it found that clients’ ability to compare quotes is limited and that “at least half of all trades are effectively ‘captive’”, because a client must close a CFD with the firm that opened it. It described as good practice firms that did not benefit from price slippage against clients, and contrasted them with “asymmetric price slippage”, where a firm “passes on adverse price movements to the client while retaining for itself any movement in the client’s favour”. It also found that firms relying heavily on internalisation were often unable to evidence that it delivered best execution or how they managed the conflicts.

The CFTC on FXCM (2017)

The clearest enforcement case on misdescribing a hedging model is the CFTC’s order against Forex Capital Markets LLC (FXCM), its holding company and two founders, announced February 6, 2017. The CFTC found that FXCM misrepresented that its “No Dealing Desk” platform had no conflicts of interest with its customers and that its role was “merely that of a credit intermediary”, while a market maker spun off from FXCM consistently won “the largest share of FXCM’s trading volume” and rebated “approximately 70 percent of its revenue from trading on FXCM’s retail forex platform” to FXCM, about $77 million from 2010 through 2014. That implies the market maker took in roughly $110 million of revenue from trading on FXCM’s platform over the period (our arithmetic, $77m / 0.70). The respondents paid a $7 million penalty and agreed to withdraw from CFTC registration and never seek to register again.

The lesson is not that market making is wrong. It is that a broker which says it has no stake in your trades, while it shares in the revenue of the firm taking the other side, is misleading its clients.

The US framework

In the United States, a firm that acts as counterparty to retail forex customers must be registered with the CFTC as a retail foreign exchange dealer or futures commission merchant and be a member of the National Futures Association. The US rules for brokers and signal sellers are covered in are forex signal sellers regulated in the US.

What can go wrong

Gap risk lands on the broker. A B-book broker holding client longs in a currency that gaps cannot hedge at the old price. When the Swiss National Bank removed its franc floor in January 2015, brokers on both sides of that risk took large losses, covered in the Swiss franc shock of 2015.
Client credit risk. If clients’ losses exceed their deposits, a broker that offers negative balance protection absorbs them. Plus500 notes such losses “are absorbed by the Group”.
Hedging counterparties. Brokers post collateral with liquidity providers and prime brokers; CMC lists withdrawing balances, reallocating hedge trading to another broker or ceasing business as mitigations if a counterparty weakens.
Conduct risk. Asymmetric slippage, last look style rejections, widened spreads around stop levels, or misdescribing the model, as in the FXCM case. The institutional version of the rejection problem is in last look explained.
Model drift. Both CMC and IG describe taking more market risk in their 2025 reports: CMC raised its risk tolerance, and IG says it widened intraday market risk limits and deployed algorithms to hedge more passively, expecting higher income retention and more short-term variability. A broker’s mix can change from year to year without clients being told.

How professionals read a broker

An institutional reviewer, or a careful retail client, would work through the following:

Capacity. Does the broker act as principal (counterparty) or agent? Read the order execution policy and terms, not the marketing.
Conflicts disclosure. Does it say it may hold client risk, and how it manages the conflict?
Revenue mix. For listed brokers, how much revenue comes from spreads and fees versus net trading gains? Plus500’s split, and CMC’s “income retention” commentary, are examples of where to look.
Slippage policy. Is price improvement passed on as readily as adverse slippage?
Rejections and requotes. How often are orders rejected, and in which direction was the market moving when they were?
Regulatory status. Is the firm registered where you live? Offshore brokers that signal groups push are discussed in offshore forex brokers explained.

The short version

A retail forex or CFD broker is normally its clients’ counterparty. It nets clients’ opposite trades against each other, hedges some of the residual with liquidity providers, and holds the rest within risk limits. Listed brokers describe this openly: GAIN’s 10-K set out the three buckets, IG says most trades offset and it hedges near its limits, CMC links revenue to “client income retention” and raised its risk appetite in 2025, and Plus500 reports client trading gains and losses separately, from +6.6% of customer income in 2024 to -1.3% in 2025. Internalising matched flow saves money at no risk; holding unmatched flow makes the broker’s result depend on whether clients win. The FCA found captive pricing and asymmetric slippage risks in CFD providers in 2014, and the CFTC sanctioned FXCM in 2017 for claiming its “No Dealing Desk” platform had no conflicts of interest while it shared in a market maker’s revenue from its clients’ trades. This is an explanation of broker mechanics, not a recommendation of any broker or model.