The FX Fixing Scandal: How the WM/Reuters 4pm Fix Was Manipulated, and What Changed

The FX fixing scandal was the discovery, in 2013 and 2014, that traders at several of the world’s largest banks had shared confidential client order information in electronic chat rooms and coordinated their trading to push foreign exchange benchmark rates, above all the WM/Reuters 4pm London fix, in directions that profited their banks. Regulators and prosecutors in the UK, US, Switzerland and the EU imposed penalties that the US Department of Justice put at nearly $9 billion for five banks by May 2015, and four banks pleaded guilty to a US antitrust felony.

In short: asset managers, index funds and corporates routinely ask banks to trade “at the fix”, a rate calculated from trading in a short window, which until February 2015 was one minute around 4pm London time. The bank commits to deal at that rate and then trades in the market to cover its risk, which gives it both an opportunity and an incentive to move the rate. The FCA’s November 2014 Final Notices document how traders disclosed their net fix orders to each other, cleared opposing orders out of the way, consolidated orders in one trader’s hands, built positions larger than their client orders, and traded heavily into the window. One RBS example made $615,000 in a single fix. The reforms that followed widened the fix window to five minutes, rewrote conduct codes into the FX Global Code, and changed how banks price and execute fix orders.

This piece builds on how the FX market is structured: the manipulation worked because the fix for each major currency was computed from a single interdealer order book (Reuters Matching or EBS) whose volume a handful of dealers could dominate for sixty seconds.

The misconception: “they fixed the price of currencies”

Headlines spoke of banks “rigging the currency market”. The documented conduct was narrower. Traders tried to move a benchmark for a few seconds or minutes, at specific times, when they held client orders priced off that benchmark. The CFTC orders found attempted manipulation, the FCA’s notices were for failures of control under Principle 3 of its Principles for Businesses, and the DOJ’s criminal pleas were for a price-fixing and bid-rigging conspiracy in euro-dollar trading. Sometimes the attempts failed: in one CFTC transcript, a Citibank trader reported after the fix that he was “hosed”, and on another day wrote “wmr is impossible”.

What a fix is, and why it concentrates risk

The FCA’s Final Notices of November 11, 2014 explain the benchmark in an annex. WM Reuters published rates through the day, and the 4pm WM Reuters fix became “a de facto standard for the closing spot rate”. For the major “trade currencies”, it was calculated from trading on a single electronic broking platform during a one-minute window, 30 seconds either side of 4pm. The CFTC’s orders add the detail: WM/Reuters captured trades and best bids and offers every second, took the medians, and derived a mid rate, with all observations weighted equally regardless of size.

Fix rates are used to value portfolios and indices and to settle derivatives. The CFTC notes that cash-settled OTC FX contracts cleared by CME reference the 4pm fix. The FCA’s 2018 occasional paper on the fix notes that MSCI and FTSE indices tracked by funds with about $6 trillion in assets use the WM/R rate as an input.

How a fix order creates a conflict

The Financial Stability Board’s Foreign Exchange Benchmarks Final Report (September 30, 2014) describes the market structure plainly. Clients place orders to trade at the as-yet-unknown fix rate. Dealers “generally accept these orders and execute them in the market as principal bearing the consequent price risk”, hedging “in and around the calculation window”, and usually give the client the fix mid-rate. The FSB concluded that “it is the incentive and opportunity for improper trading behaviour of market participants around the fix, more than the methodology for computing the fix”, that drove the problem.

The FCA’s annex gives the arithmetic. A bank receives client orders to buy EUR/USD 500 million and sell 300 million at the fix. It nets the two and is left with net client orders to buy 200 million, which it buys in the market. It has agreed to sell 200 million to clients at the fix rate, so:

if it buys below the fix on average, it profits;
if it buys above the fix on average, it loses.

Its own buying pushes the rate up. A trader who buys early, at lower prices, and then buys heavily in the window to lift the fix, sells to the client at a higher rate than he paid. The client, who was promised a neutral benchmark, pays more.

How the manipulation worked

The FCA’s notices for Citibank, HSBC, JPMorgan, RBS and UBS, and the CFTC’s parallel orders, describe the same toolkit using the traders’ own vocabulary.

Traders’ term	What it meant (FCA and CFTC descriptions)
Netting off / “leaving you with the ammo”	A chat member with orders opposite to the group’s direction trades them away with banks outside the room, so the room’s volume in the desired direction stays intact
”Taking out the filth” / “clearing the decks”	A member with orders in the desired direction nets them against opposing orders held by outsiders, removing volume that would have traded against the group at the fix
”Giving you the ammo”	Transferring orders to one trader in the room, so a single trader controls a larger share of fix volume
Building	Taking on extra fix orders from outsiders to increase the room’s share of volume in the desired direction
Overbuying / overselling	Trading more in the fix window than client orders required

The CFTC orders add that participants “used code words to evade detection by their banks’ compliance monitoring systems”, that some chat rooms were “exclusive and by invitation only”, and that one prospective member was invited for a “1 month trial” with the warning “mess this up and sleep with one eye open at night”.

Three documented examples

HSBC, GBP/USD at the 4pm fix. HSBC had net client sell orders of about £400 million. Between 2:50pm and 3:44pm, traders at four firms exchanged their net positions; one wrote that “a few more get same way and we can team whack it”. A trader at another firm netted off with a firm expected to trade the opposite way at the fix and reported that he had “taken him out”. From 3:32pm to 4:01pm HSBC sold £381 million, about 18% of it before the window while the rate fell from 1.6044 to 1.6009. In the first five seconds of the window it offered another £101 million. Its sales on Reuters were 51% of all GBP/USD selling there during the window, and with three other firms, 63%. The fix printed at 1.6003. The FCA calculated HSBC’s profit on the example at about $162,000. Afterwards: “there you go … go early, move it, hold it, push it”.

RBS, GBP/USD at the 4pm fix. RBS’s net client sell orders were £202 million, but it “built” its selling to £399 million through trades with other participants. From 3:50:30pm to 3:59:30pm it sold £167 million as the rate fell from 1.6276 to 1.6233; in the 60-second window it sold £182 million, more than 32% of Reuters sales, while the rate fell to 1.6213. The fix was 1.6218. The FCA put the profit at $615,000. One participant wrote “we fooking killed it”.

UBS, EUR/USD at the 1:15pm ECB fix. UBS’s net client sell orders were €86 million; it built its fix selling to €211 million. At 1:14:59pm it offered €100 million at 1.3092, three basis points below the best bid, which accounted for 29% of EUR/USD sales on EBS from 1:14:55 to 1:15:02. The ECB published 1.3092. Profit on the example: $513,000. A trader at another firm pasted UBS’s earlier comment about “the best fix of my ubs career”.

Worked example 1: what the RBS numbers imply

Computed in Python from the FCA’s figures. RBS bought £399 million from clients and other banks at the fix rate of 1.6218 and sold the same amount in the market. A profit of $615,000 on £399 million means its average selling rate was:

1.6218 + 615,000 / 399,000,000 = 1.62334, about 15.4 pips above the fix.

That edge came from selling early while the rate was still high. The FCA records £167 million sold before the window and £182 million in it, which leaves about £50 million sold at other times or not itemised. (The FCA does not say how it measured the profit, so treat the implied average as illustrative.) The extra £197 million of “built” selling did not hedge any RBS client. It gave RBS more volume to sell before the window and more weight inside it.

For clients selling sterling at the fix, the rate fell 58 pips between 3:50:30pm and the fix. On RBS’s own £202 million of client orders, that is about $1.17 million less in dollars than at the 3:50:30pm rate. That is not a measure of harm, because some of the fall would have happened anyway (heavy selling was coming regardless), but it shows the stakes for the client side of the trade.

Worked example 2: a stylised model of trading ahead of a coordinated fix

To isolate the mechanism, take a simple model (our assumptions, computed in Python). Every £100 million sold moves GBP/USD down 1 pip, permanently until after the fix, and the fix equals the average traded price inside the window.

Independent trading. Three banks hold £600 million of net client sell orders between them, and each executes in the window. The rate falls 6 pips across the window, the fix prints 3 pips below the starting level, and each bank’s average sale price equals the fix. Profit: zero.
Coordinated trading. Our bank (£200 million of client orders) learns in a chat room that the others will sell £400 million in the window. It sells its £200 million just before the window at an average of 1 pip down. The others’ selling then sets the fix 4 pips down. Our bank sold at -1 and bought from its clients at -4: profit 3 pips on £200 million, $60,000.
Coordinated trading with building. It also takes over £200 million of another bank’s fix orders, sells £300 million early and £100 million in the window. The fix falls to 4.5 pips down. Profit: $90,000.

In this model the bank’s profit is exactly the clients’ loss. In the coordinated case, the fix is 1 pip lower than with independent trading, so clients selling £600 million at the fix receive $60,000 less; with building, 1.5 pips and $90,000. Nothing in the model requires an extra trade to be illegal on its own. What made it work was information about other banks’ orders, which is why the FSB’s recommendations and the later FX Global Code focus on information sharing rather than banning trading near the fix.

The enforcement record

The figures below come from the agencies’ own releases and orders.

Date	Authority	Action
Nov 12, 2014	UK FCA	Final Notices (dated Nov 11) fining Citibank £225,575,000, HSBC £216,363,000, JPMorgan £222,166,000, RBS £217,000,000 and UBS £233,814,000: £1,114,918,000 in total after a 30% settlement discount, for failing to control G10 spot FX trading from January 1, 2008 to October 15, 2013. The FCA’s largest fines at the time
Nov 12, 2014	CFTC	Orders finding attempted manipulation of FX benchmarks: Citibank and JPMorgan $310m each, RBS and UBS $290m each, HSBC $275m, over $1.4bn in total
Nov 12, 2014	OCC	$950m against three national banks for unsafe or unsound practices: Citibank and JPMorgan $350m each, Bank of America $250m
Nov 12, 2014	FINMA	Disgorgement of CHF 134m from UBS
May 20, 2015	US DOJ	Citicorp ($925m), Barclays ($650m), JPMorgan ($550m) and RBS ($395m) agree to plead guilty to conspiring to fix prices and rig bids for dollars and euros; Barclays pays a further $60m for breaching its LIBOR non-prosecution agreement; UBS pleads guilty to wire fraud over LIBOR after its FX conduct breached its 2012 agreement, paying $203m
May 20, 2015	Federal Reserve	Cease and desist orders and fines of more than $1.8bn: UBS, Barclays, Citigroup and JPMorgan $342m each, RBS $274m, Bank of America $205m
May 20, 2015	FCA, CFTC, NYDFS	Barclays: FCA £284,432,000 (after a 20% discount), CFTC $400m, NYDFS $485m
Nov 18, 2015	NYDFS	Barclays $150m for misuse of last look, covered in last look explained
May 16, 2019	European Commission	€1.07bn on Barclays, RBS, Citigroup, JPMorgan and MUFG for the “Three Way Banana Split” and “Essex Express” cartels; UBS received immunity
Dec 2, 2021	European Commission	€344m on Barclays, HSBC, RBS and Credit Suisse for the “Sterling Lads” cartel; UBS again received immunity. In July 2025 the EU General Court cut Credit Suisse’s €83.3m share to €28.9m

What the DOJ pleas said

According to the plea agreements described in the DOJ’s May 20, 2015 release, euro-dollar traders at Citicorp, JPMorgan, Barclays and RBS, “self-described members of ‘The Cartel’”, used an exclusive chat room and coded language between December 2007 and January 2013 to manipulate the 1:15pm ECB and 4pm WM/Reuters fixes. They also agreed “to withhold bids or offers for euros or dollars to avoid moving the exchange rate in a direction adverse to open positions held by co-conspirators”. Each bank accepted three years of corporate probation. The DOJ put total fines and penalties paid by the five banks for FX spot conduct at “nearly $9 billion”.

What the Fed and the OCC added

The Federal Reserve’s orders found that the banks’ oversight and controls were deficient, that traders shared confidential customer information in chat rooms, and that four of the banks engaged in unsafe practices in FX sales, including in how they disclosed to customers how price quotes were set. The orders required improvements to oversight, controls, risk management and internal audit, and barred the banks from re-employing those involved. The OCC described an environment in which “unscrupulous traders discussed manipulating foreign exchange markets”.

Individuals and private claims

The corporate resolutions did not translate easily into individual convictions. In October 2018 a jury in the Southern District of New York acquitted three former traders charged over “The Cartel” chat room (Richard Usher, Rohan Ramchandani and Christopher Ashton) on October 26, 2018. In the US civil class action In re Foreign Exchange Benchmark Rates Antitrust Litigation, fifteen banks settled for about $2.31 billion; the last defendant, Credit Suisse, won a jury verdict in October 2022 (the jury found a spread-fixing conspiracy existed but that Credit Suisse had not joined it). In the UK, the Supreme Court ruled on December 18, 2025 ([2025] UKSC 48, Evans v Barclays Bank plc) that the Competition Appeal Tribunal had been entitled to refuse to certify an opt-out class action based on the 2019 Commission decisions, reinstating the Tribunal’s decision.

What changed: the benchmark

The FSB’s 2014 report made fifteen recommendations. The first was to widen the fix window from one minute, noting that the median suggestion from market feedback was five minutes centred on the hour for the major currencies. Others asked WM to use more data sources, asked banks to price fix trades transparently with a spread or fee “consistent with the risk borne”, to separate the handling of fix orders, and asked market makers not to share information about their positions “beyond that necessary for a transaction”.

WM, then part of State Street, implemented the window change on February 15, 2015. According to its November 2015 presentation to the ECB’s FX Contact Group, the calculation window became 2.5 minutes before to 2.5 minutes after the hour for all currencies, with up to 301 one-second snapshots per data source for trade currencies, and Thomson Reuters Matching data was added for the euro, yen, Swiss franc and rouble, with trade rates pooled across sources.

Did it work? The FCA’s Occasional Paper 46, Fixing the Fix? (Martin Evans, Peter O’Neill, Dagfinn Rime and Jo Saakvitne, October 2018), studied trading around the fix from 2012 to 2017 with data identifying individual traders. Its findings:

Robustness rose. A simulated outlier trade had half the impact on a five-minute window compared with a one-minute window, though the authors note the impact was below 1 basis point in both cases, because the use of medians already made the benchmark hard to move with single trades.
Representativeness rose. Short-term price reversals around the fix declined and disappeared from 2015.
Attainability fell. Tracking error for participants trying to trade at the fix rose by a factor of 2 to 5 in the largest currencies, raising costs for those who rebalance at the fix.
Behaviour changed. Dealers traded relatively less before the fix and more during it, and the total fix volume of the banks later fined fell by about a fifth.

What changed: conduct and execution

Chat rooms. The CFTC’s order records that Citibank banned multi-bank chat rooms for its FX staff in early 2013, after the period covered by the findings.
The FX Global Code. Published in May 2017, it addresses the specific behaviours in the notices: confidentiality of client information, handling of fix orders, and conflicts of interest. Its principles on information sharing are the direct descendants of the FSB’s recommendations.
Fix pricing and execution. The FSB recommended that fix trades be priced with a bid-offer spread or a disclosed fee rather than at the bare mid-rate, and that banks handle fix orders through separate processes. Clients can also use fix algorithms that execute across the window on an agency basis rather than handing the bank principal risk. These are covered in FX execution algorithms.
Controls. In practice the Fed and OCC orders meant surveillance of trader communications and of trading around benchmarks.

How professionals handle fix orders now

An asset manager that needs to trade at the 4pm fix asks a set of questions that did not exist before 2014:

Do I need the fix at all? Benchmark-relative mandates create the demand; some managers accept tracking error rather than pay for a rate promised in advance.
Risk transfer or agency? A promised fix rate is a risk transfer priced with a spread. A fix algorithm is agency execution with residual tracking error.
What does the dealer do with my order? The FX Global Code’s execution principles expect dealers to be clear about the capacity in which they act and how they handle orders, including any pre-hedging.
Can I measure it? Transaction cost analysis compares the client’s fills, or the dealer’s fix price, against the benchmark and against the market before and after.

Why it matters for anyone buying “signals”

The fixing cases show how professional FX price formation works: through order flow, information and timing in specific windows, not chart patterns. The documented manipulation was by dealers holding client orders, in seconds-long windows, on interdealer venues. A retail trader cannot see that flow, and a service claiming to know “what the banks are doing” at the fix is selling something it does not have. For enforcement against people who sell signals, see forex signal fraud cases.

The short version

Until 2015 the WM/Reuters 4pm fix was calculated from one minute of trading on a single interdealer platform per currency, while banks promised that rate to clients and then traded to cover the risk. Traders at several banks shared their net fix orders in chat rooms, cleared opposing orders, pooled and built positions, and traded early and heavily to push the rate their way; the FCA’s examples show profits of $162,000 to $615,000 on single fixes. In November 2014 the FCA, CFTC, OCC and FINMA penalised six banks; in May 2015 four banks pleaded guilty to a US antitrust conspiracy, the Fed fined six banks more than $1.8 billion and Barclays settled with the FCA, CFTC and NYDFS, bringing the DOJ’s total for five banks to nearly $9 billion. The European Commission added cartel fines in 2019 and 2021. Individual traders were acquitted in the US in 2018. The fix window widened to five minutes in February 2015, which made the benchmark more robust but harder to attain, and the FX Global Code now governs how dealers handle client information and fix orders. This is a historical and mechanical account, not trading advice.