Background

Concerned by possible manipulation of the voice spot FX market by a cohort of 12 global banks, a coalition of global regulators initiated investigations, with an initial focus on voice spot traders potentially colluding to manipulate the FX rate.

Regulators required each of the banks involved to retain leading law firms to assess the scale of the breach, alongside which the New York Department of Financial Services appointed Devon Capital and one of those law firms as a monitor to investigate one of the banks and its voice spot FX manipulation. Early on in the investigation process, Devon Capital expanded the investigation beyond voice spot trading to include e-commerce, FX derivatives and other risk management transactions

Devon’s expanded approach

With none of the law firms looking beyond voice spot, Devon applied their proprietary eDiscovery platform and searched the trade data to identify the telltale fingerprints of FX manipulation. Machine learning was used iteratively to apply and update the fingerprints across the trading data to identify further instances of manipulation. Devon was able to visualise the FX trading and order book for fixes in currency pairs, and these instances of potential manipulation were then matched with chat messages and organisational data. The outcome - Devon was able to show the results of their investigations on a single sheet

Proven results

Looking beyond the initial concern

Within a week, Devon recognized that the manipulation permeated all aspects of the bank’s FX business – not simply the voice spot market 


Increasing the scope of analysis

We analysed other products (derivatives, futures) and trading methods that were displaying behaviours suggesting manipulation


Inspecting the whole of FX

We examined the FX business in its entirety: electronic business (~80% of spot FX), derivatives with focus on exotic options, M&A and risk structuring including FX hedging of customer acquisitions and other structured transactions


Multiple Big Data Projects

Devon Capital ran four big data projects focusing on chats, market trading data and simulations to help the New York Department of Financial Services (DFS) reach a comprehensive settlement with the bank regarding all FX, voice, e-FICC and derivatives violations, unlike other banks who didn’t look further than voice spot FX.

Collusion to manipulate the Daily Fixing in the FX market

1. Trader0 has a large € order to buy at the fix. Chatroom members aid Trader0 by “building ammo” and “clearing the decks”: 

  • Firms A, B and C net sell orders with other parties (clearing the decks).
  • Firm D transfers its buy order of €xx to Trader0 (building ammo).


2. Trader0 now has a larger € order to buy.
3. Trader0 uses orders of increasing size in a stair-step pattern, moving the market price up. Trader0 has now executed part of his € fix interest ahead of the fixing, leaving a portion still to execute. Trader0 then places an order to buy well in excess of his actual remaining € need at 1.3222.
4. This order sets the fix rate at 1.3222.
5. Trader0 cancels the remaining large order volume quickly after the fixing having only been partially executed.
6. This is FX collusion and spoofing. We show the effect of this manipulation step by step on the diagram below.

after Devon’s pattern recognition identified the suspicious trading behavior chats were then linked to the data to prove collusion

  1. 13:56: Firm A: [nets its sell volume] u shud be nice and clear to mangle

14.01: Firm C [nets its sell volume] you’re all clear

14.06: Firm D: [transfers buy volume to Trader0]

 

Its not just the chats - visualising FX trading

  1. The trades and order book show significant buy-side imbalance. 
  2. What follows is stairstep ratcheting of bid volume
  3. Bidding continues to keep the desired price until the fix is set
  4. The smaller graph below tracks volume
  5. Post-fix, a large volume of bids remain, but are quickly pulled
  6. That conduct may indicate manipulation or spoofing

Devon ran multiple big data projects simultaneously, across the following data sources.

  • chats
  • trading data
  • OCCAM
  • FX traders
  • eDiscovery
  • eFICC

Source: EBS data mine 2.0 historical market data; FCA Final Notice to Citibank on the 11 Nov 2014