A practical guide to FX Arbitrage FX Arbitrage is a highly debated topic in the FX community with many unknowns, as successful arbitrageurs may not be incentivized to disclose their methodology until after it is no longer effective. Since the concept of arbitrage is so alluring, many traders are attracted to this concept but do not have many developed resources to find quality unbiased information on the subject. In this article we will discuss some of the most important aspects traders should understand about arbitrage, as it pertains to the FX market. FX vs. other markets FX markets are not traded on an exchange; FX is an over the counter (OTC) spot market i where participants exchange one currency into another. Since there is no exchange, the price that one bank may quote for an FX rate may be different than another. However, with technological developments such as the increase of internet speeds, computing power, and pricing algorithms; FX prices can be obtained from multiple sources, and since banks also trade with multiple counterparties, a close to market price can be created. This means the difference in price between different FX liquidity pools may be minimal. Price discrepancies may only exist for a very small amount of time (measured in Milliseconds ii ) and it is questionable whether these discrepancies provide ample opportunity for profit. Banks and arbitrage One question traders should ask themselves is that if there were a way to profit in the FX markets with no risk, or very little risk, why wouldn t the banks be taking advantage of it? The answer is more complex than it seems, because banks create the FX price and the algorithms involved in price creation consider multiple variables. Also banks typically only attempt to make markets, and do not normally engage in proprietary FX trading on behalf of the bank. So their role is typically that of finding the best price for their clients, and profiting by either charging a commission or a spread markup. Banks have many ways they profit from FX without needing to utilize arbitrage, for example retail exchange rates offered to smaller clients may have a spread of up to 1% of the total transaction or more, which is all profit for the bank. For customers such as travelers exchanging small amounts of money, they don t mind paying such costs (but really don t have much choice!). They don t profit a lot on each deal, but due to the large amount of volume of small transactions, this can add up to a good profit for the bank. Since they have a near monopoly on this activity, they may not see a need for investment in highly sophisticated arbitrage trading that has its own unique set of risks. These transactions are but one way banks profit in the FX market, but it provides a good example why banks are not anxious to develop proprietary arbitrage trading models. Another factor to be considered is volume; typically opaque to the average customer. If a bank has an order to sell 100 Million EUR/USD or more it may not be available at the current price. They may have to sell EUR/USD to multiple counterparties, each at a different price, providing their client with an average fill price. By obtaining the best possible price for their client, they can charge a commission or possibly markup the price by 1 pip, while still providing an excellent price to the client on a large order. If in this price discovery a bank found an off-market price which would normally be considered
arbitrage, they may take it but pass it to their client mixed with the other trades being included in the average fill price. This would be considered a wholesale price in other industries; companies may purchase products from manufacturers and markup the price for a small profit. It isn t much different here, except it is done in real-time and calculated by use of pricing algorithms. So while banks do not necessarily engage in Proprietary arbitrage directly, they have an incentive to constantly find the best FX price for their clients. How an FX price is created Banks have millions of clients that want to exchange currencies, known as deliverables (i.e. you want to send funds from Europe to the United States, whereby Deutche Bank may exchange your Euros to US Dollars at JP Morgan). Since banks have such a large demand for FX known as real money flows they need to actively trade Forex in order to exchange these funds on behalf of their customers. In the Venn diagram (left) you can see how an FX price is created. Banks have customers that want to trade their Euros for Dollars and vice versa (EUR/USD is used in above example). Each bank then places quotes to other banks which is known as price discovery this can happen in a few seconds or less. In previous times this may have involved a bank trader telephoning several banks to find information about the offered or bid rate. Price discovery is usually done by sending bids and offers at different price levels to see if there any takers and if there are not the bank may be required to buy at market rates already bid and offered. In the above example you can see the price between 2 individual banks, RBS and UBS is 1.3866, which is different from the price between UBS and JP Morgan 1.3863 and different from the price between JP Morgan and RBS 1.3864. The number in the middle of the Venn where all 3 circles intersect would be known as the market price if such a thing exists in FX. As you add more banks to the diagram, the price would change. The above example uses a simple average between 2 prices formula to negotiate on a tradable price, and the market price is average of 3 interbank prices. Actual pricing algorithms are of course significantly more complicated and we have not even considered volume or the spread, however it does illustrate how an FX price is created. The major difference between FX and other markets is that on an exchange you can only have 1 price, with 2 types of players (buyers and sellers) whereas in FX you have multiple exchanges known as ECN iii or liquidity pool. To complicate matters, individual banks may participate in multiple ECNs, so it is not always easy to determine which parties are responsible for a particular price creation.
The end result of this multi-node network is a smoothed price that is considered the market price globally. If any bank offered a significantly off market price, traders or other banks could easily take advantage of it and have arbitrage opportunity. Also bear in mind that this is all happening in seconds so without the use of advanced computing technology it would be difficult even to follow this activity. A more realistic diagram might look like this (see right). Buy side vs. Sell side In equities, the sell side is an investment bank who issues a security and sells it to the market, and the buy side are other investment banks and investors who buy the security in the market from those banks. FX is a little different, but a similar relationship exists. In FX, the terminology buy side and sell side is not appropriate as it is in equities. A better description would be liquidity maker and liquidity taker the maker being the banks, and the taker being traders or other banks. FX convolutes this even more, when sometimes these players will have a reversal of roles, or they may be a maker on one ECN and a taker on another. As there is no standard for how to manage FX order flow, banks and other institutions can get creative and have the ability to structure their order book in many different ways. And because this structure is internal and confidential, we can only speculate based on output (which we see in the markets and in public data such as BIS surveys and volume reports iv ). Viability of arbitrage in Forex Developing any type of arbitrage trading strategy requires the following elements: Access to the highest quality price information Fast network connection (Internet backbone) at multiple nodes Sophisticated computer hardware Custom software written to execute the strategy Constant updates of the infrastructure This is expensive and requires a solid understanding of the involved technologies. By meeting the above requirements it doesn t guarantee you will profit, either. Another consideration is that you could build up a specific infrastructure which will take time, and by the time it s implemented; the types of opportunities that exist may change. Of course it is theoretically possible, but how practical is it? While we don t have information about how many arbitrage traders there are, we can assume that at least several well-funded and sophisticated groups are trying to take advantage of such opportunities in Forex. Not only would you be competing with them, as the market and technology evolves, you would be also competing with infrastructure that is constantly being updated. Practical Examples The below screenshot was taken on the interbank platform Currenex during a release of Non-Farm Payrolls:
If a trader was very quick, it would be possible to click buy and sell and capture the difference in profit because the bid is higher than the ask. But unless you clicked both buy and sell at the same time, which would be impossible trading manually, you would risk that by the time you clicked the other side the price would change, and it s possible the price could change in 1 second. With use of an automated system, it would be possible to send a buy and sell order at exactly the same time, and capture the difference. This strategy is near flawless but would have a limit defined by the amount of volume available at that price. Also this opportunity may only happen a few times a month. Latency restrictions Due to the fact that markets are interconnected globally through the internet, latency should be considered when developing any type of arbitrage strategy. Arbitrage opportunities may last for only a few seconds or less. If the latency to your counterparty is 100ms, and it takes 200ms to fill an order (or more), the market may have changed during that time in which case the trade could end up being a loss. Also slippage should be considered, unless the opportunity has a large spread, a small amount of slippage could turn a seemingly profitable arbitrage opportunity into a loss. Latency is volatile and should be monitored in real time. Multiple ISPs are involved with a number of routers and network connections. Just because the current ping to your counterparty is 20ms, that doesn t mean it will always be 20ms. During a large market event, depending on the bandwidth of their network infrastructure, a large amount of price data could create additional delays. This means that any arbitrage strategy should incorporate real time network monitoring as part of the strategy. Fill Risk Another risk in arbitrage strategies is that orders may not be filled quickly or at all. If an opportunity required using more than one counterparty, it is possible that one of them may reject your order. In this case, you would have a naked real position in the market that would be subject to traditional profit and loss. If leverage was used, this could be a big risk if the market was moving quickly. Counterparty risk Any arbitrage strategy requires at least one counterparty to trade with. In the case of multi-leg arbitrage it may require more than one (an unlimited number). Quality of the counterparty should be considered. There may exist an inverse correlation with the quality of the counterparty and arbitrage opportunities. For example, a big name FX bank, such as JP Morgan, likely has a lot of experience and maybe their personnel even have some experience with FX arbitrage. They also probably have an extremely
sophisticated infrastructure, even if only because of their large budget. Finally, due to the large amount of liquidity they process, any arbitrage opportunities may be smoothed by their liquidity algorithms. In contrast, a small underfunded FX brokerage with a small amount of experience, and bare bones infrastructure, may provide many opportunities for arbitrage, but these opportunities come with risks: Failure to fill one leg of the trade due to technical difficulties They can reverse your trades They can go out of business Differences between arbitrage and correlation Many so called arbitrage strategies are actually correlations. Traders believe in their mean-reversion correlations so much that they consider them to be arbitrage when actually they are correlations. That is not to say they are not effective, just that it is not fair to call them arbitrage. Traders seek correlations in all markets but FX correlations are interesting, because of the connections between other markets and money markets which sometimes represent real money flows from one market to another. Some correlations that have been noted by traders: Pair correlations (i.e. a move in GBP/JPY is correlated with GBP/USD and USD/JPY) Stock markets correlated with certain FX pairs, such as USD/JPY being correlated with US stock markets. Correlation between certain commodities, especially Oil, and the US Dollar Correlations can be calculated by traders using Microsoft Excel, with any 2 sets of data (=CORREL). Commonly traders use daily data to calculate correlations, but any 2 data sets can be used. The result will be a correlation coefficient (a number from -1 to +1) 1 being 100% correlated and 1 being 100% negatively correlated. In example 1, data sets A and B are identical, so the correlation value is 1. In example 2, data sets are almost opposite, so correlation is almost -1. Traders who do not have significant experience, and capital / time investment, may want to consider trading a correlation strategy as opposed to arbitrage. Conclusion While price discrepancies have been widely documented in FX, capturing them may require a large investment in time and money, and may require experience in programming and trading. While arbitrage is alluring, traders should consider the risks and pitfalls associated with this type of trading. i http://en.wikipedia.org/wiki/foreign_exchange_market ii http://en.wikipedia.org/wiki/millisecond iii http://en.wikipedia.org/wiki/electronic_communication_network
iv http://www.bis.org/publ/rpfx10.htm