Remember me

Register  |   Lost password?

The Trading Mesh

HFT and Latency Arbitrage

Mon, 24 May 2010 16:11:00 GMT           

Update 3rd June 2010: As this post originally drew much of its source material from a Themis Trading white paper, in the interests of balance we would also like to include a link to TradeWorx Inc‘s Public Commentary on the SEC’s Concept Release, where TradeWorx CEO Manoj Narang debunks the existence of the Latency Arbitrage strategy. You can find the TradeWorx document at (pay particular attention to page 16, which addresses, point by point, assertions made in Themis's example of a "latency arbitrage" trade. It makes for very interesting reading).

Anyway, here is our original post on the topic…

One of the issues that has caused a fair amount of controversy around high frequency trading recently is that of latency arbitrage.

In simple terms, latency arbitrage is the practice of buying or selling an instrument slightly ahead of other market participants, by taking advantage of small delays in price dissemination. Among high-frequency traders (HFTs), latency arbitrage is an attractive-sounding strategy because it enables them to capture a steady stream of profits, with very little risk. Something of a "holy grail" in fact. So how does it actually work?

In essence, the idea behind lantency arbitrage boils down to two elements.

First, one needs to tap into raw exchange data feeds, rather than sourcing market data via the consolidated quote feeds used by the majority of the market.

Second, trading servers need to be co-located at the exchange’s data centres, thus minimizing any potential latency that would arise from sending electronic messages over physical distances.

According to their detractors, it is these two elements that give HFTs the ability to "see" prices ahead of the rest of the market and to trade on those prices at the expense of other investors.

The Raw Feed Advantage

In the US, most market participants (including professional investors, institutional fund managers and screen-traders) pick up their market data from a consolidated data feed, which includes last traded prices, best bids/offers and market depth from each exchange (plus some ATSs and ECNs). This data consolidation process introduces a small degree of latency, so by the time the data is actually disseminated out to the market, it is already a few milliseconds out of date.

High frequency traders on the other hand, source the raw price data direct from the exchanges. From those raw feeds, they are able to re-engineer the NBBO (National Best Bid/Offer) prices and market depth at faster speeds than the data consolidators (or “plan processors” to give them their official name), thus giving themselves the advantage of being able to see and therefore trade on prices ahead of the rest of the market.

The Co-Location Advantage

In a further bid to reduce latency, most high frequency traders also co-locate the servers hosting their trading systems in racks at the exchange data centres. As anyone with a basic knowledge of physics will know, due to the physical limitation of the speed of light, the less distance an electronic message has to travel from point A to point B, the less latency is incurred along that journey.

This double whammy of being able to source data direct from the exchanges and co-locate their servers at the exchanges’ data centres can give HFTs speed advantages of several milliseconds over other market participants.

The Controversy

The controversy stems from whether these two elements give high frequency traders an unfair advantage, regardless of the investment they have made in order to gain that advantage.

Regarding the raw data feed aspect, in its Concept Release document (, the SEC asks whether delays should be introduced to raw data feeds, so that there is no temporal advantage to be gained from using them versus using consolidated feeds. 

As for co-location, many would argue that it is a conflict of interest for exchanges to offer such facilities, because the exchanges should protect the interests equally of all market participants, rather than offering advantages to a select few.

If the SEC does introduce restrictions on co-location and impose delays to raw datafeeds, it will be interesting to observe the impact on the market. However, any changes introduced by the SEC will only apply to US exchanges , so one possible consequence is that the HFTs may just migrate and take their liquidity across to overseas exchanges. Unless of course market authorities around the world follow suit with similar regulations.

The controversy around "latency arbitrage" is unlikely to go away any time soon, so we will continue to monitor developments with interest.

Further Reading

For further reading on the subject of latency arbitrage and high frequency trading, there is a white paper on the topic written by Sal Arnuk and Joe Saluzzi of Themis Trading, which you can download from their site at

, , , , , , , , , , , , , , , , , , , , , , ,