Imagine that you have the fastest connection to the exchange (receive quotes 1 ms earlier than everyone else) for both stocks and derivatives.

How would you benefit from this?

Of course almost any strategy would work better on a faster connection, but I'm looking for strategies that:

  • will only work on the fastest connection (useless/doesn't work on an average connection)
  • are as simple as possible

Links to certain strategies are appreciated.

  • 1
    $\begingroup$ You are looking for HFT strategies. They are simple although often expensive to deploy for exchange colocation, FPGA feed handlers, etc... Many HFT strategies are market making strategies and related arbitrage strategies. $\endgroup$ Oct 25, 2011 at 20:15
  • $\begingroup$ Thanks, do you have concrete example because HFT is too general. I think classical arbitrage and statistical arbitrage would benefit from the low-latency connection, what else? Probably someone can add a link to most popular HFT strategies with detailed description... $\endgroup$ Oct 26, 2011 at 8:33

3 Answers 3


Jordan (@jordan.baucke) in his answer suggests that most latency arbitrages are actually market making strategies, as opposed to classical price arbitrage. While I generally agree, I can think of two exceptions:

  1. Equity price arbitrage in fragmented markets (See the fragulator for more on this). In this environment, negative spreads can arise and the quickest can take a profit. This is certainly a possibility, but probably it is not sizable.

  2. Triangular arbitrage. This is a classical example. It exists (see for example Fenn et al., for a coarse analysis at 1 sec resolution) and it might be profitable and sizable at much higher frequencies. At the 0.1 sec horizon, such opportunities are already much more frequent. Unfortunately, there are no academic studies use UHF data and can confirm this.

  • $\begingroup$ Thanks. I can't find on the referenced website what is "Equity price arbitrage in fragmented markets". Do you have direct link desribing this term? $\endgroup$ Oct 27, 2011 at 5:29
  • $\begingroup$ It isn't an established terminology. The gist of the idea is in "negative spreads". In this answer on quant.SE there is a link to a great animation of the LOB. Somewhere there you can see that at times in the consolidated LOB spreads are negative. $\endgroup$
    – Ryogi
    Oct 27, 2011 at 6:39

You're specifically interested in latency arbitrage (see, for instance, this old WSJ article). This strategy is strictly about being faster than everyone else.

You can imagine any number of instances when this would matter (see this discussion of popular algos). For instance, if you can detect another algorithm trading (this is known as a "sniffer"), and others also detect this, then you may be competing over tiny amounts returns in front-running these algos. Most orderbooks are first-in, first-out, so your position in the queue matters. Similarly, competing for returns in more traditional arbitrage across trading venues depends on speed: your faster connection between exchanges can mean the difference between earning the arbitrage return and not.

  • Market making (to collect a liquidity rebate)?
  • Flash trading (when the order is displayed to internal market participants prior to being distributed to the wider market)? *1

I think more generally strategies that implement pure latency arbitrage are related to transaction market making (providing liquidity within markets or securities) rather than price arbitrage (as they were more classically when you think of low latency market making.)

Check out this webinare - some of the subjects include:

  • What role do data centers play in high-frequency trading?
  • Are proximity and co-location the most critical factors in shaving milliseconds for firms?

1 - Of course it's my understand that most electronic markets now don't allow flash trading?


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.