Traditional Exchanges

Although the problem of latency arbitrage gets much less attention than it should, we were not the first ones to find out that High-Frequency Trading (HFT) arbitrage bots algorithms are making more money than they should (meaning, more money than they would in a better designed system). While our matching mechanism is completely novel, we were not the first ones to propose solutions either.

Michael Lewis' book "Flash Boys" is entirely devoted to the subject of HFT and how "unfair" they are to regular players. While the book makes for an interesting read, the author does not manage to separate the beneficial liquidity-providing function of HFT (market-making) from other, deleterious functions (latency arbitrage and front-running). As a reference to those that have read the book: latency arbitrage is what we are solving with our mechanism, while front-running as depicted in the book is a problem that happens due to the very badly-designed NBBO regulation created by the SEC, or due to bad (or self-serving) order routing by brokers.

Brad Katsuyama, featured prominently in the same book and founder and CEO of the IEX exchange, also doesn't seem to completely understand what the problem is. He has implemented a speed bump in his exchange, a mechanism that artificially delays the incoming orders by a few milliseconds, apparently in order to "slow things down" and prevent those pesky high-frequency traders from benefiting from their very low latency. His mechanism, however, is a symmetrical speed bump, which delays all incoming orders immediately, and as a result it does little to change the latency competition. Bots will just compete to arrive first, so that they can be processed first a few milliseconds later. Not surprisingly, after a short-lived boom due to the increased media publicity, IEX is losing market share. Adding to IEX’s woes, its attempted entry into crypto got off on the wrong foot through a partnership with fraudulent exchange FTX, a partnership that has been rescinded since the fraud became public.

The CBOE exchange has actually proposed a useful variation, which is called an asymmetrical speed bump. This means that cancel requests are exempt from the speed bump, and are processed first. This does make a lot of sense, and actually protects the market makers, but the SEC has rejected the proposal and refused to let CBOE use it after several prominent high-frequency traders lobbied against it, arguing that it "unfairly discriminates against liquidity-taking orders". We will not enter here into a detailed discussion of this ruling, but we do believe that our auction-based mechanism would do better at facing regulatory challenges.

Regardless of regulatory problems, the asymmetrical speed bump would work if implemented, and would protect the market makers from arbitrage opportunities that they are able to spot, just as our auction mechanism does. However, our mechanism offers a significant advantage in that the market maker is also protected in part against the arbitrage opportunities that it does not know about.

In practice there are hundreds of arbitrage opportunities, and for a market maker to avoid losses it would have to watch all the relevant markets and be capable of evaluating all of them correctly. On the other hand, the snipers can easily make money out of the market maker by evaluating a single arbitrage opportunity. This asymmetrical competition means that market makers would still lose a lot of money to arbitrage bots in practice, even if exchanges implemented the asymmetrical speed bump.

It is however fascinating to notice that futures exchanges in the US (that are not regulated by the SEC) are not attempting to use asymmetrical speed bumps, as neither are exchanges in foreign countries, or the unregulated cryptocurrency exchanges (as far as we are aware of), despite what we see as quite obvious advantages of it.

The lack of this technology has serious effect for the liquidity of these markets. In many instances there are several different markets for assets that are financially equivalent.

For example, there are separate markets (with separate order books) for standard, mini and micro S&P futures at CME. The mini and micro markets currently have more volume, and as a result spreads and much lower there – providing liquidity on the standard S&P futures is difficult due to the arbitrage between the markets. As a result liquidity gets fragmented: in each market liquidity is smaller than what would have been in a single joint market, and arbitrage bots reap profits arbitraging between them.

The same applies to separate but similar markets among different exchanges. For example, CME has a market for BRL/USD futures (Brazilian Real futures). However, this market is not nearly as liquid as it could be. It happens that most trading activity on this pair happens on Brazil at the B3 exchange (a pair of inverse USD/BRL – the DOL and WDO series). While it would be in the interest of CME to have a very liquid market for its BRL/USD futures, latency arbitrage between these two exchanges makes that difficult. Arbitrage traders with short cables between São Paulo and Chicago stand to profit, while institutional traders in the US who don’t have access to the Brazilian exchange pay increased transaction costs. This generates a winner-takes-all dynamic, where incumbents are hard to displace, and B3 is the incumbent regarding this specific pair.