Last night I heard a very interesting talk on innovation and high frequency trading by Professor Donald MacKenzie (a financial sociologist from the University of Edinburgh). As he told it, a Matching Machine, designed by Josh Levine, is – if I have understood this right [I’d greatly appreciate any corrections from people who properly understand HFT, for instance] – at the heart of high frequency trading run initially out of a company called Island based in Downtown Manhattan. It’s evolution from the periphery to centre stage in share trading was used by MacKenzie to tell an interesting tale of innovation. That was a tail of innovation as bricolage and ‘hinging’ between linked ecologies in a system (per Andrew Abbott) to allow an innovation to move from being minor to major. Put another way, high frequency trading ceased to be a small part of the share trading system, and shifted to become the main plank (or macro) of the system. Put more simply, a particular innovation became elemental to the system.
The story begins with the automation of the Small Order Execution System (SOES). A group of traders began to use this system to pick up stale prices (offers to buy or sell shares which were left outstanding even though the share price had moved beyond that price range). The traders started to become known as SOES Bandits because they royally skewered the conventional traders who had failed to sweep up the stale priced trades.
What Levine did was, as told, extremely simple. He built a matching machine which automatically matched offers to buy small orders with offers to sell. That is, he built a piece of software to do it. The trick was that he built this software to run at a speed of 2 milliseconds per trade. This he did through an ingenious but simple software design that did not need to save many offers to sell or buy to disk while they waited to be matched, because it quickly deleted matched sale-buys once completed. This left enough space for unmatched trades to be matched immediately. The difference between that and the conventional matching was 2 ms to 2 seconds; a difference of 1000 times. To the human mind it was instant and automation meant – I assume- more trades could be made whilst prices moved. Also, it used a smaller ‘tick price’ (difference in share price) so that smaller differences in shares could be priced in (1/256 of a cent to 1/8 of a cent under the conventional system). There were more differences to trade against. APIs were built into to enable the system to interface with other systems.
The impact on time is worth dwelling upon. The length of cables between trading platforms slowed down the trades more than the software. Trading platforms began to move towards Island; literally into the basement. There was a certain amount of negotiation with the SOES traders to keep the trades simple enough to allow the system to work. Prices were kept low and there were rebates for certain types of failed trades. There was a geographic coming together of social, economic and technological influences. A new, quick and (as the flash crash showed) sometimes unpredictable (and so perhaps dangerous) market borne.
There are other interesting dimensions to the story. Competitor platforms spread out from Island (partly programmers leaving island to assist others). Levine himself helped a competitor build a better matching machine for themselves. The culture was part collaboration and part capitalism. My own interpretation whilst listening was that it was a bit like gaming: maybe they wanted competitors to get to a standard whereby there was someone worth trying to beat. As one of the audience said, “These are such odd people to find at the heart of capitalism.” Part collaboration, part competition.
To start to see what this might have to do with innovation which might be generalizable, Mackenzie talked of the way in which innovation reached across three different ‘ecologies’ or sub-systems. I am not entirely sure I followed what these subsystems were. One was regulation: the SEC had a regulatory system premised on a degree of scepticism of the privileged insiders running trading (where there were clearly some dubious and nepotistic practices). Another was the value system of the traders themselves. The third, I guess, was a programming value system which promoted speed and elegance. Mackenzie’s critical insight seemed to me to suggest that what drives successful innovation are ‘hinges’: strategies which work in more than one ecology, that hinge between the different systems.
Thus the SEC, suspicious of the privileged insider status of normal traders, had a regulatory system that was conducive to the SOES traders. The simple, quick programming value hinged between the smallness of the trades and the ability to scale small unit profits into high volume profit. The success of the system (and who’s to say that as well as efficiency there was not also the Pavlovian gratification of instant trading) led to its shift from a small part of the periphery to the core of the system.
Now, what has that got to do with law? Perhaps nothing. Perhaps both sides of the Reinvent law debate will enjoy thinking of innovators as Bandits without being too serious about it. But perhaps there is more to it. The talk came in a week when I have met two innovative legal service providers intent on hinging between the more for less imperative of GCs (a business driver); the ability to deliver contract management and analysis more quickly than conventional providers (a systems solution); and, a particular view of how contracts and contract negotiation works which stands back from bespoke micro-management of contracts to look more at the social and economic effects of contract negotiations (a value proposition). Whilst not enabling instant negotiation of contracts, and not being able to dispense completely with micro management of contract terms, it has the potential to significantly reduce and rationalise the process of business to business contract negotiation.
The significance of this depends on many imponderables. One is the ability of NewLaw to deliver genuinely quicker, better contracts. I believe the most far sighted lawyers are those best able to understanding contracts at the macro, behavioural level. If their systems for understanding and delivering such contracts work they will appeal enormously to in-house legal teams and the business units they serve. I wonder how many lawyers have the data and theoretical knowledge to build these systems; but some – from what I have seen – do. Those who can develop build and improve the knowledge and systems quickest will likely gain significant market share if they can scale it. Quicker than one can say PriceWaterhouseCoopers.
The second significant imponderable is how big is this bit of the market. This is interesting because there comes a point at which such firms may become big enough to attract Big Law into their orbit, rather than the other way round. The micro may become the macro. I instinctively doubt it; the two markets are somewhat separate and there would need to be good reason for these middle market systems driven firms to be able to disintermediate between BigLaw and in-housers, but I doubt it a bit less today than I did on Monday.
A final imponderable is whether Newlaw will have its own equivalent of flash crash moments. Systemic approaches may be more efficient and effective, but they may also be prone to systemic, and therefore, costly errors. Taking some human error out may also take some human good out. The extent to which systems can be developed responsibly and robustly is difficult to know. Even harder to test. Regulators, I’d guess, and judges in particular, will be more sceptical of such systems and (I would guess) quicker to punish them for their errors. This may inhibit innovation but also reduce risk. The Bandits will have some hard yards before they find open country.