Locklin on science

Patented trading systems: cleaning the Augean stables

Posted in financial patents, patent law, patents by Scott Locklin on June 12, 2009

Historically speaking, there has been a lot of social good in issuing temporary monopoly powers to inventors in the form of patents.  The Venetians came up with the modern idea of the patent in 1474. The idea was to improve the industrial productivity of Venice by encouraging European inventors to move there. Most nations of the day didn’t care about productivity, as there were plenty of peasants around to do work. Venice had the most technically advanced civilization of the era: they made do with a very small, well educated labor pool. Therefore, productivity was extremely important to them. It’s little spoken of, but the Venetians were pioneers in the use of mass production and assembly line techniques in their Arsenale, hundreds of years before Eli Whitney and Henry Ford took credit for it. This, not coincidentally, happened at around the same time the Venetians invented patent law. Key to the concept was the idea that they had to reveal the nature of the invention to the Venetian Republic. The rationale was, the majesty of the law could be used to protect the inventor from people stealing his invention during the 10 years he was allowed his monopoly. The less obvious blessing of  disclosure was that the invention would be preserved for the ages for the benefit of the whole nation. This arrangement worked very well. Inventors got rich, Venice became a tremendously powerful trading nation, and this tiny country dominated the Mediterranean for two hundred years hence.


The English had issued patents before the Venetians. English law was a lot more primitive, however. Essentially, the King let his friends have monopolies. Naturally, this was open to all kinds of abuse, and it doesn’t make much sense to let some guy with a funny hat issue arbitrary monopoly power to whoever bribed him enough. Eventually the English came to their senses, and adopted the Venetian way. After they did this, the English were able to become the great military, cultural and industrial power. Later, when the Founding Fathers decided to invent America, they, equipped with formidable historical and legal minds, also adopted the Venetian way. So, America was able to prosper and invent and grow into the first nation that sent men to the moon. America is of course no longer a nation that sends Freemasons to the Moon. We pretty much don’t build physical things at all any more. The four pillars of our present “manufacturing economy” are education, medicine, software and finance: things which don’t have impressive physical embodiments.

Which brings me to the subject line of this post. One of the more famous abuses of patent law as applied to finance was issued recently to Columbia University for an algorithm commonly used to price CMOs. I’m not a structurer, and I don’t feel like digging up the information, but I’m virtually certain the “low-discrepancy deterministic sequences” claimed in this “invention” were used by physicists decades before Columbia’s nerds used them to price CMOs. I know for a fact that Ilya Sobol wrote a ton about this stuff in his papers and books and the poor guy was starving to death in Russia the last I heard. What is more, this technique has near universal currency in the business, as it’s been known for quite a long time now, and is described in many books on structuring. Yet, the patent stands. I don’t think they’re collecting any licensing fees. And everyone already knew about the technique. What was the purpose of this? Vanity? Is someone professionally prospering from filing useless patents at Columbia? Will they start suing what is left of the IBs out there who made CDO markets? It’s a mystery. Columbia has yet to attempt to enforce their patent.

Another amusing patent attempts to patent a form of pairs trade. Now, stab art is something I know a bit about. The mean reverting pairs trade was invented at Morgan Stanley in the 80s by Nunzio Tartaglia and Gerry Bamberger (Ed Thorp, of “Beat the Dealer” fame, also came up with his own version of it, and eventually deployed it with assistance from Gerry Bamberger). The idea is basically this; if you have two oil (or whatever sector) companies, and one of the company’s stock price moves, it is statistically likely the other company’s price will also move, because most oil companies are subject to the same economic forces. There is a time delay between the price moves, as the market digests the information. You assume the information from the first price move, and get paid, more or less to take the risk that the other price will also move. This sort of thing helps make markets more liquid, as you’ll have a whole new class of sophisticated market players who will help provide liquidity. Liquid markets are good. It’s nice to be able to sell or buy when you want to buy or sell. The thing is, you can’t have a public algorithm for providing liquidity. If all liquidity provider’s algorithms were all public, there would be only one price, and there would be no liquidity. Now, imagine what would have happened if Morgan had patented this instead of keeping it a Trade Secret. Everyone else would have read the patent application and immediately tried to profit from the innovation, patent be damned. Who would be able to tell? When you have many stab artists going after the same inefficiency, it rapidly goes away. The vanilla pairs trade worked great in the 80s and early 90s when only a small group of people knew about it. It’s a much more technical and complex trade to do effectively now. So, what we have here is a situation where Morgan would actually have been deprived of its innovation had they decided to use patents.


“Ed Thorp, of ‘beat the dealer’ fame: one of my favorite people: not a trading system patent fan”


So I find it difficult to understand why this guy has filed a patent on any specific embodiment of a pairs trading system. But wait, it gets better. The author of this patent refers to Ornstein–Uhlenbeck processes as “Lagrange Urenbeck.” Now, I guess it’s sort of historically OK to forget about Leonard Ornstein … maybe if you’re a rabid anti-semite. And … um … shucks, I don’t know why you’d leave his name off of Ornstein-Uhlenbeck processes: they were coauthors on the 1930 paper which described this type of statistical process. There is a vague reason to stick Lagrange’s name in there. I don’t know why you’d spell it “LaGrange” though, excepting out of some kind of personal eccentricity. What really bugs the hell out of me about this is why in the name of Brownian motion did he spell Uhlenbeck as “Urenbeck?” The author of the patent has an asian last name … most north asian languages don’t distinguish between the “l” and “r” sound (denoted byㄹin Korean for example). Could he have been dictating all this to a lawyer who perhaps didn’t know any math, or how to use a spell checker on names? Could the use of “LaGrange” instead of “Ornstein” have been some kind of personal tic that Mr. Li had? Certainly, there is no such thing as a “LaGrange Urenbeck” process, though people use Ornstein-Uhlenbeck in pairs trading all the time. I tried to find the company who was owns the patent, but they don’t seem to exist -there is no SEC 13-F form for them anyway, so I know they’re not doing pairs trades. There are a couple of companies with a similar name -maybe it is one of them, but whoever they are, they are certainly not making a profit on this trade. Because regardless of how they spell Uhlenbeck, this is a lousy trading system, and nobody will make money by using it today.


“Poor Ornstein: killed by Nazis, then deprived of his fame by Ben Li and the Westport Financial LLC”



Even if it was a rulin’ trading system, there would be no reason to talk about it for the reasons I said above. Publishing trading systems destroys their profitability. Trading systems are inherently trade secrets; not patentable intellectual property. The way trade secrets have traditionally been protected is via a three step process:

  1. paying the people who have access to the secrets very well 
  2. keeping their money under your control such that your destinies are inextricably linked,
  3. threatening them with dire legal action if they screw you over.

This works fairly well in tightly run organizations. To this day, nobody really knows how Renaissance Technologies makes their money. The three steps I listed above are how they kept the special sauce a secret for over 25 years. Would the world be a better place if they patented their ideas? No. I have a fair idea of how they work (yes, I’ll probably talk about it in some detail on my blog), and there is nothing in their technology which isn’t already in common use elsewhere. If they had patented their ideas, all it would have meant was that Jim and Lenny and Elwyn wouldn’t have gotten paid the way they did. I think that would have made the world a more boring place. The markets would have been more unstable as well.

Next up: a trading system patent application (rather than an issued patent) by … none other than world famous MIT financial engineering professor Andy Lo. I very much enjoy reading Professor Lo’s papers.  However, Dr. Lo made a big mistake. He tried to patent something I know all about. He patented a form of technical analysis (TA) using Kernel regression (KR). Just off the top of my head, I can knock this patent application out on several pieces of prior art. First off, there is an entire book on this written by John Wolberg two or three years before Andy Lo filed his patent claim. Wolberg’s stuff was actually sold on the open market, years before Lo’s patent application: another indication his application is without merit. Sure, it’s not exactly the same thing, in that Wolberg talks about using different TA patterns than Lo does, but this falls into the domain of “obviousness.” It’s also mathematically equivalent: all TA is, is a primitive form of filtering. Sticking TA patterns in a Kernel regression algorithm is non-obvious, though that idea has been around for at least a decade. Using different kinds of TA patterns once you know you can combine them with KR: completely obvious. The worst thing about this: Lo’s technique isn’t as good as the stuff Wolberg talks about. I am about to publish a package which more or less does all of this (it will be released to sourceforge here) and a hell of a lot more. I don’t think Lo’s TA patterns are useful enough to include; therefore this application fails on “usefulness.” Some of my stuff is innovative, and I intend to profit from it. The really useful innovative stuff, I’m not going to tell anyone about. That’s how I protect my IP. Releasing the rest is just a good advertisement for how awesome I am. Finally, as if all this weren’t enough, Dr. Lo released a paper on this very subject, describing his “invention” in much greater detail, and much more eloquently two years before the patent application. You’re only allowed a year before you file: otherwise it’s public domain. This application is not even well written. Lo’s papers are all extremely well written. This indicates to me this was written by a third party who profited from writing it. Dr. Lo is a creative guy, and at least he can spell Ornstein-Uhlenbeck, so I’ll refrain from further beating up his patent application like Ernie Shavers working the speed bag, though it would be easy to do so. I have no idea if the USPTO will approve this, but considering the above, I wouldn’t put it past them.

“Wolberg > Lo”

I can understand the motivation behind patenting software obviousness like 1 Click. I don’t agree with it, and think such abominations shouldn’t be allowed; but I get the motivation -it makes money for large corporations. I have a harder time understanding why anybody would patent something that will profit no large organization. So, I will make a speculation. I suspect most of these are filed by large hierarchical organizations like the late Lehman Brothers which contain lots of bored lawyers who need to create work to justify their large salaries. Law as a profession has a tendency to create busy work to justify itself; hence modern cultural innovations such as warning labels on the business ends of guns. Within vertical bureaucracies, the tendency is probably an unstoppable force. If you stop to think about it, this has a societal effect a lot like the King issuing patents to his buddies. These sorts of patents don’t reward inventors; a good trading system will only work if you keep it a trade secret. They serve no societal good in that, the information contained in a trading system patent is inherently useless, and these patents do not encourage inventors to invent things. They don’t even reward the companies which hold the patents, as they’re completely unenforceable. Mostly, these sorts of patents reward the people who are involved in patent law. When the King of England rewarded his buddies with patents in exchange for bribes, he was helping himself to free money at society’s expense. The clowns filing these patent claims are no different from English Kings: there are just more of them.


8 Responses

Subscribe to comments with RSS.

  1. tadziu said, on June 13, 2009 at 8:44 am

    Ottaviano Petrucci and Antonio Gardano were the very first printers of music in Venice. Those two obtained music printing monopoly from the Dodge and dominated the music printing business (or monopolized) for the first half of the 16th century. The former was the first to obtained monopoly, while the latter was the first to use industrialized method for printing. In fact, to show his contempt for Petrucci, one of the very first printed music by Gardano depicted a picture of a gargoyle playing a lute. The mythical creature was a mocking reference to Petrucci. While obtaining a monopoly, he did not print enough music and also his products were of perceived poor quality.

    • Scott Locklin said, on June 13, 2009 at 6:26 pm

      I was more or less trying to make a point about the dumbness of patenting trading systems, but really, the early legal system of La Serenissima is perhaps even more interesting. The dispute you mentioned could have turned into a duel of sorts. I’m pretty sure the way people settled these things back in the day was via the Arquebus or the sword, the way Cellini did things in Florence.
      Imagine if the Venetians needed to fight duels over innovations in the Arsenale. Would they have been able to mass produce all those guns and boats?

  2. Jim said, on July 22, 2009 at 6:01 am

    The secret sauce for Simons is probably the short-term data they have managed to capture and clean for so many years. I would venture to guess that no one has their database with various time frames to the tick.

    • Scott Locklin said, on July 22, 2009 at 6:10 am

      While I am sure Big Jim Simons has some really great data, I’m certain that it’s irrelevant to RenTech’s special sauce. GETCO probably has better data, and they don’t do what Medallion did. I’ll write about how I think this works eventually. Actually, I might write a paper on it, but the ideas will get popped back here.

      • Siddharth Sharma said, on July 26, 2009 at 7:12 am

        Hey, why don’t you try and make RenTec type of money with the secret sauce instead of putting it out in the public.

        • Scott Locklin said, on July 26, 2009 at 9:59 pm

          Two reasons: 1) I make money as a contractor. 2) I probably (hopefully) ain’t giving away anything important for free.

  3. kwartz said, on August 18, 2009 at 2:15 pm

    Ever had a look at Glasserman’s &c patent on some basic form of importance sampling?

    • Scott Locklin said, on August 18, 2009 at 8:45 pm

      No, I hadn’t seen that one; it’s certainly a funny one. This post came about as the result of my reading up on patent law for a job interview in a patent office at a famous non-trading company. While I didn’t get the job, it certainly filled me with an appreciation for the social utility of patent law … and disgust for the idea of patent law applied to algorithms.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: