Three months ago, my friend Francois Bellanger from Transit City wrote an article about what sounded like a peculiar book about the “algorithmiticization” of the world stock exchanges. Entitled 6: Le Soulèvement des Machines (6: The Rise of the Machines, soon to be available in English) and published at the excellent Zones Sensibles, this book is presented as written by an algorithm, Sniper (see what the bit full version of the book!), situated at “1700 McArthur Boulevard in Mahwah NJ in an “office” as large as seven football fields”. Mahwah is indeed where the New York Stock exchange is veritably situated. As the book explains in its history of the stock exchanges’ digitization, the race to profits is now counted in micro-seconds and each millimeter closer from the source can be synonym of millions of dollars gained. That is why each server in which trading algorithms work has to be at the same distance from it and the cables that link them to it are precisely measured.
The book opens with a quote by Henry Ford that can summarized its purpose. Ford, who changed the industrial paradigm of his time knew indeed something in the optimization of profits:
It is well enough that people of the nation do not understand our banking and monetary system, for if they did, I believe there would be a revolution before tomorrow morning.
6 describes a history of strategic moves and innovation, combinations of Princeton graduate bankers, mathematicians and computer engineers , in the unique goal to maximize the profits that a company can make through the stock exchange. The latter is now far from the 20th scale of time that remained human, the new paradigmatic scale of time is far below the fastest time of human reaction and the text describes several cases in which a company’s shares would loose the quasi totality of its value in the time of one second and half. It would be nevertheless erroneous to think that algorithms are only made to react fast. In reality they are embedded within a gigantic war game which consequences do not seem to ever be considered by their programmers/investors. I am including here a small portion of the inventory of these algorithms (my translation, original French version at the end of this article):
There is Iceberg, an algorithm that factions an important volume of orders into small fractions in such a way that the initial volume’s integrity escape from the “sniffers’” nose. Interactive Brokers, Thomas Peterffy’s company proposes on its website an illustrative video about these “hidden” orders: small quantities of orders appear on digital platforms as immerged parts of the iceberg before the latter melts little by little as the initial volume is spent in little fractions.
There is Shark, whose task is to detect orders that were passed in small quantities and that maliciously mask a much larger volume.
There is Dagger, an algorithm conceived by Citibank and of which an optimized version was launched in September 2012. Just like Stealth, its concurrent, Dagger is searching for the most “fluid” [liquides] but it has one difference: “A lot of algorithms that are specialized in fluidity research fail to consider their own impact on the markets […]”
There is Sumo, an algorithm conceived by Knight Capital, one of the most cutting edge market operators for high frequency trading, which exchange, just by itself, over 20 billion of dollars per day.
There is also Oasis, Sonar, Razo, Scouter, Aqua, Ninja, Arid, Night Vision, Cobra, Ambush, Nighthawk and many more that continuously spy on each other.
There are also faceless and nameless algorithms like the one that meticulously annihilated the trading capitalization of Bats Global Markets in one second and half. […]
There is this strange algorithm that, on November 30th 2009, then November 30th 2010, then November 30th 2011 and eventually on November 30th 2012, appeared systematically on the NASDAQ platform a few minutes before the markets closed. It was probably one of these programs that was specially conceived to work a few milliseconds before the market close, hoping that its competitors could not react in time.
There is this malicious algorithm that, on July 18th 2012, bought 14,214 shares of Alliance Holdings […] when their price was the highest, before selling back 14,203 shares when their price was the lowest. Everything happened in a few seconds only on eight different platforms. Since algorithms are not conceived to lose money, we can think that this was a voluntary gesture: by playing rapidly on the market’s volatility, this algorithm probably allowed an accomplice to gain a few cents on an associated title.
There is this very obscure algorithm that, on October 1st 2012, placed such a voluminous amount of orders that it was responsible for over 4% of the quotes on American platforms that day by playing on 500 different titles. […] This algorithm did not actually purchased any share but its quotes’ volume was so huge that it occupied more than 10% of the trading operation cables’ bandwidth. (6: Le Soulèvement des Machines, Zones Sensibles, 2013.)
This last example is extremely interesting as its aggressiveness associates both the immaterial realms – its volume of orders – and the material one – the consequence of its order, the saturation of the bandwidth. As I pointed out above, this war game, despite its virtuality, requires its physical means to implement itself and each material detail will very much influence its digital dimension. This is a problem when one considers the international aspect of high frequency trading. The author of 6 historicizes this geographical problem by telling us how Nathan Mayer Rothschild multiplied his personal wealth by twenty in 1815 London by learning before everyone else Napoleon’s defeat at Waterloo and spreading erroneous information. The book finishes with the speculation made by some MIT researchers who imagined that the stock exchanges could be located on archipelagos of offshore platforms between the various international market operators (my translation):
Two MIT researchers seem to have found the ultimate solution to solve this problem: In an article entitled “Relativistic Statistical Arbitage” that was published in 2010, A.D. Wissner-Gross and C.E. Free proposed to build platforms that would be at equal distances from the market operators. These platforms would contain all the necessary financial data, and since the distance to access it would be the same for all, algorithms would take exactly the same time to implement their strategies. This article supplies detailed maps [see below] on which hundreds of platforms would be spread to equal distance of all the world markets. Some of them are situated in the middle of the oceans and since it is probable that, if the project occurs, the technology used to transmit information will be micro-wave, it is possible that the future markets become islands lost in the ocean.
[…] “What is this stuff?” will ask the grandson of a wealthy American retired man when his yacht would run into an artificial island full of hypermodern technology. “That’s the Stock Exchange”, will answer his grandfather before his own retirement pensions might be annihilated by a wild algorithm… (6: Le Soulèvement des Machines, Zones Sensibles, 2013.)
This last sentence is indeed symptomatic of what the book seems to forget in order to go back stronger on it: This War Game and its various strategies have very material consequences. The whole world is still paying the price of this game that blindly control countries’ economies and individuals’ various loans, thus creating a system on which people have actually no power whatsoever to act upon. What this book might envision is that a resistance can maybe be thought from the inside of this system. The hacker group Anonymous might be able to do something about it but, more tangibly and at a less machinist level, that is what the wondeful Rolling Jubilee of Occupy Wall Street’s Stike Debt group has been contributing to do: using the principles of – not so high frequency – trading to liberate people from their student and medical loans. The principle is simple: Everyone can donate money to the jubilee and this money is used to buy debts on the market then forgive them: If you owe $10,000 to an hospital for example, the hospital, rather than waiting for years that you reimburse your debt, might sell it to a company (that will usually use more “convicing” means to have the debt reimbursed) for $500. The Rolling Jubilee is therefore able to buy and forgive a $10,000 debt with only $500. Based on this principle it is currently able to – theroretically as all prices on the market are of course fluctuating – abolish close from 12 millions dollars of debt. So far it actually bought and forgave about 1.2 millions of dollars of medical and student debt.
You can download excerpts of the 6 or order a printed copy (both in French) on Zone Sensibles’ website.
Original French version of the excerpts quoted in this article (all from 6: Le Soulèvement des Machines, Zones Sensibles, 2013.)
First Excerpt:
Il y a Iceberg, un algorithme qui factionne un volume d’ordres important en petites portions de sorte que l’intégrité du volume initial échappe au nez des renifleurs. Interactive Brokers, la société de Thomas Peterffy, propose sur son site internet une vidéo éclairante sur ces ordres dits « cachés » : les petites quantités d’ordres apparaissent sur les plateformes électroniques en tant que parties immergées de l’iceberg, puis celui-ci fond peu à peu à mesure que le volume initial est écoulé par petites tranches.
Il y a Shark, dont la tâche consiste à repérer des ordres passés en petites quantités et qui masquent malicieusement un volume bien plus important. […]
Il y a Dagger, un algorithme conçu par Citibank dont une version optimisée a été lancée en septembre 2012. Comme Stealth dont il est un concurrent, Dagger part à la recherche des valeurs les plus liquides, à un détail près : « Bien des algorithmes spécialisés dans la recherche de la liquidité échouent à prendre en compte l’impact de leur propres activités sur les marchés […] »
Il y a Sumo, un algorithme conçu par Knight Capital, l’un des opérateurs de marchés à la pointe du trading à haute fréquence, qui échange à lui seul plus de 20 milliards de dollars de liquidité par jour.
Il y a aussi Oasis, Sonar, Razor, Scouter, Aqua, Ninja, Arid, Night Vision, Cobra, Ambush, Nighthawk et bien d’autres qui s’épient mutuellement à longueur de journée.
Il y a aussi les algorithmes sans nom ni visage, comme celui qui a méticuleusement anéanti la capitalisation boursière de Bats Global Markets en une seconde et demie. […]
Il y a cet algorithme étrange qui, le 30 novembre 2009, puis le 30 novembre 2010, puis le 30 novembre 2011 et enfin le 30 novembre 2012, apparait sur la plateforme du NASDAQ, de manière systématique quelques minutes avant la fermeture du marché – sans doute l’un de ces programmes spécialement conçus pour travailler quelques millisecondes avant que les marches ferment, dans l’espoir que les concurrents ne puissent réagir à temps.
Il y a cet algorithme malin qui, le 18 juillet 2012, a acheté 14 214 actions d’Alliance Holdings […] alors que le titre était au plus haut, avant de revendre 14 203 actions lorsque le titre était au plus bas, le tout en quelques secondes et sur huit plateformes différentes. Comme les algorithmes ne sont pas conçus pour perdre de l’argent, il y a des raisons de penser que ce geste était volontaire : en jouant rapidement sur la volatilité des cours, l’algorithme a probablement permis à un complice de gagner quelques cents en jouant sur un titre associé. […]
Il y a cet algorithme très obscur qui, le 1er octobre 2012, a placé un nombre d’ordres si volumineux qu’il fut responsable à lui seul de plus de 4% des cotations du jour sur les plateformes américaines, en jouant sur plus de 500 titres différents. [ …] L’algorithme n’a pas réalisé une seule transaction mais en revanche le volume des cotations était si colossal qu’il accapara plus de 10% de la bande passante des câbles destinés aux opérations de trading.
Second Excerpt:
“Deux chercheurs du MIT semblent avoir trouvé la solution ultime pour résoudre le problème : dans un article intitulé “Relativistic Statistical Arbitage”, publié en 2010, A. D. Wissner-Gross et C. E. Freer proposent de construire des plateformes qui soient à égale distances des opérateurs de marché. Ces plateformes contiendraient toutes les informations financières nécessaires, et comme la distance à parcourir pour y accéder serait la même pour tous, les algorithmes prendraient exactement le même temps pour exécuter leurs stratégies. L’article fournit une carte détaillées où des centaines de plateformes sont réparties à égale distance de tous les marchés mondiaux. Certaines d’entre elles se trouvent en plein milieu des océans, et comme il est probable que, si le projet voit le jour, la technologie utilisée pour transmettre l’information soit les micro-ondes, il n’est pas impossible que les marchés du futurs deviennent des îles perdues dans les océans.
L’avenir du trading à haute fréquence repose sur des archipels de machines éparpillés dans le monde entier.
[…] “Qu’est-ce que c’est que ce truc ?”, demandera le petit fils d’un riche retraité américain lorsqu’un bateau de plaisance croisera une île artificielle bourrée de technologie ultramodernes. “C’est la bourse”, répondra le grand -père, avant que peut-être qu’un algorithme sauvage ne réduise sa pension de retraite à néant.”