By Hamlin Lovell
You can now trade at the speed of light (one precise estimate was 11 nanoseconds, from the motherboard to the cord facing the exchange), and hot on the heels of Inside Job, a very different type of finance movie is coming to town: Quant The Film. Anybody who feels they have a unique perspective on quant investing is invited to make a cameo appearance. Sneak preview clips of the in-progress film, shown at London’s Mayfair Hotel on June 16, so far contain the usual suspects: computer geeks talking about locating their servers next to exchanges, and drawing multi-dimensional graphs to illustrate trading strategies.
A New Paradigm Needed for Quant Investing?
One such character is former professional juggler and quant expert Paul Wilmott, who started out advising construction companies on optimally positioning dynamite sticks to blow up mountains. He now wants to explode what he sees as the consensus pillars of quantitative finance. The academic textbook obsession with arbitrage being impossible is already being questioned by the success of high frequency and quantitative algorithmic trading – so quants of all people should be the first to throw out the old paradigm, and embrace the alpha profit opportunities. Related to this, the assumption of market completeness only serves the purpose of simplifying the mathematics: in reality, there are missing markets, and these can be sources of profits. On the topic of Maths, Wilmott unsurprisingly wants to see a wider array of more advanced techniques being used. He laments that the quant is trapped in a straitjacket that represents only a tiny branch of abstract probability theory, ignoring a whole galaxy of mathematical methods. Wilmott scorns what he terms “second-year-undergraduate-level” mathematics and thinks rocket science is really needed.
Can Machines Make Music ?
Given an audience overwhelmingly comprised of quant investors and CTA allocators, the verdict of any vote was such a foregone conclusion that no show of hands was taken. The content of the discussion contained more surprises. The affinity between Maths and Music is well known, and analogies were drawn between investing, music and fashion since all involve herd human behavior, and computers are not yet generating the finest music or fashion, perhaps because humans can spot the gaps that Wilmott identified. Yet discretionary managers are increasingly employing quants to process the growing volumes of data, whilst leaving a human pilot in the cockpit.
Can a Human Hand Switch Off a Black Box?
At what point any pilot might take over from the auto-pilot was the subject of the next discussion: when quants do something sacrilegious to some of them and over-ride their own models. Many quant managers will turn down the volume when they see a price move outside the scope of their models’ look-back period, such as the gyrations accompanying Japan’s earthquakes this year, the flash crash last year, or the September 2001 attacks. This might happen four or five times every 10 years, and big central bank interventions in currency markets also can trigger exits from particular markets. Other managers say they are revising their models every few weeks or months, and others ensure permanent insurance is built into their portfolios.
The Arms Race to Kill Nanoseconds ( and Wikipedia Estimates )
One thing humans undoubtedly cannot do is trade repeatedly in millionths of a second – sometimes prices change too fast for the human eye to see, or for the human finger to punch a key. However, the oft-quoted Wikipedia statistic that high-frequency traders make up 70% of volume was disputed, especially since algorithmic trading is not automatically high-frequency trading – big institutional investors need to use algorithms to stagger large orders over a period. Moreover, the high-frequency percentage varies widely between asset classes and markets. While higher frequency traders tend to be small and pursue a diversity of strategies over multiple time frames, they can often be a stabilizing influence acting in the opposite direction of market momentum. It therefore seems improbable that high frequency is a culprit of systemic risk.
Can Twitter Interpreters Parrot Humans?
The twitter hedge fund launched a few months ago based on this academic study with Professor Johan Bollen of Indiana University, which generated an incredibly high hit rate of 85%. Machines can now read text and be programmed to interpret it, especially where news such as official data releases is very structured. Semi-structured data can also be mined through pattern recognition programs, but completely unstructured data still needs some human discretion: computers have more difficulty in framing the context of information: whether a natural disaster is the real deal, or the title of a book, film or television program. If computers can learn to identify context, semi-automated news feeds can enhance human processes, by modelling the interaction of news and other factors such as changes in analyst estimates. What matters most of all is not the data, but rather how the average or marginal investor will respond to it. But the greatest optimists envisage that five years from now, unstructured data out of context will not be beyond computers – although consciousness and sentience always will be: battle of the quants has not entered the realms of science fiction.
The nuances of news mining
Negative words alone are not enough to signify a pessimistic market implication. The phrase bad news is good news applies here, since even positive spin stories can contain more negative than positive words. However, according to this research, certain very negative words can have substantial predictive power. In fact, news sentiment can be better than one-month momentum at predicting share prices, so says the paper. Some funds are monitoring as many as 2,000 blogs to generate trading signals, and have even codified dictionaries of emoticons that are used in tweets, texts and blogs. Just as Starmine weights analyst estimates according to their historical accuracy, so too some funds are ranking users of social media and applying these rankings to create weightings.