Salesman for software company selling to HFT firms: 'We robotise events'
|June 21st, 2012, 08:49 PM||#1 (permalink)|
Aurora, Il USA
Futures Experience: Advanced
Favorite Futures: futures
Posts: 5,839 since Nov 2010
Thanks: 3,275 given, 3,321 received
Salesman for software company selling to HFT firms: 'We robotise events'
Joris Luyendijk talks to a software salesman about adjusting to the world of high-frequency trading and the need for speed
Among the most read interviews on the blog is this computer programmer at a high-frequency trading (HFT) hedge fund. To understand HFT, he explained:
"Compare the movements of shares on the stock market with waves. Our company is like a surfer trying to spot a wave, ride it for a tiny moment and get out again before it breaks. On any given day, our computers buy and sell shares tens of thousands of times, holding them for very short periods, sometimes even less than a minute. No human being, or collection of human beings, could do the volume of trades computers are doing at the moment at stock exchanges across the world."
Today's case study operates in this world. We are meeting in Broadgate Circle in the heart of the City for a coffee and then lunch (ravioli for him). He is a jolly man in his early 40s with an easy manner.
"By now 84% of all stock trades on the New York Exchange are by HFT computers. This leaves 16% for human traders. Automated trading is the future, and my company sells software to hedge funds and investment banks that helps them improve their algorithms – the programme used by their HFT computers. Specifically it helps them respond to events ahead of the rest of the market.
"HFT is about being the fastest and exploiting that. There are two categories. You can invest in the hardware, hoping to gain a microsecond, ie a thousandth of a thousandth of a second. Better cables, more powerful computers, moving that computer physically closer to the stock exchange … Our business deals not in micro- but in milliseconds. Or you can invest in software. Our software analyses the news and interprets its impact on market sentiment – within 15 milliseconds. Usually the market reacts to events with a time lag of seconds, up to several minutes; if you know where the market is heading before it actually does, that's very lucrative. Not only in equity (shares) but also foreign exchange (FX), bonds ("fixed income") and commodities.
"Our clients are programmers and engineers who build the algorithms used by HFT computers.
"How does it work? While the markets are open there is a constant stream of events in the real world impacting those markets; from announcements by companies (takeover completed, quarterly profits lower than expected) and agencies (new drug approved by Food and Drug Administration, merger approved by Antitrust authority), to macro economic events and unexpected things like earthquakes, coup d'etats or exploding pipelines.
"All these come to us through reports by wire agencies: Reuters, Bloomberg, Dow Jones. What our software does is recognise the language in those reports and make them machine-readable. It identifies the company involved, and picks up terms such as "increase", "profits", "warn". Ultimately the software converts these into a one-sentence line that appears in your inbox – say, "British Airways profits up" or "BP production down". It adds a scale of minus five to plus five (red to green), depending on how positive or negative the event is for the company involved.
"Next the 'quants' (maths geniuses), programmers and engineers at hedge funds get to work by analysing how a similar event impacted the share price in the past. Simply put, they feed this into their algorithms, and next time when profits are up by that degree for British Airways their programme will try to get ahead of the most likely movement of the share price in response.
"In the old days, you'd read something in the paper, walk to the stock exchange and adjust your portfolio. Today, a computer decodes the electronic version of that paper and adjusts its portfolio in milliseconds. Whoever gets there first can make millions and millions.
"Most of my clients are in New York and London. Usually it's a couple of dozen maths geniuses fronted by a handful of friendly doormen and posh-looking secretaries. Together they try to persuade institutional investors and extremely wealthy individuals to hand them lots of money to invest with their algorithm.
"You know what's funny? I hardly meet Americans or English people in London or New York. At those hedge funds and quantitative trading desks, it's mostly Russian, Indian, Chinese. Very, very, very bright people with incredible maths skills. It's a world that takes some adjusting to. For one, they don't do small talk. When one of them picks me up from reception and we ride the elevator, I have learnt not to start chatting away about, say, the weather. They simply don't seem to understand. They think you're attempting to communicate something apparently important about meteorological conditions. Same thing with innocent jokes … blank stares. At a typical meeting I get seated at a very long table, with around 12 engineers from the fund opposite me. I talk, they nod and you can tell: they know their stuff.
"I guess most have mild forms of Asperger's, absolutely. I don't mind that, by the way. 'No' really means 'no' for these people, and 'yes' is 'yes'. You might argue that what I do – sales – is the total opposite of quants.
"High-frequency trading is incredibly expensive in the sense that the hardware is very costly, and so is the software. Only a limited number of players can muster those funds. Basically a cartel has cornered the market. If you are rich enough to get a piece of the action, then great for you. If not, well, there it is.
"We robotise events, you might say. What we don't like is journalists trying to write beautifully. If in your first sentence you say 'Facebook' and then you say 'the social media giant from California', our programme struggles. The more standardised the reporting, the better. I do marvel at the inefficiency of it all. We have events translated into language, then decoded back again in machine-readable chunks.
"Cynicism is another thing. Traditionally machines are not good at understanding what you mean by 'IBM is so absolutely, totally great … not!' We are getting very good at this, picking up nuances. When people include emoticons, we include that.
"Apart from news wires, we are also analysing social media, as these can be important for market sentiment, too. We work across the world and we've learnt not only to translate but also decode from one language to the other. Take microblogging in China, which is huge. Suppose you want to monitor the response to a new product by Sony. Chinese people are very unlikely to say 'This new product sucks'. They will say it's 'not as good as we had hoped'.
"By now I'd say we are getting that 98% right. You don't need a literally perfect translation, with a one-on-one correspondence between the languages. You need to get the central message, the 'product sentiment'.
"The chances of speculators manipulating social media for their own purposes? Well, this is not our concern, all we do is extract whatever is being said and make it machine-readable.
"There is safety in numbers, in the sense that if you follow hundreds of bloggers, it tends to cancel or filter out the nonsense.
"There are cases of false rumours ricocheting through the system like an echo chamber. At some point somebody began to tweet that the Swedbank ATMs in Riga were empty because the bank was bust. It even got picked up by respectable wires, and Swedbank ended up tweeting photos of functioning ATMs to quell it. There is a risk here, yes, as so much in social media is about passing on others' stuff without checking it.
"We have an archive of all the news wires of the past decades. Everything that ever happened in those years is in there, allowing you to identify patterns. Basically, anything you and I could think of has already happened somewhere at some point in history. You want an example. The Facebook IPO. You know it's being hyped so you go back in history to earlier hyped IPOs to establish a pattern you can trade on. For instance, you see how the big banks will prop up the price on the first day of the IPO, then let it collapse.
"Theoretically you could feed events for the entire world into a programme, but that makes for a very slow programme. So we offer sectors – oil or pharma – or countries or regions.
"For Greece, you might programme for the leftist leader Alexis Tsipras. You would need a meta-interpretation, though, of whether a Greek default would be good or bad. Once you have that, you can determine for every statement by Tsipras how this will make default more or less likely, and by what degree. This will allow you to trade on the possibility of a default ahead of others.
"In high-frequency trading your computer buys a share as its price is moving, then sells it immediately on to the next buyer, all within that same price movement. The algorithm is a middle man, exploiting a tiny advantage that comes from being there faster. I suppose high-frequency trading has no value to society. That's not our concern. At the end of the day it's a buy-sell market.
"Perhaps in the future we can cut out both the PR departments at public companies issuing the misleadingly positive press releases, as well as the journalists passing these on.
"A hedge fund, or companies themselves, or perhaps some other authority will issue machine-readable statements. Then everyone will have the same information at the same moment in time.
"It's not easy getting appointments in banks to promote our product. You can't just call up Goldman Sachs. Trade fairs are good places to establish contacts. I remember when we started out and I got my first appointment with a vice president (VP) in a bank. Boy, was I excited and proud. So I went to see this guy, who had his own phone – a big deal in banks, apparently. He had his own office, too. A tiny room, but with a window – another big deal in banks, apparently. Then I discovered decisions are taken far higher up the tree. Big banks have thousands of VPs – it's quite a junior position!
"I rarely get to speak to board members of banks. It's interesting how low key they are. Traders pull up in their Ferraris, they want to be seen. Board members arrive in silent limousines with tinted windows and quietly take the elevator to the top floor where there are deep layers of carpet, everybody speaks in hushed voices and there are lots of beautiful men and women asking if you want more excellent coffee.
"Investment bankers and hedge fund types are strikingly competitive and seem to turn everything into a game of one-upmanship. One celebrates his child's third birthday by hiring a big boat, so now his colleague has to get an even bigger boat from his child's birthday.
"How I know these things? They tell me. They laugh as they tell me, making fun of one another. But they all go along with it. It's never enough and they never seem satisfied.
"They buy an expensive yacht with their bonus only to discover even bigger yachts in the harbour. They buy an $8m apartment on Central Park not realising they have to come up with $9,000 in service costs each month. This is how they get trapped. Or they trap themselves.
"Is it all about greed? With the quants I feel they derive genuine pleasure from cracking problems, from writing the best and most elegant programme. Making more money than anyone else is a marker of that, an expression they are the smartest guys in the room. At some point the money becomes a symbol, does it really matter whether you have 300 or 400 million?"
Salesman for software company selling to HFT firms: 'We robotise events' | Joris Luyendijk | Comment is free | guardian.co.uk
|The following user says Thank You to kbit for this post:|