Re: <nettime> The secret financial market only robots can see

[David Mandl <dmandl@panix.com>]

Hi Armin--

On Sep 30, 2013, at 9:07 AM, Armin Medosch <armin@easynet.co.uk> wrote:

> In principle, it is important to differentiate between different forms
> of algorithmic trading. There are on one hand, large investment banks
> and hedge funds who hold large portfolios of different types of stocks
> and equities; they also need fast computers and fast lines, but just
> because they need to keep track of lots of different positions and
> their relations to each other - together with news and lots of other
> things happening in real time;
> 
> those are the companies who employ Quants, people with high level
> mathematical and/or theoretical physics knowledge to design the
> software and the 'products' traded, but the trading itself is not
> really high-frequency, the final decisions are still made by humans
> and there are a number of trades a day or even more, but nothing
> approaching nano-second stuff.

Right, I wouldn't call this algorithmic trading at all. This is general risk management: "How exposed am I to interest-rate movements, shifts in volatility, big drops in a particular industry or the market as a whole?" etc. There are standard measures for these things, usually called the "Greeks": delta, gamma, vega, rho.

The trading involved here can be minimal, depending on how often the desk wants to rebalance their portfolio to neutralize these risks. That could be once or a day or once an hour. But the main work and "intelligence" here is in the systems that compute these risk measures, and the people who decide what measures are meaningful to them. The trading doesn't have to be anything fancy. No algorithms needed, usually.

> High Frequency Trading is a special case of algo-trading and that
> really is a world of its own; according to one insider, the big
> investment banks and hedge funds are not really good at it at all,
> because it is based on a different mentality - very much a kind of
> nerd / hacker type mentality - so that mostly new companies are doing
> it who follow this special mindset. the algorithms used are relatively
> simple, you don't ned the brain of a quant to write one, but it has
> to be very reliable; the strategies applied are aiming at very low
> risk as opposed to the risky 'over the counter' deals of hedge funds;
> software base is mostly Linux and open source and the entry level
> for firms relatively low; my source claimed that HFT was actually a
> 'democratization' of speculation, because in a few years everybody
> would be able to do it.

In my experience, the talk about "open source" in finance is exaggerated. All these systems run on Linux boxes, but that's pretty much where it ends. The algorithmic stuff tends to be proprietary, whether it's any good or not, and written in C++ or Java. There's no end of articles about how Wall St. has "gone open-source," and I don't get it. They might use the gnu C++ compiler, or use cvs for source-control, but that's been going on forever. I don't think I've ever worked on a trading or risk-management system that wasn't proprietary. In fact I know someone who knows someone who used to work with a certain Russian programmer who was arrested and had his life more or less ruined because he was suspected of stealing the secret algorithmic code from the firm he was leaving in order to bring it to the hedge fund he was going to. Or so I've heard.

While quants aren't needed to write the trading algorithms, they definitely write a lot of the code that gets baked into these systems--for figuring out what the fair price of some esoteric derivative is, for computing the desk's risk (see above), etc. There's a pretty clear separation of responsibility between the programmers and the quants, though the former know a hell of a lot about the market and the latter know a lot about software. Whatever else you can say about them, most of these quants, at least at the top firms, seriously know their mathematical shit. Many, but not all, of the programmers are equally on top of the technology. This is pretty specialized knowledge, which is why good Wall St. techies are paid as finance people rather than as programmers. Historically, no programmer in any other industry could make anything like what Wall St. tech people made, though I've heard that's changing with some people at Google etc. At the same time, Wall St. firms are getting stingi
 er. (Yeah, things are bad all over.)

> I was also surprised to learn about conditions in this industry. You could say that this was a kind of Fordism of financialism, where you have very few analysts but many coders and data base maintainers; they are all employed with 38 hours jobs, lots of holidays and on the job training and, while salaries are higher than almost everywhere else, they are very much lower than totally out of poportion bankers' boni.

See above. Certainly almost no Wall St. programmer's salary can touch a trader's salary, but they're still far above what other tech people make. It's possible that in continental Europe (for example) some people in finance work 38-hour weeks, but I've never heard of that in the US or London.

> This just confirms that there is a general tendency in society to mystify the workings of machines, whereby the commodity fetishism applied to machines just conceals the real mechanisms of social power as carried out by people, corporations, powerful interest groups. HFT is not that bete noir of banking as what it has been protrayed by some. If it is a good thing I dont know and have serious doubts about the 'democratization'.

I agree with you about the mystification of tech. As my colleague Steve Bong at the Register (http://www.theregister.co.uk/Author/2497/) is so good at pointing out, it keeps charlatans and confidence-tricksters feeding at the trough here and there.

Cheers,

   --Dave.

--
Dave Mandl
dmandl@panix.com
davem@wfmu.org
Web: http://dmandl.tumblr.com/
Twitter: @dmandl
App.net: @dmandl
Instagram: dmandl


#  distributed via <nettime>: no commercial use without permission
#  <nettime>  is a moderated mailing list for net criticism,
#  collaborative text filtering and cultural politics of the nets
#  more info: http://mx.kein.org/mailman/listinfo/nettime-l
#  archive: http://www.nettime.org contact: nettime@kein.org