The Dynamics of Capitalism

Mountains and waves. The tram pulls in at the terminus, and as I step out, that's what I can see: mountains and waves. I'm standing on the bank of the Zürichsee; to my right is the city, pretty, clean and dull. In front of me the lake stretches out, its waters choppy in the fresh March breeze. All around the mountains, daintily capped with snow, hem me in.

Behind me is the railway station and behind that is an old textile mill which has been converted into office space. At one end of the building is a glass cube, about four stories high. This is where I'm headed. It's the headquarters of Olsen & Associates, the company I have come to see. In the building's facade, made a single giant mirror by the angle of the winter sun, the mountains from across the lake are reflected.

Olsen and Associates (O&A) are part of an increasingly common phenomenon: they sell predictions for price fluctuations on the international financial markets. Financial modelling is another of those industries which has blossomed parasitically on the back of the personal computer boom, and it's easy enough to find a trading model to suit your needs whether you dabble in a bit of dealing in your lunch-break or whether you do it for a living with a large investment fund at your disposal. Less easy is finding a model that works. And if you are lucky enough to find that, then the chances of being to understand why it works are almost nil, even if you're the person who wrote it.

But I've come to O&A because they are supposed to be different. Founded in 1986 by Dr. Richard Olsen, they have quickly established a serious reputation in the financial world - not an easy thing to do in an arena where computer modelling, although dabbled in by many, is still regarded as something of a pipedream. By all accounts, though, O&A is a very unusual company - as far as I've been able to make out it's part academic research unit, part blossoming business, and part vision.

I cross the railway tracks and find my way around to the cube's entrance. The door is open, and there doesn't seem to be anyone about, so I go in. Inside there is a metal stairwell and four buzzers. All of them have "Olsen & Associates" written on them. I press them all, one by one, and nothing happens. So I walk up a flight of stairs, open a door at random, and go inside. And then come straight back out again, for heading towards me at a great pace from between the pleasant banks of shelving stacked with econometric journals is an extremely large brown dog. This is not the image I had of Swiss banking.

Eventually I find the reception desk (located on the top floor, as it turns out) and am introduced to the coffee machine whilst Dr. Olsen is informed of my presence. This is no ordinary coffee machine. No packets of Lyons Continental Blend for this baby. Here you simply put the roasted beans into the top, select for strength and style, and press a button. Perfect coffee every time.

I'm busy trying to break this wonderful contraption when I hear a soft, slightly stilted voice behind me. "So, do you like our new coffee-maker?" It is the voice of Richard Olsen, and he welcomes me to Zürich whilst removing the empty cup from my hand and showing me how to fill it without causing thousands of francs worth of damage. "And how do you like Olsen & Associates?" Well, I tell him, it's not quite what I expected.

As I'm to discover, that was exactly the intention. When Olsen set up his company in 1986, giving it a structure that had never before been seen in the Swiss banking community was very much part of his vision. O&A is very unhierarchical, very informal. There are very few ties to be seen, and quite a lot of jeans. The phones don't ring too much because within the office everyone communicates by email. The whole atmosphere is very West Coast, very Wired. "I'm a believer that in today's environment effectively the employee owns his job," he explains in slighty faulted English (English is the office language at O&A). "Provide him with the liberty and the tools to make it his castle, and then you make the best out of him and the relationship. In part this is rationale, in part this is thinking about evolution of the labour market, and in part it's the need of just making life more enjoyable. I have a vision that this company should be a model for other companies to go by."

Olsen has lots of visions. The Swiss son of a banker father and an academic mother, when he was twelve years old - when the rest of us were worrying about he remembers asking himself why there was no scientific technology in banking. It wasn't until he got to university (he did a PhD on the Interaction between Law and Society at Oxford TK) that he realised that the reasons had very little to do with practicality and rather a lot to do with convention. Having little time for convention, he set out to change things. He worked as a floor trader for a while, and discovered that whilst "traders were using my concepts in practice without realising it," they were also making decision after decision with almost no support of any kind. He saw in this a business opportunity, the possibility that an organisation which developed and sold value added information could flourish. Perfect predictions were not what was required - even if possible, that would just put everyone out of a job. What was needed was quality information that could tell a trader which way the markets were moving, give him or her a baseline of trends from which to work from, let him or her quickly gauge how risky a particular trade was. And Olsen saw in his idea more than just a business opportunity. Imagine if it were possible to do this, he thought to himself. Imagine if you could provide a tool that everyone could use to spread their risk, but which wouldn't have the side effect of changing the structure of the markets. The whole nature of capitalism would change; financial projects could be longer term, less businesses would go bancrupt, people's savings wouldn't be wiped out. In a very real sense the world would be a better place.

"You have to understand one thing about Richard," I am told by one of his employees over lunch. "He could have spent his whole life playing golf if he'd wanted to." The employee was referring not to Olsen's handicap but to his wealthy background. Olsen set up O&A not to make money - in contrast to most who attempt to predict the financial markets - but because he believes that his product can change the world. He admits this to me rather sheepishly: "One of the deeper motivations for me to set up the company is actually my belief that it's horrendous what we're doing to nature. If you see American tourists in slacks up in the mountains walking around on the glaciers taking photographs I think they're well prepared in comparison to how we're prepared to tackle our current environment." He grins and goes to the window, pointing out to the glittering Zürichsee. Yachts are moored all along the water's edge. "Businesses are like boats," Olsen says, "and boats need to attract water - or liquidity - in order to stay afloat." Trouble with water is that it's full of waves. And with waves come risk. "Basically what we're doing here is, we're forecasting the waves. We're like a powerplant on the seashore which extracts energy from waves. This is what we're building. It's an abstract powerplant. And obviously it's not electricity which comes out, but it's cash with which you can then buy electricity." It's a slightly bizarre image, but what he means by it is this: the viccissitudes of the financial markets are a potential problem for everyone, whether it is an individual investing with a pension fund, or a multinational buying up raw materials. At the same time though, the markets are an incredibly useful way for companies to raise funds quickly. What is needed is a way to keep the advantages whilst limiting the disadvantages. Olsen believes that this is what O&A is offering the world. By forecasting the viccissitudes they allow investors to minimise risk whilst still being able to invest and make money. If it works, then markets will have been tamed. They will no more be wild animals but will instead be beasts of burden.

Olsen's enthusiasm is infectious, and it certainly infects all of his employees, right the way from the relative old-timers to Mark Lundin, a physicist who just a few months ago gave up a research position at CERN to come and work for O&A. What convinced him to work for Olsen? "When I was trying to decide, Richard said to me: 'Economics is in the same position today as quantum mechanics was in the 1920s.' That made up my mind." Like the rest of the team, Mark is convinced he is involved in an area of science which is about to explode - and where an individual can still make his or her mark, a combination which is increasingly rare these days.

But wait a minute. The centrepiece of modern economic (or "econometric") thinking is something called the Efficient Markets Hypothesis (EMH). In its strong form, this says that the current price of a piece of stock already reflects all the relevant information that's available. The idea is that markets are efficient and prices only respond to unpredictable news. So where was Olsen's value added information going to come from?

Olsen was convinced that the EMH didn't tell the whole story. Not only that, but with the entire financial industry paying lip-service to it it served as a psychological barrier that prevented anyone from breaking through to a greater level of understanding of price fluctuations. Why shouldn't price movements obey rough laws, asked Olsen, laws which at least constrain the behaviour of markets if not totally determining them? Laws like those which physicists discover - laws which describe the ingrained habits of the world.

Certain that his hunches were right, in 1986 Olsen hired a bunch of physicists, put them in this glass cube on the banks of Lake Zürich, and let them loose. The key thing at that stage was data. Classical economic theory was generally based on studies of daily or even weekly market data. Although there were many price changes during each working day, the only ones that tended to be recorded were closing prices. This meant that you had around 250 data points a year to work with. For Olsen and his team this was not enough. They began to collect tick-by-tick data on the foreign exchange and interest rate markets from the stock market feeds of Reuters, Knight-Ridder and Telerate. Suddenly, they had anything up to 18,000 data points each day. (Indeed, O&A's database is now so large that they provide all of the data for RiskMetrics, the well-known value-at-risk methodology of US investment bank JP Morgan, recently made available in the public domain.) "One econometrician said to us, 'Why are you collecting all of this data? Why don't you just generate it randomly?'" recalls Michel Dacorogna, one of Olsen's early recruits. "The data the old economists had told them the markets were a random walk. But this new data was saying other things, although many people seemed too scared to work with it." A random walk is a relatively simple idea, nicely interested by something statisticians call "the drunken sailor problem". Visualise this: tanked up with rum and good times after nine months at sea a sailor stumbles out of a quayside pub with the idea of heading back to his ship. He can step either to the left or to the right, but he is so drunk that it doesn't make any difference to him which way he steps - and the direction of each step relies in no way upon the direction of the previous one. This is a random walk.

Dacorogna had just finished a Postdoctoral thesis in Solid State Physics at Berkeley when he came to work for O&A; he had been working out how compute the distance between atoms in a crystal lattice from Schrödinger's quantum mechanics equations. I ask him why he had suddenly switched to working in the financial world, and he tells me that he was motivated by the research, by the idea of being at the frontier of science. This is not the answer I expect (I thought he would talk about the money). Dacorogna had been convinced to make the change when another of Olsen's physicists, Ulrich Müller, had shown him the work that Benoit Mandlebrot, the mathematician who discovered fractals and pioneered their use, had been doing back in the 1960s.

Financial data and fractals have always gone hand in hand - one of the first data sets that Mandelbrot had applied his theories to was that of fluctuating cotton prices over several centuries. "Forget about chaos," Dacorogna tells me. "We have never found any chaos in our data. There is no simple formula that generates such behaviour as is found in the foreign exchange markets. A fractal is a more general thing. A fractal produces a measure." The measure Dacorogna, Olsen, Müller and the rest of the team were really interested in was that which would measure time. One of the reasons that physicists like Müller and Dacorogna were willing to pitch in with Olsen was that he had some very interesting ideas about time. Hitherto, economists had always thought of time as outside of the system, a straightforward linear measure that all other changes could be related to. But the idea which fired up the Olsen team (as only physicists can be fired up) was that time was instrinsic to the system - that it was somehow dynamic and that a measure of time was needed which could stretch and contract as the system changed. I asked Dacorogna what this meant. "Time eats things," is what he said. "We define time dynamically." Perhaps it's difficult to explain the massive impact of this statement to someone without a background in science or philosophy. With few exceptions, the history of science has been characterised by the tendency - almost never questioned - to use time as an independent measure of an experiment (or of the world in general, in the case of philosophy.) Think of all those chemistry graphs you drew up in school, which were always plotted against time. In the twentieth century, physics has finally begun to break free of this notion that time is independent of space and action. And now the Olsen team - perhaps because a large proportion of them were physicists, and so were doing what came naturally - were junking the idea that time was an independent measure of the financial markets. It was as if they were introducing relativity theory to economics.

Olsen's explanation is just as visceral. "Actually what you can do is, when there is a price movement, you can say how much time has elapsed. Market time. Suddenly you realise that time is very, kind of, you know... you can put your hands around it." But how could fractals connect price changes and time? I still hadn't understood, so I turned to some of the many papers which O&A put out when they were still more academic research centre than rapidly growing business.

This is what I found. Think of the mountains reflected in the windows of their offices. The silhouette of the mountain range is neither one dimensional nor two dimensional - it is somewhere in between, and is said therefore to have fractional, or fractal dimensions. A piece of paper scrunched into a ball is a fractal object; it has more than two dimensions but less than three. Fractal objects have two main characteristics. The first is that they are "self-similar across different scales". In other words, I can look at smaller and smaller portions of them, and the image I get will still be pretty much the same (imagine zooming in on that craggy mountain range.) Secondly, the apparently random nature of a fractal pattern is not really that random at all; each point is constrained by the position of the point which precedes it.

In the same way that this is true of a mountain range it is true of the price fluctuations in a capital market. Because each price depends to some extent on the previous price, the market is not a random walk, despite what the EMH says. Remember our drunken sailor. One of the things that made his random walk a random walk was the fact that each step he took was totally independent of the previous steps. And amazingly the self-similarity across scales holds true too - believe it or not, in the foreign exchange market the pattern of volatility measured at ten minute intervals is similar to the pattern of volatility measured at hourly intervals, and similar again when measured at daily intervals. Olsen's team found that this scaling law applied for lengths of time up to two months and beyond. Not only that, but the scaling law also holds for the frequency with which prices change direction. Market prices didn't obey a random walk however you looked at them.

So if the EMH is wrong, and price does not obey a random walk, what does it obey? After lengthy examination of their ever-expanding database (today the largest of its kind in the world), the team discovered that foreign exchange prices had a "drift exponent" of 0.58. Back to our drunken sailor. If his quayside stumbling is truly random, after 100 steps he will end up neither to the left nor to the right of where he first began (tossing a coin 100 times you'd expect to get around 50 heads and 50 tails.) In this case, we can say that his drift exponent is 0.5 - i.e. that he has no bias towards the left or the right. But say he has a dim idea of where his ship is, despite all the rum he has drunk. Now after 100 paces he always ends up somewhere to the right of the pub door - and this means that his drift exponent is no longer 0.5 but nearer, say, 0.6. (If for every 100 flips of a coin you got 60 heads and 40 tails, you'd begin to get pretty suspicious - in fact, you'd think you had a biased coin.)

To put this into context, look at the behaviour of the European Exchange Rate Mechanism (EMS). When the EMS was constrained in the early 90s, with its currencies only being allowed to fluctuate within certain bands, the drift exponent of the market for these was around 0.5. But when the bands were broadened in 1993, the drift exponent changed - to 0.58. More amazingly, when O&A did a similar study of the interest rate markets (which are not unified like foreign exchange markets but divide up into Western and Eastern regions) they found something very similar. They had found their first law of the markets. Dacorogna calls it the "Capitalism Constant", and hypothesises that any free market situation will have a similar drift exponent. Perhaps unsurprisingly, when O&A first tried to get the paper published, no journal would take it because it contradicted the EMH.

I was beginning to get excited. I had come to Zürich expecting a bunch of dry mathematicians with some cobbled together ideas and a big computer. What I've got is a group of physicists who are as philosophical switched on as anybody I've ever met. Sure, they've got a big computer, but it's not computer wizardry which was at stake here. These people already had the concepts and the insight - the computer was just the tool with which they were proving they were right. And what made the project so beautiful was that it was grown from the bottom-up. Unlike just about every other econometrician, Olsen had not arrived on the scene with a set of pre-determined mathematical axioms which he then set out to prove. Rather, he had had a hunch about the way that the markets organised themselves, and had gathered the necessary data and people in order to find out if his hunch was right. It's astonishing when you think about it, but this approach has never been tried in economics before. It's hard to swallow when you think of how much is at stake. The economic policy of every company, of every government, is based upon a set of theories that were largely dreamt up in the nineteenth century and which while they have been extensively tested mathematically have never been made to fit with the "real world". And then along comes Richard Olsen, who is so determined to develop a radically empirical set of economic tools that he won't even leave time out of the system. For Olsen there is no objective measure - there is just what there is.

Let's talk volatility. To help us we have Dr. Ulrich Müller, known affectionately as Ully, who comes into the meeting room and throws open the windows, filling the room with the breezes coming off the lake. "I'm a fresh air man," he explains. "I hope you don't mind." Suddenly I'm rather cold, despite the strong sun, and it's annoyingly appropriate - volatility is a measure of how quickly market conditions can change. Contained within this measure is an indicator of an investor's risk: the more dramatically things can change, the riskier it is to be involved in them.

Ully explains to me that traditional economic theory holds that markets are homogeneous - or devoid of internal differences and distinctions - and efficient and therefore adapt quickly to new information which supposedly travels through them instantly. On the other hand the data that O&A were collecting showed that price movements display patterns that indicated that the markets were not homogeneous at all. In addition, they found that periods of high and low volatility persisted for much longer than should be the case if the markets followed a random walk. Indeed, the data showed that there was a direct connection between levels of volatility and levels of market activity. So sensitive was this connection that it showed up the fact that on the Tokyo Stock Exchange the brokers take a long lunch break, on London Stock Exchange they take a short lunch break, and on Wall Street they don't break for lunch at all. The data was reflecting the distribution of people, computers and trading activity on the planet. And the more people there were, the more trading activity there was, and the more prices there were. O&A were discovering that the way prices behaved was intimitely bound up with the kind of people who were operating in the market - and that there were many kinds of people, not simply the stereotypical "rational consumer" that classical economics had always posited. In a way it's obvious - it's just that traditional economics had never accepted it as a possibility. The challenge now was to build a model of it all.

The fact that volatility comes in clusters had been observed and modelled before. That model was called ARCH, which stands for Auto-regressive Conditional Heteroscedasticity. In words, "heteroscedasticity" means that there are coherent patterns to be found, "conditional" means that they only occur under certain conditions, and "auto-regressive" means that the patterns exhibit some kind of memory over time - i.e. that they relate to previous pattterns. In other words, ARCH said that prices don't follow a random walk. But ARCH is an observation of a phenomenon, not an explanation. It suggests that markets are inefficient, but it doesn't give you any reasons for this. It tells you what. Ully thought he could tell you why - by tacking an "H" on to the front and making ARCH into HARCH. The "H" stands for heterogeneous.

The O&A team had realised that the key thing about the markets was that they are not homogeneous at all, but are made up of all sorts of different types of people. Fine, but what differentiates these groups from one another? Answer: the time-scale of their trading horizon. That is, how far ahead they look when they make a trade. Those with the shortest timescale are the floor traders, who have a certain amount of trading cash per day and who make money by projecting trends that last maybe only a few seconds and certainly no more than two or three hours. The floor traders don't care what the price levels are; all they care about is which way they are moving. At the other extreme are the central banks who are not interested in such minor fluctuations but only in big peaks and troughs and long-term changes. In between them are derivative market traders (or "hedgers"), the head dealers (the bosses of the floor traders) and the corporates (like chocolate manufacturers buying up cocoa stocks.)

Now the importance of fractals and intrinsic time begin to fall into place. The idea behind HARCH was that certain types of volatility cluster belong to certain types of trader time horizon, and this made the key mechanism of the market the dynamics or transmission mechanisms of information between traders. In other words, O&A were trying to understand the markets as a problem of information flow between people with different perspectives on reality, perspectives which were defined by their time horizons. And these time horizons had a fractal relationship with one another.

The EMH had been totally rejected by O&A. They were now arguing that not only did the markets not follow a random walk, not only did they exhibit coherent patterns over time, but there was no such thing as the perfect rational trader. Traders fell into broad, overlapping categories, and trading decisions that made sense for one type of trader made no sense for another. There wasn't even a single volatility measure for the market as a whole. The market was no longer to be thought of as a single, coherent, efficient machine. It was a mass of nebulous and interacting mechanisms whose parts did not add up to a definable whole.

My favourite metaphor was the one of the mountains. Olsen prefers to use the lake to illustrate his ideas, and once again he gestures out to the Zürichsee. He explains that any unpredicted event which effects the markets, the outbreak of war, say, or the collapse of a bank like Barings (things that economists call "fundamentals"), is the equivalent of dropping a stone into the lake. At O&A they're not interested in trying to predict the stone; what they think they can do is predict the pattern of the waves - those same waves which introduce risk into liquidity. He offers another analogy: "You always have this components approach, like in the atomic model, where you have electron orbits. These orbits are embedded; each orbit is embedded in the next orbit out, which is embedded in the next orbit out and so on. When an electron gets into an excited state it skips up into the next orbit. And a very similar thing happens in the financial market. Price changes have certain orbits - certain orders of magnitude of size - which enclose the market like rings. At the centre, you have small explosions, say a 5% basis point price move or one of the big traders like Soros selling. If the price gets excited enough by this activity, it might move up to the next orbit, to a larger order of magnitude." Olsen was saying that the orbits corresponded to the various time horizons. A small "explosion" might cause a certain amount of excitement amongst the floor traders, but not be significant enough to incite their bosses to trade. But if the traders get really hot under the collar their bosses get caught up too, and this next level activity feeds back into the market, generating more price movement and more trading activity. Again, things may stop there, or they may move up to the next level. It's a very human way of looking at the financial markets.

Unlike most financial prediction tools, the O&A Information Service (OIS) distinguishes between forecasting and trading. There are four tools: directional forecasting, which gives you a view of the short-term future of the market; historical analysis, which identifies "overbought" or "oversold" positions, market extremes that indicated optimal times for taking a profit from a position or opening a new position; the trading models themselves, fully computerised systems which do all your financial interpretation for you, right down to telling you "now go short, now go long"; and the potted version of all these which is available over the Internet (www.olsen.ch). If you want the full monty you pay O&A around 100,000 francs and they install a couple of Sun workstations into your office, which are linked by modem to their computers in Zürich, and give you a fairly comprehensive training course and full customer support. If you want the Internet version, you set up an account and then pay a very reasonable 40 Swiss cents every time you download a market "snapshot". The services currently cover 84 currencies and operate continuously all year round, and are all based on continuous monitoring and analysis of real-time foreign exchange data.

Although it doesn't yet turn a profit, the Internet service is key to the whole process - and to the vision. Olsen believes that the combination of widespread distribution of his forecasting information via the Web and the possibility of actually trading on the net - which will lower deal transaction costs and move activity away from the centralised stock exchanges - will in the end lead to more efficient markets with narrower stock spreads, which would be better for everyone.

But OIS is no get rich quick scheme. According to Felix Fluck, Head of Marketing at O&A, the system is like a radar, something to help you "fly in fog." The idea is to give traders a long-term edge over the market, to help them lengthen their perspectives and spread their risk. For Fluck, the O&A building holds within its walls the promise of "a future of stable worldwide financial trading. Sure, they'll be the occasional hurricane, but on the whole you'll be able to grow your crops." The image that all the members of O&A come back to again and again is that the financial markets behave like the weather, and using their software is like getting a weather forecast. Two or three months ahead, it gives you a pretty good idea of the season that we're in. A week ahead we can see storms gathering off the coast, although we don't know yet whether they're going to break or not. Tomorrow it's going to rain, that's 90% sure. But this afternoon it's going to be warm enough for you to give that barbecue.

I was slightly disconcerted to discover that no one at O&A actually uses the software to trade themselves. According to the employees, things didn't start out that way - it's just that none of them actually had the time to spare to dabble. Eventually the practice became institutionalised - Olsen felt that the team might become unnecessarily compromised if they were trying to trade for themselves whilst simultaneously trying to give unbiased advice to clients. However, so that he couldn't be accused of not putting his money where his mouth was, in 1993 he set up a Benchmark fund. This fund - which now runs at $6 million with a leverage of $3 million, giving a trading amount of around $20 million - is executed by two banks solely according to advice given by an O&A trading model. The idea is that the fund will provide an independent barometer of the quality of the company's software. And it's not just for company money - other investors can add their investments to it as well (I thought this sounded like a great idea until I discovered that the minimum investment was $100,000). Unfortunately, the fund is not yet the jewel in O&A's crown that it might be - it was started in 1993, one of the worst years for investments on record. But it says something that whilst 19 out of 20 funds lost money that year, the Benchmark fund broke even, and has made money every year since, if not on a spectacular scale.

Olsen predicts that one day, all traders will use his software, or something like it. And the beauty of selling trading models that have the notion of a heterogeneous market at their very foundation is that they won't be rendered useless by widespread use, like simpler systems that can be defeated by their own popularity - once everyone starts to use them, they're no longer any good. Everybody who uses OIS will use it in a different way. As he puts it, "As long as not all men love the same woman, our system will work."

This is a big claim, and to get some expert opinion I email Doyne Farmer. Farmer was the bloke who, along with Norman Packard and various others, built computers into his shoes in order to try and beat the roulette wheels in Las Vegas. He now runs a markets prediction company of his own called, appropriately enough, the Predictions Company. I have already spoken to Olsen about Farmer; the two men know each other's work. One of the things that differentiates their approaches is that whilst Olsen is extremely - and most unusually - open about his techniques and findings, Farmer is very secretive. I asked him what he thought of Olsen's claim that the OIS could become widespread without changing the markets themselves. "I agree with Olsen that as time goes on more and more people will use machine-based forecasting tools," Farmer told me. "However, I disagree in that I believe that this will have a strong effect on the markets. Markets operate by supply and demand: buying pushes the price up, and selling pushes it down. Everyone, even Olsen, agrees with this. From this it immediately follows that the market dynamics are altered when the pattern of trading is altered. My guess is that Olsen's response is that the alterations are sufficiently complex and diverse that they will tend to cancel."

This is the gamble that Olsen is taking - it is not so much to do with the science as with the vision. In the beginning, he followed his hunches that the market exhibited general laws, and he seems to be proving himself right. Now he is gambling that because his system has diversity built in at its very foundation he can go public without destroying the very patterns he has discovered. "The main thing that distinguishes Olsen's system from others who sell forecasts is that the underlying science in his case is good," says Farmer. "Generally, only people who don't know what they are doing make their forecasts public. I believe that broadcasting information will ultimately be bad for him: as more and more people bet money on his forecasts, the patterns that he has detected in the historical record will disappear, and his system will cease to work. In contrast, since we [at the Predictions Company] keep our forecasts and methods secret, we have control over how much is invested based on them and we can avoid pushing the market so hard that the patterns vanish."

I can't resist asking Farmer for an answer to that key Wired question: Richard Olsen, lunatic or prophet? Farmer obliges with a great reply: "Lunatic and prophet are not mutually exclusive categories; prophets are frequently both. Furthermore there is a large space in between for 'ordinary guys'. To me, Olsen just seems like a guy with a good idea who is trying to run a business. But then maybe I am a lunatic too, and just incapable of recognising a fellow traveller."

If it's impossible for Doyne Farmer to judge Richard Olsen, then it's certainly impossible for me to do it. I'm not a trader, let alone a mathematician, and besides, it is simply too early to tell whether or not the OIS is going to prove to be a robust prediction mechanism. In the end it boils down to whether the figure in the bottom right hand corner of the spreadsheet is positive or negative. But everything that I've seen in Zürich tells me that if it's at all possible to get an edge over the financial markets, then O&A is the place it's going to happen. If nothing else, deciding to try and model the markets from the bottom up, with minimal conceptual baggage and with prices as the only raw material, is an approach at once so obvious and yet so revolutionary that it is almost bound to yield results. What they are trying to do at O&A is build a map of the markets, to create some kind of geography through which traders can navigate. It's as pointless to think that the markets could ever be completely transparent as it is to think that we could ever control the weather - both systems are so complicated as to render such ideas meaningless. For the weather and the market - both dynamic combinations of liquid and solid, water and land, mountains and waves - what we need is some idea of seasons and some idea of the likely range of possible behaviours of the system in the coming days and weeks, in order that we might best be able to gauge when to put to sea. As far as the weather's concerned we're doing pretty well. If Richard Olsen is right, then it won't be too long before we're doing pretty well with the markets too.

- Wired UK #2.07, July 1996