Here is a sentence I did not expect to write in a financial planning blog. There is a British start-up whose stated aim is to predict the future, and it appears to be rather good at it.
The company is called Cassi. It was co-founded by Dr Keith Dear, a former RAF intelligence officer who did three tours of Afghanistan and later advised a Prime Minister, and Al Brown, a former Royal Engineer and physicist who now runs the technology. Their product is an AI system that forecasts how uncertain events will turn out, and attaches a probability to each.
For years, the gold standard in this strange discipline has been human. The psychologist Philip Tetlock famously identified a group of ordinary people he called “superforecasters”, who, through temperament and practice, could predict world events markedly better than chance and better than most experts. They are the benchmark everyone else is measured against.
This summer, on a public leaderboard that scores exactly this kind of forecasting, Cassi’s system ranked at the very top, statistically indistinguishable from those human superforecasters. On questions drawn from real betting markets, it became the first system to beat the superforecasters’ median. The firm’s own marketing claims it forecasts fifty per cent more accurately than people.
How does a machine get good at telling the future? Not through magic, and not through anything resembling a crystal ball. Through information. An AI can ingest and cross-reference an almost unlimited quantity of data, and the more relevant information you can hold in view at once, the better you can spot the patterns that hint at what happens next. Prediction, at least the near-term, this-is-more-likely-than-that kind, turns out to be substantially a problem of information volume. And information volume is precisely what machines are built for.
Which is interesting on its own terms, and significant for anyone who owns investments.
Think about what a share price actually is. It’s a collective, real-time forecast: the market’s best guess, right now, about the future prospects of a company, expressed as a number. Every buyer and seller is, in effect, forecasting. The price is the average of those predictions.
Now drop tools like Cassi into that process, connected to the markets, digesting every filing, every data release, every scrap of relevant news the moment it lands. The forecasting gets faster and sharper. The market’s collective guess gets closer to correct, quicker. In the jargon, price discovery improves: prices come to reflect what’s knowable almost as soon as it’s knowable.
That sounds abstract but it has a very concrete consequence.
The entire business of active fund management rests on a single proposition: that a clever, well-resourced manager can spot prices that are wrong, buy the bargains, avoid the duds, and beat the market after fees. It’s a hard living even now. Most active managers don’t beat the market over the long run, and the ones who do are fiendishly difficult to identify in advance.
Make the market’s forecasting sharper and faster, and you make that job harder still. The mispricings the active manager hunts for get rarer, smaller, and shorter-lived, hoovered up by machines before a human has finished reading the headline. You are asking a person to out-predict a system that can hold more information in its head than any human ever could, in a race that’s already difficult to win.
So where does that leave an ordinary investor? Oddly, in a rather comfortable place.
If beating the market is getting harder, the sensible response is to stop trying to beat it and simply own it: a broad slice of the world’s companies, held at the lowest cost you can find, left alone to compound. You’re not trying to be smarter than the machines. You’re declining to play them at their own game, and pocketing the difference in fees.
There’s a pleasing irony in it. The more powerful the forecasting technology becomes, the stronger the case for the least clever-looking strategy there is. Own everything, cheaply, and be patient. The robots can do the predicting. You just have to do the owning.
We help people in Hook and across Hampshire, Surrey, Berkshire, Sussex and Kent invest in a way that doesn’t depend on out-guessing anyone, machine or human.
Altor Wealth Management LLP is authorised and regulated by the Financial Conduct Authority (FCA number 769033). Registered office: Landmark House, Station Road, Hook, Hampshire RG27 9HA. This article is general information, not personal advice. The value of investments can fall as well as rise, and past performance is not a guide to the future.