An introduction to our investment philosophy
Some of the most pressing problems of the world (and life) would be better understood by recognizing the patterns which produce them. To separate signal from noise, wheat from chaff.
Indeed, life itself is a “dance of pattern” (hat tip, Alan Watts). Right down to the level of DNA in cells and up to the level of cosmic supernova there are structures – patterns – which underpin reality.
Survival itself is predicated on success in rapidly recognizing meaningful patterns. Snake! Our peripheral vision and autonomic nervous system (“fight-or-flight”) is wired to respond to threats. Wild berries! Our acute frontal vision is wired to locate camouflaged prey or gather nuts and fruits.
To get the pattern wrong – or to miss it entirely - can be fatal. Don’t eat the red berries!
G2’s aim is to use machine learning to find patterns in the universe of data. Our work combines insights into the brain (neuroscience), markets and AI/data science.
The human brain has evolved over eons to become an incredible pattern-recognition technology. Our ability to link the past, present and future (through patterns) is what separates us from our animal ancestors. Apes are pretty smart and even adaptable, but they can’t play chess or drive a car.
But the thinking brain is not infallible. Indeed it cannot be – the biological load would be too heavy. We would not have survived a “least time, least energy” revolution. In most of life the various heuristics and biases we have are not fatal. Things like loss aversion (prospect theory), anchoring, recency bias and over-optimism all serve us well … most of the time.
Hence the brain’s powerful pattern recognition function hits its limits at complex, noisy, data-rich problems (like investing).
Indeed, the stock market is a wicked learning environment. How or when do you know you are right? How can you brain discern patterns when feedback is delayed, clouded by randomness and obscured by behavioural biases. To paraphrase Churchill: the market ist a riddle, wrapped in a mystery, inside an enigma.
While markets are somewhat efficient, Ben Graham’s allegory of Mr Market remains the most useful frame of reference. That is, Mr Market is a manic-depressive who is driven in turns by panic, euphoria and apathy. Graham stresses that Mr Market is there to serve you not guide you.
The mathematics of efficient market theory are elegant and compelling … and wrong. Their fatal flaw is the assumption that errors (among market participants) are independent. Really!?
Our view is that markets are mostly efficient, most of the time. Indeed really good investment opportunities are ephemeral and rare. A good investment process needs to combine counter-intuitive qualities. Example: patience and aggressive opportunism. Another example: discipline and adaptability. Another example: neither take comfort nor discomfort in being “with the crowd”.
One thing which strikes us, apart from the whoosh as the years fly by, is how revolutionary recent advances in computing, data science and machine learning are. Consider the increasing number of ‘high order’ pattern recognition tasks where computers dramatically outperform humans: chess, poker, quiz shows, cancer diagnosis, driving, game of GO, et cetera.
Things are indeed moving fast. This is because the AI revolution is bottom-up, not top-down. It is led by doers/practitioners, not theoreticians/academics.
Even though we live and breathe this (exponential) world, there have been times when we failed to foresee the pace of change. Yet in one way or another, these developments have all been ideal for G2, as we apply our adaptive strategy and the latest in AI/machine learning to stock picking.
A personal anecdote to give a sense of history: when one of the G2 partners was working on his PhD (just) ten years ago, experts were still training computer systems. Now, computer systems are beginning to train the experts.
Nevertheless we are far from being techno‐utopians; our mindset is much more engineering than evangelical. In fact, these years in the business of systematizing stock picking has taught us how far away we are from easy solutions. Like the curmudgeonly old‐timer, we find ourselves increasingly skeptical of expectations and claims made about AI.
The bottom line is that, just like humans, machines can make mistakes … it’s just that they tend to make fewer mistakes (just ask grand chess master Kasparov who lost to IBM’s Deep Blue in 1998). To say another way, machines are just a little less dumb than humans.
MAN VERSUS MACHINE
Whether ML can beat human managers in stock picking is no longer a theoretical argument. G2 has demonstrated, over 6 years (and 1000’s of trades) that the machine outperforms >90% of 1000’s of comparable human managers, with less volatility/draw-downs. The jury is no longer out.
A key advantage the machine has in competing against humans is that it can assess the entire stock universe deeply and rapidly. It then becomes a “rejection machine”, saying no to more than 96% of stocks rather than analyzing a few and getting to yes. “The difference between successful people and really successful people is that really successful people say no to almost everything.”
To say this another way, the system specifically focusses on the “tails” of the stock distribution (the outlier stocks which are priced inefficiently). As Mandelbrot said, Everything that matters is in the tails. True, that.
One of the criticisms is that pattern recognition is necessarily backward looking. That’s fair, but we have demonstrated success over 6 years and have done particularly well in periods of unpredictable market stress: January 2016 (Oil/China), Brexit, Jan 2018, Oct 2018 (Momentum crash). It’s not that these events are predictable – it’s that markets/people tend to react in predictable ways – panic (in simple terms) – and that creates opportunity.