Join NexChange - the professional
network for the financial services
industry - and receive a free one-
year subscription to Forbes
New Report Shows AI Hedge Funds Are Crushing Their Human Overlords
Hedge Funds, FinTech
If there is one constant in the world of quantitative investing and mechanical trading strategies it is change. In a report on artificial intelligence (AI) entering the hedge fund space and doing so with a bang, Eurekahedge noted the changing alternative investment landscape in a January report. The report noted AI hedge fund firms with machine learning are outperforming “traditional quants,” benchmarked by the CTA/Managed Futures index, and doing so with low correlation. These learning and self-adjusting hedge funds are also beating traditional human-based fund managers.
AI Hedge funds beat other alternatives during the quantitative era (2010 and beyond)
Eurekahedge’s AI/Machine Learning Hedge Fund Index, which tracks the historic performance of 23 hedge funds, has outperformed both traditional quant and more generalized hedge funds since 2010. The AI hedge funds provided investors annual returns of 8.44% over this period. This compares to the Eurekahedge CTA / Managed Futures index, which delivered 2.62% over a similar period, the Eurekahedge trend following index, up just 1.62%, and the Eurekahedge hedge fund index, up 4.27% since 2010. In 2016, the AI hedge funds are up 5.01%, closely matching traditional hedge funds 4.48% performance.
Why and how did these intelligently aware machines outperform their rivals?
They “why” question can be answered by understanding two underperformance issues involving traditional quant strategies.
With the popularity of quant hedge fund investing strategies, illustrated in the report by noting record $40.5 billion in inflows over the past two years, the highest level in the history of the strategy. In the first two quarters of 2016, inflows totaled $10.8 billion, again a record high level.
With this popularity has come traditional hedge fund problems. The widespread diffusion of similar quant strategies who control the lion’s share of assets under management has led to a crowding effect, depressing returns.
While too much money chasing the same market indicators is a performance dampener, so, too, is the historical research model. “Trading models built using back-tests on historical data have often failed to deliver good returns in real time (as previously identified trends have broken down).” While the Eurkahedge report didn’t dive deep into this concept, many algorithmic fund managers and market analysts have noted that during the central bank quantitative easing period historic correlations and market relationships were less meaningful.
Given there is a performance dampening in traditional quant strategies, how do AI hedge fund vehicles work to deliver superior returns?
AI hedge fund industry leader Yoshinori Nomura reveals his secret sauce to a degree
Perhaps one of the industry leaders of intelligent AI hedge fund design is Yoshinori Nomura, a director at the JPY507 billion Simplex Asset Management.
Nomura has a Master of Science in physics and majored in Nonlinear Non-equilibrium Statistical Mechanics – all of which served as the inspiration behind his investment strategy. But taking apart the strategy into understandable components is what makes him unique along with his mathematical mind.
Among his three driving factors to algorithmic trading success is “the model should be as simple as possible.” Simplicity is in the eye of the beholder at times, but here one of the best performing AI hedge fund managers shows his secret sauce to a degree.
Among several keys to success is a system that monitors the market environment and adapts strategy to best fit current conditions. Certain analysts have developed methods to monitor and even handicap probabilities relative to a change in market environment.
Currently Nomura has two primary factors it attributed to success, both based on price history. He considers the market environment for price persistence, used in a trend following strategy, and the market environment for mean reversion. Traditional quant strategies often separate these concepts and do not alter their strategy accordingly. He notes the issue:
Generally speaking, CTAs tend to use trend following strategies which works well when there is strong momentum in the market but does not work well in mean reverting environment. Counter trend following strategies are just the opposite. Our strategy looks at momentum and mean reversion at the same time so that it can adapt to both market conditions.
Another key to success is his back testing model. Historical back tests that blindly cull 20 years of performance history are not as relevant relative to existing markets, where nearly from 60% to even 80% of stock market trading is now estimated to be done through algorithmic methods.
To combat the problem of historical performance not being as meaningful to current performance, Nomura uses what he describes as Walk Forward Test (WFT). The process, as he explains, looks for recency relevance to weight analysis:
‘Walk Forward Test (WFT)’ is a verification method of sustainable predictive capability of a quant strategy as an alternative to back test. New quant strategies will always suffer from a ‘chicken and egg’ problem, and have a tough time convincing investors that a good looking back tested result will not fail in real time and suffer from the classic ‘betrayal of the back test’ phenomenon. WFT avoids this ‘betrayal of the back test’ and directly evaluates the sustainability of the predictive capability of the new quant model. A good looking back test result might be dubious because the quant analyst knows the result of market at first place and can choose the best combination of trading strategies in the back test period of time to make it look the best, however this ‘optimisation’ does not usually lend predictive capability to the model. On the other hand, the result of WFT is based only on prediction by the model which does not use any future information in prediction. WFT uses historical time series (ex. 20years) and the model firstly looks back certain period of time (ex. 1 year look back from 19 years ago) to learn and adjust to the market condition of its present (ex. 19 years ago). Then, the model predicts the future and trades for a short period of time (ex. a week). After the profit/loss of the trade is fixed, it looks back the same period of time again to refresh the model’s predictive capability and repeats this process until the end of the historical data. If the model has no predictive capability, the result of WFT is miserable, but if the result shows stable returns over time, it is very difficult to say the model does not have any predictive capability since all the result are implemented only by the prediction. At least, if a human trader with a good track record can be considered to have sustainable predictive capability for some reason, an AI model with a good WFT result can be considered in the same way.
Something must be working based on performance, but there are caveats. While the AI hedge funds have been outperforming both traditional quants and human hedge fund managers, they have a difficult time understanding causation in some cases. For instance, why was the Brexit market response different in some respects from the Trump election? These issues where events without historical precedent occur present problems, the report noted.
This article was originally published in ValueWalk.
Photo: chupacabra runner