Stock selection by an independent model and the use of a pure artificial intelligence quantitative model have resulted in a significant outperformance for one of the latest additions to the Pictet AM family of funds. This success has generated a rare enthusiasm among investors. The fund has raised over €1 billion in assets in 14 months..
When it comes to finding the shortest, fastest or most scenic route on a road trip, GPS already uses artificial intelligence. Unfortunately, there is no GPS as such on the stock market to find the best performance or growth stock. This is because financial data is not static (like a road) and its impact can also vary according to the economic climate. Can creating an effective stock selection model using artificial intelligence work?
At Pictet AM, we have been experimenting for a number of years and the initial results have been convincing. Since its launch in March 2024, the Quest AI-Driven Global Equities strategy has used this technique.
In fact, the QUEST team (Quantitative Equities & Solutions Team) has developed its own model based on artificial intelligence, with the aim of achieving the most optimal selection of equities, to systematically and progressively generate outperformance that can have a profound long-term impact. A performance differential of 1.5% per annum after fees on an international equity index (MSCI World) represents over 15% additional returns in ten years ,” comments David Wright, co-manager of QUEST and Head of Quantitative Investments at Pictet AM.
But before such results could be achieved, an enormous amount of data, almost as many tests, patience and several years of work were required. ” For each stock and to try to isolate its potential future performance, we need a wide range of data, at a high frequency, over long periods that only artificial intelligence can process, ” explains David Wright.
Building, training and, above all, meticulously testing machine learning models based on a large arsenal of data, including fundamentals, analyst sentiment, prices and market activity, is a time-consuming task. The combination of this multitude of data, which only AI is capable of processing, has made it possible to analyse the stock markets and put them into a “box”, using these new capabilities to predict market trends as accurately as possible.
Finally, the components that determine the performance of each stock were put together to create an optimised portfolio comprising the stocks most likely to outperform.
This global equity portfolio also aims to outperform, regardless of the stage of the economic cycle and the markets.
But it’s not easy to model. This is because financial data is constantly changing, and even identical data can tip share prices in both directions. Logically, bad economic news has a negative impact on the market, but paradoxically it can also have a positive effect in other circumstances. For example, poor economic data can encourage rate cuts and a rise in equities. A prediction model must be able to handle such situations, and measure their impact according to the circumstances.
A model that evolves automatically over time
To achieve this, the model adapts and corrects itself. The use of the concept of machine learning has become essential. This technology aims to teach machines to learn from data and improve with experience, rather than being reprogrammed to take account of these changes. With this technique, the model has gained in efficiency and is capable of capturing the relationships between data over a wide historical horizon, with rigorous testing.
The time scale has a different influence
It has been observed and considered that each share is initially influenced by investor sentiment rather than fundamentals. However, the macroeconomic cycle is taking over again in the long term. Artificial intelligence, via machine learning, is capable of estimating this short-term impact using a multitude of data, even the smallest, each of which is weighted, such as market activity, calendar effects and investor positioning. For example, a stock surrounded by pessimism, with a high short position and/or multiple negative analyst opinions, will benefit from a more solid rebound thanks to the surprise effect in the event of good news (and therefore a better performance). The rebound will be more violent than if it had already been favoured by analysts.
Alongside this macroeconomic data, equities also have their own universe, which is often even more complex, making it difficult to build models,” admits David Wright.
Today’s model may not be tomorrow’s model
The time scales have also been unpacked. The characteristics that influence short-term and long-term action are not the same. “We believe that, over short-term horizons, investor behaviour rather than company fundamentals is the main driver of stock market returns
The black swan
However, despite extensive data collection, certain events remain unpredictable and absent from the model. This is known as the black swan… United Healthcare, the American stock, was a recent illustration of this. The company has faced a number of headwinds, including the murder of its manager, an obviously unforeseeable event. The share price fell immediately. When an invisible and unquantified risk appears, the value is automatically removed from the list of values.
Launched on 28 March 2024, the Quest AI-Driven Global Equities strategy got off to a good start, with a cumulative performance of 14.88% in the first year compared with 13.71% for the MSCI World EUR benchmark. At the end of May this year, on a rolling year basis, the fund showed a capital gain of 11.08% compared with 8.74% for its benchmark, an increase of 2.34%. It should be noted that the outperformance was maintained during the market downturn in the spring.
At the head of the portfolio, it is not surprising to find three champions of artificial intelligence with high prospects: Microsoft, Apple and Nvdia.
“Quest AI-Driven Global Equities’ rapid crossing of the billion-dollar mark is impressive, but it is not the first strategy at Pictet Asset Management to have done so. However, not many have,” said Bruno Hellemans, Head of Distribution Belgium-Luxembourg at Pictet AM.
Finally, a new strategy (long/short) using the same technique has been added to the range, offering absolute return and diversification with no correlation to the global stock market.
