{"id":289018,"date":"2026-02-24T09:17:13","date_gmt":"2026-02-24T09:17:13","guid":{"rendered":"https:\/\/www.quoniam.com\/?p=289018"},"modified":"2026-05-21T09:28:32","modified_gmt":"2026-05-21T09:28:32","slug":"machine-learning-stock-portfolios","status":"publish","type":"post","link":"https:\/\/www.quoniam.com\/en\/interview\/machine-learning-stock-portfolios\/","title":{"rendered":"Machine learning in stock selection: Beyond linear thinking"},"content":{"rendered":"\n
\n

Carsten, why are traditional linear models no longer sufficient for equity investments today?<\/strong><\/p>\n\n\n\n

A classic linear model \u2013 as used in multi-factor approaches \u2013 assumes that effects are linear and additive. Put simply, if a factor is twice as strong, it also has twice the effect. And if several factors are positive, their effects simply add up.<\/p>\n\n\n\n

This assumption is robust and easy to interpret \u2013 which is precisely why it has been the basis of quantitative equity models for decades. The only problem is that reality is often not so straightforward.<\/p>\n<\/div><\/div>\n\n\n\n

\n\n\n
\n
\n
\n

Not everything on the stock market is straightforward. That’s exactly where we use machine learning.<\/p>\n\n <\/div>\n <\/div>\n

\n
\n \"\"\n <\/div>\n
\n

Carsten Rother,
\nCo-Head of Research Forecasts<\/p>\n\n <\/div>\n <\/div>\n<\/div>\n<\/div><\/div>\n\n\n\n

\n

Non-linearity essentially means that a relationship is not proportional or constant. The effect depends on the context or threshold values.<\/p>\n\n\n\n

Let’s take value as an example: a linear model assumes that the cheapest stock has the highest expected return \u2013 and everything in between is uniform. In practice, however, we know that some stocks are cheap because their business model has structural problems \u2013 “cheap for a reason”.<\/p>\n\n\n\n

This means that very low valuations can have completely different effects depending on the quality of the company. It is precisely this context dependency that is a non-linearity.<\/p>\n<\/div><\/div>\n\n\n\n

\n

Can you make that even more tangible?<\/strong><\/p>\n\n\n\n

A very clear example is what are known as thresholds.<\/p>\n\n\n\n

Think about debt. The difference between low and moderate debt is often hardly relevant. However, if a company exceeds a critical mark \u2013 a threshold \u2013 the risk increases disproportionately.<\/p>\n\n\n\n

A linear model would say that more debt is uniformly worse. In reality, it’s more like this: for a long time, little happens \u2013 and then it tips over.<\/p>\n\n\n\n

Such tipping points or interactions between factors are typically non-linear. Machine learning helps us to systematically identify these patterns \u2013 not only in individual cases, but across the entire investment universe.<\/p>\n<\/div><\/div>\n\n\n\n

\n

Are you replacing your classic multi-factor model with machine learning?<\/strong><\/p>\n\n\n\n

No. The linear model remains our foundation.<\/p>\n\n\n\n

Our approach is deliberately two-stage: first, we create a structured, economically sound linear forecast \u2013 specifically, we apply our established multi-factor model based on value, quality and sentiment. These factors are combined to provide a transparent, robust baseline estimate for each stock in the universe.<\/p>\n\n\n\n

We then systematically check whether there are any patterns that the model overestimates or underestimates. Technically speaking, we apply machine learning to the residuals of the linear model \u2013 i.e. to the unexplained part of the return (see Figure 1). Machine learning thus acts as a corrective, not a replacement. Depending on the interaction or threshold effect identified, the forecast can be adjusted upwards or downwards.<\/p>\n<\/div><\/div>\n\n\n\n

\n
Figure 1: Comparison of machine learning and linear model<\/h5>\n\n\n\n
\"\"
Source: Quoniam Asset Management GmbH<\/figcaption><\/figure>\n<\/div><\/div>\n\n\n\n
\n

In portfolio management, both components remain transparent: the contributions from value, quality and sentiment are visible separately, and the machine learning component appears as an independent building block. The overall forecast is the result of the controlled interaction of both elements.<\/p>\n\n\n\n

This approach has been an integral part of our forecasting systems since 2018 and is fully embedded in our research and production process.<\/p>\n<\/div><\/div>\n\n\n\n

\n

Greater model complexity sounds attractive at first. But when does it become a problem?<\/strong><\/p>\n\n\n\n

That’s a valid question. In financial markets, we work with a relatively low signal-to-noise ratio meaning that genuine return signals are often weak and difficult to distinguish from random market fluctuations. It’s easy to find patterns in historical data that exist purely by chance.<\/p>\n\n\n\n

That is the real risk: a model may look excellent in the past \u2013 but only because it has ‘learned’ coincidences that will not be repeated. This would be dangerous for investors because apparent predictive power would then disappear in the future.<\/p>\n\n\n\n

That is why we discipline our models very strictly:<\/p>\n\n\n\n