Top Traders Unplugged Sep 25, 2026 1h 6m · 50m saved
With Nick Baltas, Managing Director and Head of R&D for Cross-Asset Delta-One and Commodity Systematic Trading Strategies at Goldman Sachs
Nick Baltas said almost 45% of S&P 500 stocks now have a negative beta to the index, meaning they tend to fall when it rises. At the peak of the dot-com era, the share was just under 20%.
Most of the debate about AI in investing is about whether a model can pick trades. The paper Baltas brought to this interview asks whether a team of AI agents can run an institution's whole asset-allocation process, and his main objection is that no one can reproduce its results.
"So about half of the stocks in S&P 500 have a negative beta to the index which I think is just mind blowing."
Nick Baltas, who runs research and development for systematic trading strategies at Goldman Sachs and lectured in finance at Imperial College Business School before that, has been a regular guest on Top Traders Unplugged's Systematic Investor series since 2022. In this episode he reviewed two research papers on how portfolios get built.
The full interview is covered here so you can skip it. 66 minutes of audio, 16 minutes of reading.
Here are the 13 insights that matter.
Key Takeaways
Almost 45% of S&P 500 stocks have a negative beta to their own index, against just under 20% in the dot-com era
Baltas said AI stocks are forming an index within the index, and the rest are drifting away from it
Rates have doubled their contribution to trend-following returns this month, and he called it the third-best year in 15
Switching a trend model's speed on the VIX added no significant value over a fixed 10-month signal
Only three months in nine years fell in the high-volatility zone
He would rather scale signals by volatility than switch models on a trigger
An AI "CIO" overruled its own agents' vote in the paper's March 2026 run, and ended at 44% equities against a 60/40 benchmark
AI-built portfolios cannot be properly backtested, because the models know the future and do not give the same answer twice
The binding constraint in asset management is human bandwidth, not data or models, according to the paper
1. A 45% Negative-Beta Reading
Asked what had caught his attention, Baltas pointed to a chart from Christian Mueller-Glissmann, who runs multi-asset strategy research at Goldman Sachs, featured in a Financial Times article that week. It shows the share of S&P 500 members with a negative beta to the index over a three-month window. Beta measures how much a stock moves with the market; a negative reading means it tends to move the other way.
Half the index now moves against the index
"So about half of the stocks in S&P 500 have a negative beta to the index which I think is just mind blowing." — Nick Baltas
The dot-com peak was less than half of today's level
"Last time that we had this situation was in .com, but at the time the number was just a tad less than 20%. It's now almost 45%." — Nick Baltas
He said the cause is that the S&P 500 is weighted by market value, so a small group of large companies sets its direction.
AI stocks have become their own index
"The AI dominance is effectively building its own kind of mini index within the index and all the rest are becoming less correlated to it." — Nick Baltas
The host asked whether energy stocks were part of it. Baltas said it could be, but that the Goldman Sachs note focused on concentration, on whether bonds still protect a portfolio, and on the opportunity the split creates for dispersion trades, which bet on stocks moving apart from each other.
