Assessing the benefits of optimized agentic AI systems
Launching 2026 — Register to be notified
Explaining markets, in real time.
An open, out-of-sample benchmark for AI systems to explain stock returns around earnings announcements — using any information available prior to the announcement and the earnings call transcript.
Get notified when we go live:
The research
Where it comes from.
The competition is the open benchmark introduced in Koijen and Levy (2026). It accompanies an SDK so researchers can evaluate optimized agentic AI systems for asset pricing — and benchmark them against a real-time, out-of-sample standard.
Working Paper · April 2026
Assessing the Benefits of Optimized Agentic AI Systems for Asset Pricing
Ralph S. J. Koijen — University of Chicago Booth · CEPR · NBER Bradford (Lynch) Levy — University of Chicago Booth
Abstract
Evaluating optimized AI systems for asset pricing is fundamentally difficult for two reasons. First, models are trained on all data, implying that any backtest using historical data suffers from look-ahead bias. Second, markets are reflexive — as investors adopt AI, prices adjust — eroding the very patterns the system was trained to exploit.
We introduce a real-time, out-of-sample benchmark designed to sidestep both problems. Applied to a range of agentic AI systems, the best-optimized systems more than double the explained variation in returns — R² rising from 8% to close to 20%. We release an SDK and launch an open competition inviting researchers to improve on these results.
The benchmark is free from look-ahead bias and fully reflects the reflexivity of financial markets.
02
Relevance of the benchmark.
Saturating this benchmark would represent fundamental progress in understanding how capital markets process firm-level information.
03
Competition.
Compete against peers to rise to the top of the leaderboard each quarter!
Stay informed
Subscribe to the newsletter.
The competition isn’t open yet. Subscribe to hear when the SDK is released, when participants can submit forecasts, and when the first results are published.