Testing SOTA LLMs in Trading for 2 Years: They Lose to Static Baselines
hyperparticle · x · 2026-07-30
An empirical test on LLM trading capabilities reveals that current state-of-the-art models perform poorly, significantly lagging behind simple static baseline strategies.
- Reasoning doesn't help: Higher reasoning capabilities in models do not translate into better trading performance.
- Behavioral flaw: When losing money, models (like Sol) tend to trade less frequently rather than adjusting strategies to recover losses.
More from Research
- UMich team gets $750K NSF grant for transparent AI learning tools — QVeraLiao · 2026-07-30
- AI Can De-anonymize Internet Users for Under $4 with 90% Accuracy — luisdans · 2026-07-30
- GRAM Paper: Precisely Removing LLM Dangerous Capabilities via Gradient-Routed Modules — burkov · 2026-07-30
- Yale Scholar Shares Practical Guide on Using AI in Household Finance Research — TaniaBabina · 2026-07-30
- Experiment: Injecting Fictional Lore as Executable System Prompt into RAG Crawlers — BitcoinsOrganizer · 2026-07-30
- llamppl: Integrating Large Language Models into Probabilistic Programming — xuanalogue · 2026-07-30