ResearchNovember 17, 2026
-7
❄️
LLM Traders: Only Memorization Beats the Market
A study finds LLM trading agents do not outperform simple strategies; their edge came from memorization.
#LLM#trading agents#benchmark#finance#evaluation

🔥 What happened
Jackson Wang tested five LLM trading agents from four vendors before 598 earnings announcements and after 409 bad-news drops. No agent beat simple rules like “just buy everything.” The only edge came from training data: a GPT model called 59.2% of directions correctly before its training cutoff, but only 53.3% after.
💡 Why it matters
If you build LLM agents for event trading, you’re likely measuring memorized knowledge, not skill. The benchmark is look-ahead-free, with all inputs cut at decision close — yet the apparent edge evaporates post-cutoff. That’s a wake-up call for any backtest that leans on LLMs.
⚡ Our take
Most “AI trading bots” are expensive random generators with good memory. Anyone who skips a blind post-cutoff test is selling investors an illusion.
The title, summary and analysis of this item were produced automatically by an AI system and have not been editorially reviewed. They may contain errors, bias or omissions — when in doubt, read the linked original source.
Deep Dives & Similar Intelligence

GPT-6.2 Sol: Pricier, Not Better
AI modelsNovember 18, 2026
Alibaba open-sources its code review agent
Tools & ProjectsNovember 13, 2026

165M Model Trained for $4,200
ResearchNovember 15, 2026

Microsoft VP ships GPU task manager
Tools & ProjectsNovember 12, 2026

Claude secretly learned protein language
ResearchNovember 17, 2026