ChatGPT Got 25 of 26 Real Trade Setups Wrong, Reddit User Finds
kaljakin · reddit · 2026-10-11
A Reddit user tested ChatGPT (GPT-6 in chat, high mode) on 26 real swing-trade setups he had personally traded successfully in the past, using daily charts of Intel and UDR. It got only 1 right — far worse than random chance.
- Method: show a daily chart, ask "after a few days you see this situation — will you go long/short or wait?"
- The author stresses the model's chart reading was accurate; the failure was interpretation. It treated indicators as absolute signals and ignored context: calling RSI above 60 not overbought (in a downtrend it is), and treating EMA 90 as resistance in a choppy, trendless market.
- After receiving feedback, the model lost confidence and waited through roughly 10 profitable opportunities, endlessly updating analysis and adding new entry conditions; once it regained confidence it bought the top and got burned again.
- It also ignored volatility, over-interpreting noise, and read volume from only a few days rather than the whole chart.
Conclusion: the model shows clear beginner biases and is consistently, predictably fooled by the market.
More from Models
- OpenAI's Decisions API hits GA: near real-time routing up to 10x faster than GPT-6 Luna — OpenAIDevs · 2026-10-23
- DeepLearning.AI weekly: GPT-6.1 hits 97.9% byte parity on COBOL migration — DeepLearningAI · 2026-10-23
- Marker's lite mode scores 75.1 vs 42.3 for Azure Read on OmniParseBench — VikParuchuri · 2026-10-23
- Datalab launches lite OCR at $0.50 per 1,000 pages, 20-30x cheaper than Azure, Google and AWS layout APIs — VikParuchuri · 2026-10-23
- LiveNerf's 29-day data finds no nerfing of Claude Opus 5.5 — TheOnlyVibemaster · 2026-10-23
- PHP Foundation reveals Anthropic's Claude cyber model scanned PHP, validated with OpenAI's Daybreak Red — saibharadwaj · 2026-10-23