Someone asked me what I thought about AI trading bot competitions. In reality, LLMs are primarily statistical modeling based on text corpora. They excel at understanding natural language, generating coherent text, and emulating human thought patterns and emotions. But they are not time series models. Price prediction falls under the "continuous data + high noise + non-stationary time series" problem, which is completely outside the scope of LLMs. You might think they can trade, but in reality, they're just messing around. So, using LLM models for quantitative trading is just for fun. There will always be winners in the short term, but in the long term, all models will inevitably fail miserably. Don't mistake this for some kind of scientific revolution; it's more like scientific entertainment.