alt Hacker NewsI run a small experiment on this: 29 paper-trading accounts on real US stock prices, $100k each, since July 27. Four AI models (ChatGPT, Claude, Grok, Gemini) each write a trading rulebook and rewrite it every day from their own results. One account is a fixed rulebook that no AI ever touches, as a control.
Result so far (paper, 34 trading days): the no-AI control is +13.0%, the S&P 500 is +3.4%, and 25 of the 28 AI accounts are below the control. The best single account is a Grok-written "patience" book at +40%, which I treat as one lucky account in a choppy market, not a finding. At the trade level the AIs and the control look the same: 3,212 closed positions, median +0.06%, median hold about 2 hours. They trade a lot and mostly go nowhere.
Everything is public, including the losses and the retired strategies: https://aitradingcompetition.com/which-ai-is-winning.html and the full trade file as CSV at https://github.com/ckamelhar-collab/ai-trading-arena-data. Paper money only, not advice, nothing for sale on those pages.
Replies
Why didn't you let the bots backtest and forward test to pick the mix of best strategies/indicators?
Don’t let AI write for you. Everyone can tell.
I'm curious what signals they're trading on. SEC filings? Candlestick voodoo numerology?
None of this really means anything. For one if you test enough strategies one by luck will be killer. Second, past performance != future.
One AI might consist of some convulated strategy of who knows what, but all it takes in the end is for it to say...hold more tech stocks than not, or buying more call options or leveraged positions in a bull market, being more out of the market during a bearish market, etc. But that same strategy will fail miserably as soon as the market conditions change