Fable 5.1, GPT-6 Astra and GrokBot compete in a one-week trading strategy contest
Each of the three models ran automated backtests for a week: Fable 5.1 produced 1297 winning strategies, while Astra built 1018 and barely hit usage limits, making it the most cost-effective.
I put Fable 5.1 vs GPT-6 Astra vs GrokBot in a trading competition for 1 week. They had 1 job: find profitable strategies. Here's how to run it yourself: Step 1: Get your models Claude desktop: http://claude.ai/download ChatGPT desktop: http://chatgpt.com/download GrokBot: http://x.ai/bot Step 2: Give it a backtesting engine None of these models can backtest TradingView strategies on their own. So set up a free account at http://tradingkit.com, click My Account, copy the line of code. Paste it in with "hey there buddy, could you install this locally for me". Copy your authorisation code from the same page, paste that in too. Step 3: Check the connection Prompt each model with: "Backtest a simple strategy on BTCUSDT 1 hour, just to make sure we're connected" Step 4: Get the Prompts Visit http://aitradingarena.com, click the key icon, all 3 prompts are in there free. Step 5: Build the dashboard Paste the dashboard prompt from AITradingArena. You'll then get every stat per strategy: drawdown, forward test, Monte Carlo for every trade. Step 6: Onboard the model Paste the onboarding prompt, it teaches the model how to backtest and record results. Step 7: Start the Sprint Paste the sprint prompt, walk away, come back in a week. My results: Fable 5.1 won with 1,297 strategies, 139 were good enough for forward testing. GrokBot built the most strategies at 1,828 but only 3 were profitable. GPT-6 built 1,018 and barely touched my usage limit, which makes it the best value of the 3. Real talk: I wouldn't trade any of them as they are. They all go into incubation for 3 months first. But Fable 5.1 was the winner of this test.