I set out to use Ai to produce an automated trading strategy that could outperform the market. GPT4 and I collaborated over a couple evenings to create a Pine Script strategy for TradingView that would give Buy and Sell signals. I explored a variety of methods but settled on a strategy that utilized Heikin Ashi candle stick patterns, Bollinger Bands, and MACD. It backtested as magic.
Note: this project has now been extended and brought live as a subscription service in the form of algomountain.com.
Every trader understands the importance of limiting your own bias and emotion when it comes to successfully trading market movements. Create a plan, trade the plan, and feel as little as you can about it. The problem is this is easy to propose and difficult to do.
I set out to code a strategy that synthesized my personal learnings from trading with proven investment strategies and indicators. At the time that I was coding this GPT4 could not search the net. It only had knowledge about Pine Script V4 and Trading View was running V5. It had of course access to the internets knowledge about trading prior to Sept 2021 but was super hesitant to provide any actual strategies or recommendations.
I was able to cleverly prompt the system, provide it with updated code translation information for V4 to V5, and in the end code an automated trading script that profoundly beat the market in backtesting. Trusting backtesting is risky due to the difficulty of proving that you'd actually get orders filled at the time and the price that backtesting shows. Still, the strategy was successful on both Crypto and Stocks (when optimized) and was able to be adapted to ones chosen level of risk.
Manually traded the signals of the strategy proved successful and I'm currently in the process of implementing it as an algo-trading strategy.
Strategy Performance on $SOL
