Pssst, You! The original story is on Medium! Please clap 50 times and share the article with a friend interested in investing, finance, or AI.
|
Every time I open up my NexusTrade account, I keep expecting to see one thing — my ChatGPT-Generated Portfolio destroyed, down 30% for the week and 70% for the month. Yet, the reality is far from that. Not only are my ChatGPT-Generated Portfolios profitable, but they are significantly outperforming the market. By a lot.
|
The “One and Done” Optimized Portfolio is up 34%. since the start of the experiment
|
To be absolutely transparent, these results are not backtest results. They are results obtained from real-time paper-trading. This article will summarize the experiment and give the current, unaltered results as they stand. For the previous article in the series, check out the following article.
|
The purpose of this experiment was to show the average everyday investor what they are missing out on when using traditional brokerage platforms. Currently, there is no brokerage that comes close to matching NexusTrade in terms of speed, configurability and raw functionality. The AI-Powered Chat allows users to effortlessly perform financial research and generate fully automated trading strategies.
|
Talking to Aurora, the AI-Powered Financial Assistant
|
During this experiment, I decided to test several different ideas to see how each of them would perform in the market. This included:
- An “un-optimized” ChatGPT generated portfolio (booted from the experiment due to not making ANY trades)
- A ChatGPT-generated portfolio that was optimized at the beginning of the experiment
- Two ChatGPT-generated portfolios that are re-optimized periodically differently
- A portfolio with “Buy and Hold SPY” as the strategy.
|
Let’s discuss how these portfolios are doing.
|
Aside: Introducing Genetic Optimizations For All!
|
For months, I’ve been talking about the genetic optimization engine as a powerful tool to bring your trading game to the next level. And now, thanks to the recent surge in growth in NexusTrade, I’ve upgraded my servers, and can finally launch the genetic optimizer for a full-scale public release!
|
The genetic optimizer works by running thousands of backtests in the backend with slightly different variations of your portfolio. And then, it continuously makes modifications to your portfolio that maximizes some objective function.
|
The Optimization Config in NexusTrade
|
For example, if you optimize a portfolio using “sortino ratio” and “average drawdown” as your fitness functions, the portfolio will generate children portfolios that are better than your original portfolio in those ways! In other words, the children portfolios will have a higher sortino ratio (better risk-adjusted returns) and a lower drawdown (a low percent change from peak to trough).
|
Unlike other optimization techniques which will generate one portfolio, genetic optimizations specialize in generating POPULATIONS of unique, powerful solutions.
|
NexusTrade also includes built-in tools to tackle the overfitting problem, one of the most pervasive flaws with all optimization techniques. This includes:
- Automatically splitting your data into a training set and a test set
- Allowing users to generate new individuals “spontaneously”
- Enabling users to control the mutation rate, mutation intensity, and the initial positions that the backtests contain
- Allowing the training set to be split into multiple windows and taking an average of the results.
|
From my experiments, these configuration options SIGNIFICANTLY improve the optimization process and allow for the generation of unique candidate solutions.
|
Finally, NexusTrade is the only platform that’s doing anything like this. While you can set up custom indicators in the Robinhood app, you don’t have nearly as much configurability and tools at your disposal as you do with NexusTrade. Refer a friend and let them learn what a differentiated trading experience is like.
|
Buy and Hold Baseline Portfolio
|
I talk a lot about “buying and holding SPY”, and for good reason. It’s the simplest, most effective way to become a millionaire passively. 99% of trading strategies don’t come close to the effectiveness.
|
However, the AI-Generated Portfolios do. As you can see, this strategy had an outstanding 8.7% gain over the course of 2 months. For a purely stock-trading strategy, this is considered incredible! But it pales in comparison to every single one of the optimized AI-Generated Portfolios.
|
One and Done Optimized Portfolio
|
The One and Done Optimized Portfolio was optimized at the beginning of the experiment. Just like in the previous updates, it has the best performance by far, increasing 34% in two months. This is over 3 times better than the Buy and Hold SPY Trading Strategy!
|
The One and Done Optimization is DOMINATING the market
|
Sliding Window Optimized Portfolio
|
The Sliding Window Optimized Portfolio was initially optimized at the start of the experiment, and is re-optimized every 1-2 weeks. During each re-optimization, both the start date and the end date are both increased so that the model learns to make decisions on the most recent data. The goal is that by re-optimizing the portfolio in this way, the model will learn and adapt to an ever-changing environment.
|
The Sliding Window Optimized Portfolio is DOMINATING the Market
|
Expanding Window Optimized Portfolio
|
The Expanding Window Optimized Portfolio was also optimized at the start of the experiment, and is re-optimized every 1-2 weeks. The difference here is quite subtle. During each re-optimization, only the end date is incremented. This is so that the model learns to make decisions on all available data in the history.
|
The Expanding Window Optimized Portfolio is DOMINATING the Market
|
As the results above suggest, the optimized ChatGPT-portfolios are currently doing amazing. They are vastly outperforming expectations by a significant margin during real-time paper trading.
|
Don’t get me wrong; these results are undeniably outstanding. But while we’ve shown that the optimization process learns how to enter into a position, we haven’t shown evidence to be remarkably good at exiting a position. These past couple months were a pure bullish rampage, and the portfolios can go down as easily as they go up.
|
The model still hasn’t shown to make many exit decisions during live-trading. What this could indicate is that it learned a marginally better strategy than buy and hold that entered into positions more cautiously. What would really convince me on the efficacy of these portfolios is if the algorithm decided to sell before the stock market crashes. That would be incredible!
|
Nonetheless, I’m more than happy with the results of this experiment. It’s safe to say that AI is at least marginally helpful when it comes to automated trading. Imagine 10 years ago if you said that you could talk to an AI-Powered agent and it would bring your ideas to life for you. You would have thought I was crazy!
|
AI has so many use cases in finance, investing, and trading. This experiment proves that, at the very least, AI can help improve your trading decisions. It’s too early to conclude that purely AI-generated portfolios are inherently superior, but the mind-blowing thing about this experiment is that these portfolios were generated in just minutes.
|
I’ve said this once before, and I will say it again. Imagine if we paired AI with expert-level traders. The traders can generate different strategies across thousands of lowly-correlated assets. Testing, iterating, and improving their strategies becomes trivial, and the gap between traders using AI and traders not using AI will soon become very evident. There’s no way a human-being can match the speed and efficiency of a super-intelligent agent. The world of finance is going to change dramatically.
|
Thank you for reading! Stay tuned for our next update on these portfolios. Interested in learning how AI can be applied for your unique use case? Check out NexusGenAI. Want to try out the AI-Chat for yourself? Create an account on NexusTrade today!
|
|
