Section 1.1: Portfolio Performance Overview

So, how have our portfolios fared?

Buy and Hold SPY Strategy

The stock market has seen considerable volatility in recent weeks, with the S&P 500 dropping by 3%.

Experiment: Buy and Hold SPY

Surprisingly, the unrefined ChatGPT portfolio delivered the best results by a significant margin. The strategies employed were leveraged bullish approaches, and given the current bearish market conditions, the portfolio that didn’t initiate any trades outperformed the others.

Experiment: GPT-Generated Portfolio

While it might be tempting to conclude that ChatGPT possesses exceptional trading acumen, this is not the case. The foundational strategy produced by ChatGPT contains parameters that are highly improbable to succeed under normal circumstances. Nonetheless, it’s amusing to note that the untouched GPT-Generated Portfolio achieved the best outcome.

Backtest Performance of GPT-Generated Portfolio Over 5 Years

Next, we have the "One and Done Optimized Portfolio," which was optimized only once at the experiment's inception. This portfolio performed reasonably well, with a loss of about 6% in the past two weeks.

Experiment: One and Done Optimization

Moving on to the Sliding Window Optimized Portfolio, which undergoes weekly re-optimizations. For each new iteration, we adjust the start and end dates. Unfortunately, this portfolio did not perform well, resulting in a loss exceeding $1,000 in the last two weeks.

Experiment: Sliding Window Optimization

This outcome is notably worse than simply holding TQQQ, the asset this portfolio trades.

Buy and Hold TQQQ Performance

Finally, we have the Expanding Window Optimized Portfolio, which operates similarly to the sliding window version. The key difference is that the expanding window portfolio maintains a constant start date while incrementing the end date with each optimization. This portfolio performed the poorest, incurring even greater losses than the sliding window approach.

Experiment: Expanding Window Optimization

These early results provide us with some intriguing insights.

Psst! This portfolio was initially shared on Aurora's Insights. Subscribe if you’re keen on AI, Finance, and their intersection.

Section 1.2: Key Lessons Learned

Risk Management Considerations

A critical factor I overlooked while selecting the new optimized portfolio was risk management. The optimization process produces a variety of high-performing portfolios. Some may yield higher returns but with increased risk, while others might offer lower returns with reduced drawdowns. Initially, my focus was on raw gains, leading to the selection of portfolios that heavily invested in TQQQ—a highly volatile triple-leveraged stock. This strategy works well in a bullish market but fails catastrophically in bearish conditions, as evidenced by our results.

In retrospect, a more balanced portfolio akin to the One and Done Optimization Portfolio would have provided better options during both bullish and bearish market phases.

Choosing the Right Fitness Function

During the optimization, I emphasized the Sharpe ratio, Sortino ratio, and overall percentage change. Although this approach is valid, it caused me to overlook the potential drawdown of certain portfolios. In hindsight, evaluating both maximum and average drawdown could have led us to select a less risky option. In the next round of optimization, I considered drawdowns more carefully and intentionally chose portfolios with lower risk profiles.

Replacing the Sliding Window Portfolio

Updates to the Optimization Engine

We’ve also made improvements to the optimization process by resetting the initial positions in the portfolio. NexusTrade aims for maximum configurability, so we’ve enhanced the process by including various options for initial positions, ranging from random placements to retaining current positions.

Bug Fixes 🐛

Initially, there was a bug affecting the optimization of compound indicators, which prevented correct crossover and mutation operations. This issue has now been resolved.

NexusTrade Referral Program

Our optimization engine is a standout feature of NexusTrade but requires substantial computational resources. Therefore, we’re gradually rolling it out to a select group of power users on the platform.

Want early access? Follow these steps:

1. Confirm your email on the profile page.
2. Invite friends using your unique referral code.
3. Ensure your friends confirm their email addresses.
4. Email us at [email protected] to request early access.

What rewards can you expect? They are remarkable!

• Invite 2 friends for 1 week of access 😲
• Invite 5 friends for 1 month of access 😱
• Invite 15 friends for 6 months of access 💀
• Invite more than 15 for a special invitation from Austin Starks, the creator of NexusTrade 🥳

Referring friends is the sole method to gain access to these premium features for FREE.

Go to your NexusTrade Profile