9. Creating a New Composer Symphony with AI

You can create a new symphony from anywhere on the Composer platform by accessing the dropdown menu attached to the Create button in the top navigation bar.

Developing a New Symphony

Once you’ve started a new symphony, there are three primary ways to build it:

1. Manually entering information into blocks – Construct the symphony from scratch by adding and configuring blocks individually.
2. Pasting existing blocks from other symphonies – Reuse logic from previously created symphonies to streamline development.
3. Using “Create with AI” – Leverage ChatGPT to generate a symphony based on specified parameters and investment goals.

Clicking this button opens up an AI chat box:

Let’s start with a basic example by asking ChatGPT to generate a symphony similar to the TQQQ For the Long Term symphony we edited earlier.

This strategy is extremely straightforward, and as expected, its performance results are somewhat underwhelming.

This symphony outperforms SPY over longer timeframes, but it comes with significant drawbacks. Investors would need to endure extended drawdowns and, at one point, a months-long period of flat returns after rotating into BIL in early 2022.

To introduce more diversification and potential resilience, let’s ask ChatGPT to add foreign ETFs to the strategy. This adjustment may help reduce drawdowns and improve overall risk-adjusted returns.

While this adjustment creates a more complex symphony, it does not improve performance. The updated logic results in an annualized return of just 0.8% with maximum drawdowns of 2.2%—a performance so weak that simply holding cash might have been a better option.

The attempt to add foreign ETFs significantly weakened the symphony’s performance. Now, let’s test a different approach—rotating into volatility ETFs (tracking the VIX) when QQQ’s relative strength becomes excessively high.

This request improves annualized returns to 7.4% a year and reduces the maximum drawdown to just 0.6%, which is at least better than a money market fund:

Despite multiple adjustments, our new strategies continue to underperform the original symphony. Since adding foreign ETFs weakened performance and volatility ETFs didn’t help, let’s take a more unconventional approach—replacing the underperforming foreign ETF with a popular Bitcoin-related ETF.

This is a pretty simple swap, as it just replaced VXUS with BITO while adjusting the If/Else blocks leading to the BITO block:

This approach seems to have done the trick, as our annualized returns have more than doubled from our initial TQQQ-or-bust approach.

One noticeable difference in this backtest preview is that the green line, representing our ChatGPT-generated symphony, starts much later than the pink line for SPY. This is because BITO, our Bitcoin-related ETF, has only been available for trading since mid-2022. Since Composer can only backtest assets with available historical data, it’s impossible to test this strategy prior to BITO’s launch.

Does This Make Our “TQQQ, Bitcoin, and Volatility Strategy” a Failure? Not necessarily. As mentioned throughout this guide, backtest results should not be viewed as a guarantee of future performance. A strategy that has performed well from 2010 to the present could fail next year if market conditions shift beyond its ability to adapt. Conversely, a symphony with lackluster backtest results might thrive in live trading.

A short backtest period is often a red flag, but the only true test of a symphony’s effectiveness is how it performs in real-world market conditions. There are two ways to gain confidence in a strategy’s Out of Sample (OOS) performance:

1. Invest real money – The most direct way to evaluate performance.
2. Monitor the symphony without investing – Let it run and observe how it behaves over time. This is what most users do with symphonies found on the Discover page, aside from the newest ones.

Ultimately, past performance never guarantees future results—the best way to validate a symphony is to let it interact with live market conditions and assess whether it meets expectations.