Optimizing the Optimizer: Creating a better AI Trading Strategy Generator
My no-code algorithmic trading platform has one extremely unique feature that other platforms simply cannot replicate: the optimization engine.
The optimization engine (also called the optimizer) is an AI optimization algorithm. It’s based on the old-school “genetic algorithm”, and allows you to improve your trading strategies automatically based on past historical performance.
Picture the naive approach. You run a backtest and sees how well it performs. Then, you update a parameter and re-run the backtest. If re-running the backtest improves the performance, then you keep that change. Otherwise, you try something else.
Genetic algorithms do this on a more massive scale. It uses a “natural selection”-inspired algorithm to automatically find the best parameters for your strategy. You can optimize for percent change, drawdown, sortino ratio, or more. You can also optimize multiple variables at the same time. This is called multi-objective optimization.
While multi-objective optimization should theoretically be useful, the reality is most users couldn’t extract the value from it due to poor UX.
This article will discuss the flaws of the genetic optimization engine within NexusTrade and how they were rectified in the most recent update.
What’s wrong with the “Optimization State”?
In fairness to me, the screenshot above was a production bug. It’s more than useless and doesn’t help the user whatsoever. This bug has been in production for around a couple weeks.
If it was working properly, the optimization state should’ve looked something like the following:
This is marginally more useful. We can see how this strategy is different from the original strategy. The white text signifies everything that didn’t change and the red text signifies the differences between this strategy and the original.
In addition to the poor UX for the Optimization State, each “Optimization Vector” also had a virtually unreadable gargled mess.
Here’s how I fixed these issues.
Show, don’t tell: the secret to good UX
When I was at Cornell University, I took a class called Creative Writing. The class was focused on writing short stories and poems that brings out the writers creativity.
The biggest lesson I learned from that class, “Show, don’t tell”, can be applied for good frontend design. Here’s how I implemented it.
For one, I streamlined the UI for the optimization vector. You don’t see an unreadable mess; you get a big blue button that says “CLICK ME!”
After the clicking the button, you’ll see this new page:
If you’re a NexusTrade user, then this page should look familiar. It’s a similar page for viewing strategies in the strategy library.
Within this page, you can seamlessly test out the newly generated strategy in a backtest. You can actually SEE how it performs, and not guess based solely on the optimization performance.
You can update the date range and the initial value, and after the backtest runs, you get a list of useful statistics.
This includes:
- The Shape Ratio: a measure of returns normalized by how risky the portfolio is
- The Percent Change: the raw change in portfolio value divided by its initial value
- The Maximum Drawdown: the largest drop in portfolio value
Amongst other useful metrics.
Now, the trader is far better likely to assess whether an optimized portfolio actually meets their goals!
Finally, after scrolling down, you can see the new strategy for the portfolio.
Conclusion: My AI Strategy Generator is actually useful
I’m neither a UX Designer nor a frontend engineer. I’m a software engineer that’s passionate about finance and artificial intelligence.
As a soloprenuer, I sometimes make mistakes. And the original design for the optimizer was definitely one of them. I got so excited in the potential of the feature that I failed to present it in a way that actually brings value to my users.
This new design is my first step in rectifying this. The genetic optimization engine is one of the most unique and powerful aspects my algorithmic trading platform. The ability to automatically find the best parameters for a strategy is extremely useful, and is something other platforms cannot easily replicate.
I hope that by updating the UX, people are able to see just how powerful this feature is! The optimization engine has transformed from a “gimmick” to a useful tool for traders to improve their trading strategy. Don’t believe me? Try it out yourself!
Visit us at DataDrivenInvestor.com
Subscribe to DDIntel here.
Featured Article:
Join our creator ecosystem here.
DDI Official Telegram Channel: https://t.me/+tafUp6ecEys4YjQ1
