I've been experimenting quite a bit with using AI as part of the strategy development process, and I recently tried something a little different. I asked Claude to come up with a technical indicator designed to identify potential blow-off tops, situations where price has accelerated in a parabolic manner and may be vulnerable to a reversal.
Claude came up with an indicator it called BlowoffScore, which combines several characteristics of price behavior into a 0–100 score. It implemented the indicator in WealthLab and then I ran it through some analysis.
I tested BlowoffScore across historical market data in WealthLab and looked at how stocks behaved after reaching unusually high readings. The results gave me a chance to see whether Claude's hypothesis held up against actual market history rather than simply accepting a plausible-sounding AI explanation. I documented the entire experiment here:
Blog:
https://www.wealth-lab.com/blog/i-asked-claude-to-build-a-trading-indicator-then-i-backtested-it
I also made a video showing the process:
https://youtu.be/ByLjFFREMZQ
I think this is where AI becomes particularly interesting for quantitative traders. Instead of asking AI to magically produce a profitable strategy, we can use it as an idea generator and development partner, then use WealthLab to determine whether those ideas actually survive contact with historical data.
I'd be interested to hear how others are using AI in their own strategy development process. Are you using it mostly for coding, generating trading ideas, analyzing results, or something else?
Claude came up with an indicator it called BlowoffScore, which combines several characteristics of price behavior into a 0–100 score. It implemented the indicator in WealthLab and then I ran it through some analysis.
I tested BlowoffScore across historical market data in WealthLab and looked at how stocks behaved after reaching unusually high readings. The results gave me a chance to see whether Claude's hypothesis held up against actual market history rather than simply accepting a plausible-sounding AI explanation. I documented the entire experiment here:
Blog:
https://www.wealth-lab.com/blog/i-asked-claude-to-build-a-trading-indicator-then-i-backtested-it
I also made a video showing the process:
https://youtu.be/ByLjFFREMZQ
I think this is where AI becomes particularly interesting for quantitative traders. Instead of asking AI to magically produce a profitable strategy, we can use it as an idea generator and development partner, then use WealthLab to determine whether those ideas actually survive contact with historical data.
I'd be interested to hear how others are using AI in their own strategy development process. Are you using it mostly for coding, generating trading ideas, analyzing results, or something else?
Rename
QUOTE:
Instead of asking AI to magically produce a profitable strategy,...
Yes, this doesn't work. The solution returned is oversimplified.
QUOTE:
... we can use it as an idea generator and development partner,...
Yes, asking it for a specific solution works better. But you still need to work out the kinks.
And you still need to integrate several specific solutions into one giant integrated solution that confirms all the specific triggers.
AI is clever at C# coding. For example, it presented a method to allocate storage on the stack using malloc (without using the "new" operator, which allocates a .NET object on the heap) so as to avoid garbage collection later. That's a neat trick I never thought of.
Most definitely, I learned so many new tricks using AI to aid in coding these days!
Your Response
Post
Edit Post
Login is required