Connect your Claude, Codex or OpenClaw for full transcripts on millions of podcasts. Two week free trial, cancel anytime.
Start free trial →Ditching Hourly · Agency Business · November 2026
Asked for the most common mistake among traders using AI, Dave Mabe says it is having the model write their own backtester instead of teaching it an existing tool. The model underestimates the edge cases.
…mistake I see or the most frequent mistake I see now is people using Claude to create their own backtesting software. So there's lots of backtesting software out there that works. And a much better approach is have teach Claude how to use the existing backtesting software. But the first way is that
because intuitively they probably have the or certainly intuitively on my part, I might have fallen into the same trap because I know it's good at making software. So why do you think it doesn't work?
I think you like backtesting software is more complicated than Claude thinks it is. And there's all these edge cases, there's stock splits, there's Reverse splits, there's corporate actions. Oh, wow. There's all sorts of things that can skew your back tests, like can skew your data if they're not accounted for. And just not anticipating that. And the first thing that Claude's going to do is, yeah, we can write a back tester, Jonathan. Yeah, no problem. No problem. And then, you know, two months later, they realize they've got this big pile of technical debt that they don't even realize. And they're just kind of stuck. And
I mean, unlike perhaps a non-technical founder who's building, you know, what they think is the next greatest to-do app, someone who's using it to build their own back tester is losing money.
Right.
Most almost certainly. Yeah. So they notice. It's not just technical debt, it's debt.