Using AI in Trading: Why Claude Adds Speed, Not Judgment

Everyone assumes AI takes the guesswork out of trading research. It doesn't. It just moves the guesswork somewhere you're less likely to check it.

That's not a knock on Claude. We use it constantly. For research, for building tools, for testing strategies faster than we ever could by hand.

But every time we've gotten burned using AI in this business, it's because we forgot one thing. Claude is a fast assistant, not an oracle.

It will write code, crunch numbers, and explain results in seconds. It can also be confidently wrong just as fast.

You are the analyst. Claude is the intern.

Why Does a Confident Answer Feel Like a Correct One?

Here's a real example from our own research. We had Claude calculate Average True Range on our futures data. ATR is a volatility measure we use to size stops and targets.

The number came back roughly four times too big. A normal ~50-point range was showing up as ~200 points.

Claude never flagged it. It just did the math it was asked to do, on the data it was given.

It had no way of knowing that number was physically implausible for the instrument. That's not a math problem. That's a judgment call.

And here's the trap. The answer looked clean. Formatted, confident, done in a second. If we'd taken it at face value, we'd have sized every stop and target off a volatility read that was four times off. That's not a small rounding error. That's blowing an account in slow motion.

AI is a calculator, not a detective. It won't question whether a 200-point ATR makes sense on something that normally moves 50.

That's your job. Every time. No exceptions.

Does Speed Ever Replace Judgment?

No. And this is the part traders underestimate the most.

Claude can scan dozens of strategy ideas in the time it takes a person to manually test one. That speed is genuinely useful. It's why we keep using it.

But speed creates its own blind spot. Volume bias.

When you're reviewing 50 results instead of 5, you stop deeply understanding any single one of them. You start trusting the output table instead of interrogating what's actually in it.

We've caught strategies that looked statistically strong. Solid profit factor, decent win rate. And they made zero structural sense once we actually looked.

A mean-reversion approach applied to a trending instrument is one example. On paper, fine. In reality, a strategy fighting the exact behavior of the market it's trading.

AI gives you speed. Judgment is still 100% on you.

If you can't explain why a strategy works in one sentence, you don't have an edge. You have a coincidence that hasn't been caught yet.

What's the One Prompt That Catches This?

We've built one habit into every AI-assisted research session that catches more bad ideas than any technical check we run. We ask Claude to argue against us.

The prompt is simple. "Assume I'm fooling myself. What are the three most likely reasons this result is not a real edge?"

That's it. No special setup, no complicated framework. Just a direct request to stop being agreeable and start being skeptical.

It works because it forces a moment of real scrutiny into a process that otherwise rewards moving fast and feeling good about the results.

Left alone, Claude wants to help. Ask it a leading question and it leans toward yes. So we stop leading it. We point it at our own work and tell it to find the hole. Half the time it finds one we would have shipped.

This connects straight back to REPs, our core framework of Risk management, Edge, and Psychology. The psychology piece isn't about staying calm during a losing streak, though that matters too, as we've written about in How To Handle Trading Losing Streaks.

It's about staying skeptical of yourself even when a tool just handed you an answer that sounds right. Especially then.

Is This Really a Psychology Problem?

Mostly, yes. The technical fixes are simple to describe. Sanity-check the numbers, verify the data, test the logic.

The hard part is doing that consistently when the output looks polished and the deadline is close and you want the strategy to work.

That's a discipline problem wearing a technical costume. Everyone knows the checklist. Almost nobody runs it when the answer already looks good and they're tired and they want to move on.

It's the same discipline we teach around position sizing and journaling. Process over profits. The process matters more than any single result, which we go deeper on in Trading Risk Management Strategy: The Psychology Edge and in Positive Expectancy: Finding Your Trading Edge.

Patience is key, whether you're staring at a chart or a spreadsheet Claude just generated for you.

We talked through more of this on Edge Up Podcast Episode 077, "Using Claude and AI in Trading," available on Spotify. It includes a few examples that were too long for a blog post but worth hearing in full.

What Should You Actually Do With This?

Next time Claude, or any AI, hands you a clean result, don't move to the next step yet.

Ask it to argue against itself first. Check one number by hand. Ask if the result would still make sense if you explained it to someone outside trading in one sentence.

None of that slows you down much. All of it keeps the speed AI gives you from turning into a shortcut around the judgment that was always supposed to be yours.

Consistent growth in this business has never come from moving faster than everyone else. It comes from staying skeptical of your own results longer than everyone else is willing to.

Mahalo for reading and trade well! Glenn & Reid | Hawai’i Trading Academy