discretion is the new edge

Most people building bots right now never had an edge to automate in the first place.

This article is for two kinds of people. Traders who already make money reading charts and are starting to think they need to automate, and people getting into trading who are being told the first step is to build a bot.

If you learn to use a calculator before you learn your times tables, you'll never notice when the answer is obviously wrong. You end up copying down whatever's on the screen. That's what happens when you automate something you don't really understand.

Painting of a line of blind men holding staffs, the leader tumbling into a ditch and pulling the next one after him.

A lot of people automate because it feels more "scientific". AI tools have become cheap to use, backtesting is easy and a spreadsheet full of profitable trades can feel like proof for something working.

When someone builds a bot and it stops working they often respond in two ways.

Some just keep building bots. If the next version works they never have to admit they didn't understand the first one, and sometimes it can work for a while, which is exactly why it's so dangerous. It can waste so much of your time.

The others decide trading doesn't work. They convince themselves that there's no edge left or that the market's rigged, so they quit. Neither fixes anything, because the bot was never the problem.

Loop diagram: build a bot, works for a while, stops working, build another one, with an exit from stops working to quit.

So why does this make discretion the edge?

Information is not an unsolved problem. Everyone has the same charts, the same data and now the same AI. What separates people is not what they know, but what they actually do with it in spite of fear, boredom, ego and exhaustion. What are you actually going to do in REAL-TIME.

The most profitable traders I know have spent an endless amount of hours reviewing their own trades. Replaying the market at critical historical moments in time and embedding this into their brain. All of this requires screen time and understanding first principles and why things happen to begin with.

I believe someone coming into this business from scratch and going at it as a fully discretionary trader has a far better chance of making it than someone coming at it from a fully systematic AI standpoint. Not necessarily because "AI is bad" (I'll get to how I'd actually recommend using it soon), but because the AI route is much more likely to drag you away from first principles thinking. You start asking the AI what works instead of working it out yourself.

Chart of skill over time: the screen time curve keeps climbing while the AI shortcut curve flattens out, with the widening space between them labelled the gap.

How do the best traders use AI?

More efficient reporting, customised tooling, automating an edge they already understand. I built my own journaling software personalised specifically to make it easier for me to keep track of different niches within my own edge.

I like to think about this by reverse engineering my discretionary trading. What am I currently doing manually that could be automated or made more efficient using AI.

What is my constraint as a trader and how can I use AI to resolve it.

I had the problem of not being able to calculate my position size as fast as I would like based on where I was wrong on my idea and the risk I wanted to place on the thesis. I know the constraint, the solution was to build a quick access position size calculator which I could hit on hotkey at any time.

There is still and always will be plenty of room for new traders to come into this business and make money. The success rate is as slim as it always has been, the only thing that's changed is the distractions keeping people away from the actual productive thing they need to be doing.

The gap between a newer trader and the best traders was already quite big to begin with. Now that everyone is being sold the "make me a trading bot, make no mistakes" dream, that gap has now widened 100x.

I recommend using Hyperliquid. Sign up with my link for 4% off your trading fees.

Affiliate link.