Free PDF field guide · Agentic trading
How I build an AI trading agent that learns from every trade.
And how you can too, even with zero coding experience.
- Why the next indicator, signal service or course was never going to fix it
- The 60% win-rate trap, and the everyday habit behind it
- Why a plain trading algorithm won't work, and what an agent does differently
- The one rule I hard-coded before the agent ever touched an order
- How to build a system that checks every rule and learns from every outcome
→ not enough evidence yet. kept v13.
Illustrative. Not actual trades or results.
Research cited inside
Inside the guide
Seventeen pages. No hype, no signals, no shortcuts.
The honest story of 18 years spent searching for the right indicator, signal and strategy, and the agentic system that finally ended the search.
p. 06
The 60% win-rate trap
Six winners, four losers, and the account still goes down. The everyday habit that quietly flips the math, and why a better entry signal won't save you.
Fictional illustration from the guide. Same win rate, different habits.
p. 07
The one rule I hard-coded before the agent ever touched an order
And why automation without it doesn't fix your trading. It just helps you fail faster.
p. 11
The three layers every agent I'd trust is built on
Why a perfect signal still isn't allowed to trade on its own, and which layer gets the final say.
p. 12
Learning, with boundaries
How the agent turns wins and losses into tested improvements, and the one thing it must never be allowed to change.
p. 15
The five-minute process audit
Seven questions that show exactly where your own process leaks, before you write a single line of code.
All 17 pages, free.
The core idea
An algorithm repeats your mistakes. An agent learns from them.
After a full workday, I expected myself to be focused, patient and disciplined at the charts. I was asking a lot of a tired person. Signals and indicators tell you when, but you still have to execute. An algorithm executes, but it repeats its rules forever, including the flawed ones.
A self-learning agent is different: it writes its own rules. You give it a goal and risk limits it can't touch, and every rule it comes up with has to prove itself on data first. Before a trade it weighs news and market context alongside its strategy, the way a good trader would. It tests every idea, validates changes against evidence, learns from every win and every loss, and can never touch its own risk limits. Page 11 shows how it's built.
Illustrative sequence. Not a record of actual trades or results.
Where do the rules come from?
- Signals & indicators
- Someone else's
- Algorithm / EA
- You write them, they never change
- Self-learning AI agent
- It develops its own and proves them on data
Who executes?
- Signals & indicators
- You, whenever you can
- Algorithm / EA
- The code
- Self-learning AI agent
- The agent, inside your risk limits
News & trend shifts
- Signals & indicators
- You notice them, if you're watching
- Algorithm / EA
- Ignored, or a fixed filter at best
- Self-learning AI agent
- Weighs news and market context before every trade
After a losing week
- Signals & indicators
- You look for the next one
- Algorithm / EA
- Nothing changes
- Self-learning AI agent
- It investigates and tests a fix
Learns from outcomes?
- Signals & indicators
- No
- Algorithm / EA
- No
- Self-learning AI agent
- Yes, from every trade
Risk limits
- Signals & indicators
- Your willpower
- Algorithm / EA
- Hard-coded
- Self-learning AI agent
- Hard-coded, and it can't touch them
Your goal, its rules
You set the goal and the risk limits. The agent finds the rules and keeps only what survives the evidence. Boredom, FOMO and revenge aren't inputs.
Limits with teeth
The risk layer can veto any trade, and the agent can't raise its own limits to win a loss back.
Runs while you work
Routine decisions stop depending on your constant attention. You supervise instead of babysitting charts.
Learns from every trade
Wins and losses become tested, reversible improvements. Never a gut-feeling rewrite after a bad week.
Who wrote this
“Trading was supposed to buy me freedom. Instead it kept me glued to a screen.”
I'm Hendrik. I read The 4-Hour Workweek at 18, and one idea never left me: build something that works even when I'm not working on it, on holiday, with my family, asleep. That was 18 years ago. I've tried almost everything since, and nothing tested me like trading.
The signal came in at 2:14pm. I was in a meeting. I opened MetaTrader on my phone under the table, already late, and entered the trade without really checking it. It lost. And it kept happening.
I tried every angle of it. Expert advisors. Signal services. Indicators by the dozen. Trading software, courses, a shelf full of books. I even won a BDSwiss trading challenge against more than 500 traders.1 And I still couldn't make trading fit into my life. It always ended up as a second job.
When AI arrived, I used it to build my own trading systems. Faster code, faster backtests, same problem. Only when I built a self-learning AI trading agent, one that tests, validates and learns inside limits it can't touch, did trading finally start to fit my life. The guide is the honest version: what worked, what didn't, and what is still my job.
1 Individual result in a single competition. Not typical and no indication of future results.
Read it tonight. Implement it tomorrow.
Short on time? The quotes, the three-layer framework and the five-minute audit give you the core in ten minutes.