How to build a trading system is usually treated as a search problem. People go looking for the system, as though it exists somewhere finished and the work is finding it. Then they collect indicators, test a few things, feel briefly certain, and abandon the whole thing after a bad fortnight.
I want to reframe it, because a system is not something you find. It is something you write down. In any other business you would call it an operating procedure, and nobody would expect to discover theirs on a forum.
The test I use is blunt. Could you hand your system to a competent stranger, walk away for a week, and come back to find they took roughly the trades you would have taken? If not, you do not have a system. You have preferences, and preferences quietly change shape under pressure.
Before any of this, the standing rule on this site: no entry, stop or target discussed should be treated as a signal.
A System Is a Document, Not a Feeling
The reason this matters is not tidiness. It is that an unwritten system cannot be wrong, and anything that cannot be wrong cannot be improved.
When the rules live in your head, every trade is quietly renegotiated at the moment of entry. The setup was almost there, so you took it. The stop was a little tight, so you widened it. Afterwards there is no way to tell whether the plan failed or whether you simply did not follow it, because the plan was never fixed enough to be departed from. Six months of that produces a trader with a lot of experience and nothing to show for it.
A written system solves this by making deviation visible. That is its entire job in the first year. It is not there to be right. It is there to be a stable reference point, so that when results come in you can tell which of two very different problems you have: a system that does not work, or an operator who is not running it.
Those problems have opposite treatments, and mistaking one for the other is how people end up rewriting a perfectly good system every few months. I have written separately about when to retire a strategy that stopped working, and almost every premature retirement I have seen traces back to this confusion.
How to Build a Trading System in Four Documents
Four pieces. None of them long. The discipline is in writing them before you need them, while you are calm and have no position open.
1. The market and the hours
Name the instrument and name the window. Not "gold, mostly", but the specific market and the specific hours you will be at the desk, on the days you will be there.
This sounds trivial and it is the constraint everything else inherits. A system that requires attention at hours you cannot attend is not a system you own, it is a system you will break. If you are fitting this around employment, the capacity arithmetic of trading with a full time job is the place to settle the hours honestly before you write anything else down.
2. The setup, written so someone else could find it
This is the part people think is the whole system, and it is roughly a quarter of it.
Write the conditions that must be present, in language that does not depend on your judgement. "Trend is up" is not a rule, because two people will disagree about it. Something a stranger could verify from the same chart is a rule. If a condition cannot be checked by someone else, it is intuition, and intuition may well be valuable but it cannot be tested, so it does not belong in this document.
Be equally specific about what disqualifies a trade. Most operators find their exclusions do more work than their entry conditions, because the money is usually lost on the trades that should never have been taken rather than the ones that were taken correctly and lost anyway.
3. The risk ceiling
Two numbers: what a single position may risk, and what the account may lose in a day or a week before you stop.
The second is the one people leave out, and it is the one that determines whether a bad run stays a bad run. A per-trade limit without a per-period limit still permits you to lose a great deal in an afternoon, one correctly sized trade at a time. That is the ceiling for the whole business, and it belongs in the one page trading business plan alongside the mission line.
4. The review rule
Decide now, in writing, what evidence would make you change the system, and how much of it you need.
Without this, every drawdown becomes an open question, and open questions get resolved by mood. With it, a bad month is simply a bad month until the pre-agreed threshold is crossed. This is the document that protects the other three, and I will spend the rest of this article on why it is so hard to write honestly.
Design for the Drawdown You Will Actually Have
Here is the number that changes how people write the review rule.
I simulated 100,000 years of a system with a genuine edge. The assumptions: it wins 45 percent of the time, winners return twice what losers cost, risk is 1 percent of equity per trade, and it takes 20 trades a month. That works out to an expectancy of 0.45 times 2 minus 0.55 times 1, which is plus 0.35R per trade. Every simulated year below is a year in which the system is working exactly as designed.
The deepest drawdown reached inside the year:
- In the median year: 9.6 percent.
- One year in four: 11.6 percent or worse.
- One year in ten: 14.1 percent or worse.
- One year in twenty: 15.9 percent or worse.
- One year in a hundred: 19.9 percent or worse.
Put plainly, a system with a real edge exceeds a 10 percent drawdown in 43.2 percent of years. Nearly half. It exceeds 15 percent in 7.4 percent of years.
Now consider what most people write into their review rule without thinking: if I am down 10 percent, something is broken. On these numbers that trigger fires in almost half of all healthy years. An operator using it would abandon a working system roughly every other year, then go looking for a new one, and repeat.
The same applies to months. Across the simulation the average year contained 1.6 losing months, and 19.7 percent of years contained three or more. A quarter with two red months in it is unremarkable. It feels like evidence, and it is noise.
This is why the review rule has to be written from the distribution rather than from the feeling. Your threshold should sit outside what a working system normally does, not inside it. Change my assumptions and the numbers move, which is the point of stating them: run it with your own win rate and payoff and you get your own thresholds rather than mine.
How Long Before You Know It Works
The other half of the review rule is time, and here the honest answer is uncomfortable.
Small samples say very little. A confidence interval is the standard way of expressing how much a sample can be trusted, and the NIST handbook describes it as a range of values likely to contain the true figure rather than a single answer. Applied to a win rate over thirty trades, the range is wide enough that a promising result and a mediocre one are statistically indistinguishable.
Which leads to a rule I would defend strongly: do not judge a system on a sample you would not accept from anyone else. Thirty trades is a rehearsal. It tells you whether you can follow your own document, which is genuinely useful and is not the same as telling you whether the document works.
So run the first stretch as an operations test rather than a performance test. The question is not what did I make. It is: did I take every trade the system specified, did I skip the ones it excluded, was the size correct each time, and is the record complete enough to review. Those questions have clear answers after thirty trades. The profitability question does not.
The metrics worth watching in that period are the ones that move first, which I have set out in the key metrics every trading business should track. Adherence is a leading indicator. Equity is a lagging one, and by the time it speaks clearly, months have passed.
The Four Ways This Goes Wrong
Having watched a fair number of these documents get built and abandoned, the failures are repetitive.
Building it after a loss. A system written in the two days after a painful drawdown is a monument to that drawdown. It will be too restrictive in exactly the area that hurt you and unconsidered everywhere else. Write it flat, or write it after a win you are suspicious of.
Optimising it before running it. Adjusting rules to fit history feels productive and is mostly how a system stops working in advance. A slightly worse system you actually follow beats a finely tuned one you do not.
Adding rules after every loss. Each loss suggests a filter that would have avoided it. Add enough of those and you have a system that would have been perfect last year and cannot take a trade this year. Rules should be removed roughly as often as they are added.
Leaving the review rule blank. The most common of the four, because it is the only document that constrains you rather than the market. Without it, the system lasts exactly as long as your patience, and the numbers above show that patience is asked for more often than most people expect.
The routine that keeps all four in check is unglamorous and daily, and a pre-market routine is where the document turns into behaviour.
What I Would Want You to Take Away
A trading system is four short documents and the willingness to be measured against them. The market, the hours, the setup, the disqualifiers, the two risk numbers, and the rule that says what would change your mind.
Write them badly and revise them. A mediocre written system beats an excellent unwritten one, because only one of them can be improved, and improvement is the whole business. The operator who has run a modest system faithfully for two hundred trades knows more than the one who has tried nine brilliant ones for thirty each.
If you want the structure this fits into, the free one page trading business plan has a line for each of these numbers, so the system and the business it belongs to live on the same page instead of in different notebooks.
Frequently Asked Questions
How do I build a trading system as a complete beginner?
Start with the two documents that do not require market knowledge: the hours you can genuinely attend, and the two risk numbers. Those are decisions about your life and your capital, not about the market, and they constrain everything you write afterwards. The setup rules come third and will be revised many times.
How many trades before I know my system works?
More than thirty, and probably more than a hundred. Over small samples a confidence interval on a win rate is wide enough that a good system and a mediocre one look alike. Treat the first thirty trades as a test of whether you can follow your own rules, which is a question the sample can answer.
What drawdown means my system is broken?
Lower than most people assume, if you set it by feel. In the simulation above, a system with a real edge exceeded 10 percent drawdown in 43.2 percent of years and 15 percent in 7.4 percent of years. A threshold inside that range fires on healthy systems. Set yours from your own assumptions, deliberately, before you need it.
Should my system be automated?
Automation enforces the rules but does not write them, and it adds failure modes of its own. If your document is not clear enough for a stranger to follow, it is not clear enough to code either. Write it first, in words, and decide about automation afterwards.
How often should I change my system?
Only when the evidence you specified in advance arrives. If you find yourself changing it after individual losses, the problem is the missing review rule rather than the system. Averaging 1.6 losing months a year, a working system will regularly present you with reasons to interfere.
Can I use someone else's system?
You can start from one, and you will still have to write your own version, because the hours, the capital and the risk ceiling are yours. A borrowed system with your name on it tends to get abandoned quickly, since nothing in it was decided by the person expected to follow it.
About Rex
I'm Rex. I spent years running operations before I ever placed a trade, which is why this journal treats an account as a small business with procedures, ceilings and a set of books rather than a series of opinions about the market. More about how I run the channel.
Risk disclaimer: This article is educational and is not financial advice, an offer, or a recommendation. Trading gold (XAUUSD) and other leveraged products carries a high risk of rapid loss, and most retail accounts lose money. No entry, stop or target discussed should be treated as a signal. Simulated figures rest on the assumptions stated in the text and are not a prediction of any real result; external sources are linked so you can check them.