“I tested it with a large sample size.
I followed the rules.
And yet, for some reason, my live results look nothing like what I saw in testing.”
Has the strategy stopped being profitable?
Or has the market changed?
But this is not the first time this has happened.
No matter what you test or how much time you spend, you follow the rules in live trading, yet it still does not work.
This article explains why this keeps happening, what causes it, and how to solve it.
■“Having Rules” Does Not Mean “Having a System”
Put simply, the problem is that you have rules, but your system is incomplete.
Most traders have entry rules.
They have exit rules too.
They have also tested those rules.
So they believe they have a system.
But that is not enough to call it a system.
Suppose you have a rule that says, “If A happens, do B.”
Most traders have rules in this form.
At first glance, it looks as though the decision has already been made.
But to repeat the same decision consistently, there is still more that must be defined.
When exactly do you confirm that A has occurred?
Under what conditions should B not be triggered, even if A has occurred?
What would invalidate A?
In what order must the conditions that make up A appear?
If even one of these is undefined, you will have to supply the answer in the moment.
Faced with the same A, you will do B one day and not do it another.
And this does not apply only to A.
B, and everything before and after it, has its own decisions that must be defined in the same way.
When all these decisions fit together without contradiction, exist at the right places and times, and every possible answer leads to a defined next step, the whole structure becomes a system.
Having a few rules is not the same as having a system.
■Build a Reproducible System
A system must be reproducible.
Reproducibility does not mean producing the same profit every time.
It means handling the same conditions the same way every time.
If you handle the same conditions differently in testing and live trading, then even if you follow the same rules, you are not running the same system.
Reproducibility does not come from the number of rules you have.
It depends on whether the trigger that begins each decision and the next step that follows it are defined at the right place and time.
At this point, some people may think, “If human judgment can drift, why not automate everything?”
Automation can certainly reduce execution drift.
But it does not solve problems that already exist at the rule design stage.
Automation can only repeat processes that have already been defined.
If necessary conditions or branches are missing, automation will not find and fill them for you.
If the design is crude, all you get is an incomplete system repeated with precision.
The issue is not whether a human executes it or a program does.
The system being executed must be complete first.
Most traders do not have rules at that level.
What makes this even more difficult is that it is hard to notice what is missing.
It took me a long time to recognize this problem myself.
If you do not recognize it, all that remains is the question, “Why is this not working when I am following the rules?”
You probably do not believe that you lack reproducibility.
After all, you have rules, and you follow them.
Then why, after testing it, do you still discard the strategy as “unusable”?
How long will you keep repeating this?
You are left with no choice but to blame the strategy or the market, so strategies keep getting discarded before their true potential is ever established.
If testing had been completed properly, all that should remain is execution.
No additional testing should be necessary.
Before you throw away another strategy, stop for a moment.
What if the problem was not the strategy, but the fact that the system running it was never complete?
If you leave that unchanged, the same thing will happen with the next strategy.
You test it, move into live trading, fail to make it work, and begin searching for another strategy.
No matter what you choose, you return to the same place.
There is only one thing you need to do to escape this loop.
Do not change the strategy.
Do not simply add another rule whenever one occurs to you.
Build a reproducible system.
■“It Is Impossible to Create Rules for Everything the Market Can Do”
You may be thinking exactly that by now.
You are right.
You cannot predict everything the market will do next.
But a system does not decide what the market will do.
It decides what you will do when something happens.
If the conditions are met, execute.
If they are not, do not trade.
If it is still too early to decide, define in advance whether you will continue monitoring or stop.
An unexpected market move does not require your next action to become unexpected too.
The same applies to discretion.
Define when it may be used, what you will look at, and how much choice is allowed.
When those boundaries are defined, discretion becomes part of the system.
By contrast, if your rule is merely “decide when the time comes,” you cannot verify whether you made the same decision in testing and live trading.
A complete system is not one in which the market never does anything unexpected.
It is one in which an unexpected event never leaves your next action undefined.
■Before You Throw Away the Strategy
You follow the rules, but live trading does not match what testing showed.
When that happens, do not immediately blame the strategy or the market.
First ask whether the system running the strategy was truly complete.
But everything described so far is only a small part of what a system requires.
In addition to entries and exits, you must define invalidation conditions, conditions under which you do not trade, what you monitor, and position sizing.
And it is not enough for each rule to exist in isolation.
Every rule must exist at the right place and time, fit together without contradiction, and operate as one continuous structure.
Only then can you repeat the same process under the same defined conditions and test the system with a large sample size.
Only then can you use those results to determine whether the system has an edge.
Building all of this is a skill in itself.
It is not as simple as adding another rule whenever you notice something missing.
The hardest part is finding the gaps you do not even realize require a rule, then integrating them without breaking the whole.
TRADING SYSTEM ARCHITECTURE is the blueprint for turning your strategy into a reproducible system that allows probability to work.
It lays out, as one complete process, what to define and in what order, from rule design and testing to position sizing based on your test results.
If you have not yet built a reproducible system with an edge, I want you to use it.
Before you search for another strategy, finish turning the one you already have into a system that can be tested properly.
Without a system with an edge, you have not even reached the starting line in trading.