1. Discretionary vs systematic: what really changes when you automate a decision
Theory 5: What algorithmic trading really is
A trader who follows a system by hand, with pen and paper and without a single line of code, is doing systematic trading. An automated algorithm may not be.
With the fundamentals of markets, execution, assets and risk now covered, this level finally enters the territory that gives this course its name: algorithmic trading.
Before designing anything, it is worth clarifying precisely what really changes when a trading decision goes from discretionary to systematic. It is a distinction that gets muddled all the time.
What a discretionary decision is
Discretionary trading is trading in which a person, at the moment of deciding, weighs the available information and makes a buy or sell decision using their own judgment in real time. That information might be charts, news, earnings or intuition accumulated through experience.
There is no rule written in advance that mechanically determines what to do in every possible situation. Each decision is, to some degree, a fresh interpretation of the context.
What a systematic decision is
Systematic or algorithmic trading defines in advance — before the moment of decision arrives — a set of precise, unambiguous rules that determine exactly what to do in every possible market situation.
It then executes those rules, and there are two equally valid ways of doing so:
- Manually, by a person who simply follows the system without deviating.
- Automatically, by a program that applies the rules with no human intervention at the moment of execution.
The difference is not human versus machine
This is where the contradiction from the opening comes in. The key difference is not 'human versus machine'.
A trader can follow a completely mechanical system by hand, without a single line of code, and that is still systematic trading.
And the reverse holds too. A badly designed algorithm, with parameters set arbitrarily by its creator with no statistical basis whatsoever, can end up being as discretionary in its origins as any intuitive decision. Only automated in its execution.
The real difference lies in whether or not there is an explicit, reproducible rule, defined before the outcome is known. A rule that any other person, or any computer, could apply in exactly the same way and reach the same decision given the same information.
The three practical advantages of a system
This distinction matters for several practical reasons.
The first is consistency. A system, by definition, always applies the same rule: in any market condition and in any mood of whoever is running it. Discretionary trading is inevitably exposed to the psychological biases we saw in the previous level. The same trader may react differently to the same price signal depending on whether they have had a good or a bad day.
The second is verifiability. An explicit rule can be tested against historical data before a single real euro is put at risk, which is the backtesting we will look at in the next module. A purely intuitive decision does not allow you to do that rigorously, because there is no fixed rule to apply retroactively and consistently to the past.
The third is scalability. A system can watch hundreds or thousands of assets simultaneously. That is materially impossible for a discretionary trader, who can only pay conscious attention to a limited number of things at once.
Where human judgment still wins
None of these advantages automatically makes systematic trading 'better' than discretionary trading in every case.
An experienced discretionary trader can incorporate nuanced, shifting context that no system based on past patterns would know how to interpret correctly. Think of a new geopolitical situation with no clear historical precedent. The system has never seen a comparable situation in the data it was designed or tested on.
Automating is not improving
What is true, almost without exception, is that a badly designed system is not automatically better just because it is automated. We mean systems with ambiguous rules, no rigorous backtesting, no risk management defined in advance.
Automation multiplies the speed and the execution consistency of a decision, good or bad. Not the intrinsic quality of that decision.
The rest of this level is devoted precisely to understanding what it takes for that decision, before you automate it, to rest on a solid foundation.