What transfers from discretionary trading (and what doesn't)

If you have been trading by hand for a while, deciding each trade in context, you have been doing discretionary trading, and you are not starting from zero. A large part of what you know is exactly what a systematic trader needs. Some of it, though, only works inside your head, and a computer cannot use it until it has been rewritten. This lesson sorts one from the other.

What carries over unchanged

Market knowledge. You know that liquidity thins out at certain times, that spreads widen around news, that gaps happen overnight and that some instruments behave differently from others. A computer knows none of this. Every one of these facts becomes an assumption in a test, and traders who have lived through them make better assumptions than people who have only read about markets.

Risk discipline. Sizing a position from the distance to the stop, thinking in R-multiples, deciding the loss before the entry: all of this transfers directly. Systematic trading makes it stricter, because the rules apply to every trade without exceptions.

The journaling habit. A journal that records the setup, the reasons, and the result is a small, hand-built data set. Systematic trading is the same discipline at a larger scale, with the computer doing the record-keeping.

Scepticism. Experienced traders have learned to distrust a setup that worked three times in a row. That instinct is exactly what protects you from the most common mistake in quantitative work: believing a test result too quickly.

What has to be rebuilt

Pattern recognition. Discretionary traders often say "I know it when I see it." That may be true, but it cannot be tested. Seeing a pattern involves dozens of small judgements about trend, distance, shape and context. To test the idea, each of those judgements has to become a condition with a clear answer.

In-the-moment judgement. Skipping a trade because "it doesn't feel right" is a decision with no definition. In a systematic approach it either becomes a rule, such as "skip entries within 30 minutes of a scheduled release", or it goes away. If you keep overriding the system by feel, you are no longer testing the system; you are testing yourself, without records.

Memory as evidence. The trades you remember are not a fair sample. Big winners and painful losers stay vivid; ordinary losses and the setups you passed on fade. Hindsight bias makes past signals look more obvious than they were, and confirmation bias makes us notice the cases that fit. A systematic test replaces memory with a count of every occasion the conditions were met, good and bad.

Turning a judgement into a function

Take a common discretionary idea: "buy a pullback in an uptrend." To make it testable, each part needs a definition. One possible version:

• "The trend is up" becomes: the 50-day moving average is above the 200-day moving average.
• "Price has pulled back" becomes: the 2-period RSI is below 10. RSI is an oscillator between 0 and 100; a very low reading means recent closes have mostly been down.

In Python, that judgement becomes a function: a named block of code that takes inputs and returns a result. Here it takes three numbers and returns True or False.

def pullback_entry(ma_50, ma_200, rsi_2):
    """Buy a pullback in an uptrend, written so that nothing is left to judgement."""
    uptrend = ma_50 > ma_200   # "the trend is up"
    pullback = rsi_2 < 10      # "price has pulled back"
    return uptrend and pullback
 
 
print(pullback_entry(ma_50=105.0, ma_200=98.0, rsi_2=6.5))
print(pullback_entry(ma_50=105.0, ma_200=98.0, rsi_2=35.0))
print(pullback_entry(ma_50=95.0, ma_200=98.0, rsi_2=6.5))

True
False
False

The first call is an uptrend with a deep pullback, so the rule fires. The second is an uptrend without a pullback. The third has a pullback but no uptrend. In each case the answer follows from the numbers alone, and there is nothing left to argue about after the fact.

This definition is not the only one, and not necessarily a good one. That is the point: once a judgement is written down, you can test it, compare it with alternatives, and throw it away if it does not hold up. A feeling cannot be thrown away so cleanly.

A translation table

Most discretionary habits have a systematic counterpart. The habit does not disappear; it changes form.

Hybrid approaches are fine, if you measure them

Many traders end up somewhere in between: a systematic scan or filter finds candidates, and a person makes the final decision. That can work. But it adds a component, your judgement, that the test did not include. If you take that route, record every signal the system produced and whether you took it. Then you can compare the trades you took with the ones you skipped, and find out whether your judgement is adding value or quietly removing it.

Where this leaves you

You bring more to systematic trading than you might expect. What changes is that each belief has to be stated precisely enough to be checked, and the check has to include every case, not the memorable ones. The next lesson gives that check a single number to aim at: expectancy.