Trading Rules: A Rule Needs a Number

Michael Whitman · Trading since 2017, full-time since 2020 · Reviewed 2026-09-06

Trading rules are the individual instructions inside a trading plan, and a rule only counts as one if two people reading it would take the same action. That requires numbers rather than descriptions, because a rule without a threshold cannot be checked or broken.

How it works

A candlestick chart of the site's shared price history. The headline on the chart reads: Specific enough that somebody else could apply them. — Specific enough that somebody else could apply them. Illustrative chart - not real market data.

A rule is an instruction that produces the same action regardless of who reads it. That is a higher bar than it sounds, and most written rules fail it.

A gently rising stretch of the long price series with an account equity curve beneath it. The headline on the chart reads: If two people read a rule differently, it is not a rule. — If two people read a rule differently, it is not a rule. Illustrative chart - not real market data.

“Only trade strong trends” is not a rule. Two people reading it disagree about what counts, and one of them is you on a different day. “Only take long trades when the 50-period average has been rising for at least ten bars” is a rule, because it has a number and it produces one answer.

A declining stretch of the long price series. The headline on the chart reads: Every rule needs a number or it cannot be checked. — Every rule needs a number or it cannot be checked. Illustrative chart - not real market data.

The number is what makes it checkable. Without one, you cannot look back at fifty trades and count how many followed the rule — which means the rule cannot be evaluated, improved, or shown to have been broken.

Fewer, harder, logged

A calmly advancing stretch of the long price series with a slowly rising equity curve beneath it. The headline on the chart reads: Five rules you follow beat twenty you do not. — Five rules you follow beat twenty you do not. Illustrative chart - not real market data.

A long rule set is usually an unfollowed one. Twenty conditions produce constant partial compliance, which is the same as having no rules and a great deal more paperwork. Five that are actually applied are worth more than any number that are not.

A flat, quiet stretch of the long price series. The headline on the chart reads: A rule with an exception is a preference. — A rule with an exception is a preference. Illustrative chart - not real market data.

“Unless it looks especially good” is the phrase that dissolves a rule. Every discretionary exception is applied precisely when the pressure to act is highest, which is when rules exist to help.

If an exception is genuinely warranted, write it into the rule with its own number. A rule with a documented exception is still a rule; a rule with an undocumented one is a preference.

A strongly rising stretch of the long price series with a gradually rising equity curve beneath it. The headline on the chart reads: And a broken rule only exists if it was recorded. — And a broken rule only exists if it was recorded. Illustrative chart - not real market data.

Enforcement is the log, not willpower. A broken rule that nobody wrote down did not happen as far as any later review is concerned, and the memory of it will be reconstructed around whether the trade worked.

A single column in a spreadsheet — rule followed, yes or no — turns compliance into a countable number. Sixty percent compliance is a fact you can do something about; “I mostly follow my rules” is not.

In practice

A flat but volatile stretch of the long price series with an account curve that breaches its limit. The headline on the chart reads: Change them on a schedule, never after a loss. — Change them on a schedule, never after a loss. Illustrative chart - not real market data.

Rules changed after a loss are fitted to the most recent event. The sample restarts, the change is never tested, and the process repeats. Changing them on a fixed date, with the log in front of you, is the version that can improve.

A candlestick chart with a volume histogram beneath it, with the volume histogram emphasised. The headline on the chart reads: Participation filters are the one rule that uses new information. — Participation filters are the one rule that uses new information. Illustrative chart - not real market data.

A volume condition is the one rule type that adds an input the price rules do not already have. Most rule sets are several conditions computed from the same closes, which the confluence page treats as the standard way traders manufacture agreement.

A long-horizon candlestick view of the same price series. The headline on the chart reads: On a longer horizon fewer rules are tested per year. — On a longer horizon fewer rules are tested per year. Illustrative chart - not real market data.

A slower method tests its rules fewer times a year, which means it takes longer to know whether they work — a real cost of long-horizon trading that is rarely stated.

A candlestick series containing several opening gaps, with the largest opening gap marked. The headline on the chart reads: No rule survives a gap, which is why size is a rule. — No rule survives a gap, which is why size is a rule. Illustrative chart - not real market data.

A declining stretch of the long price series, with the entry price and the level at which a stop would trigger drawn as horizontal lines. The headline on the chart reads: The stop rule is the one most often broken. — The stop rule is the one most often broken. Illustrative chart - not real market data.

The stop rule is broken more than any other, and it is the one whose breach costs most. Which makes it the first candidate for automation: an order placed at entry does not require a decision later, and the moment the decision would have been made badly is removed entirely.

And each rule-following trade still costs a share of a bar. Illustrative chart - not real market data.

Following the rules perfectly still costs 2% of a median bar’s range per trade on this site’s shared history, which is why a rule limiting frequency is as important as any rule about entries.

A 72-bar candlestick section of the shared price history. The headline on the chart reads: The market has not agreed to any of them. — The market has not agreed to any of them. Illustrative chart - not real market data.

What trading rules are not

They are not a strategy. They are how a strategy is written down.

They are not guidelines. A guideline you sometimes ignore is a preference.

They are not improved by adding more. Compliance falls as the count rises.

And they are not a substitute for sizing. A perfectly followed rule set at the wrong size still ends an account.

When it fails

A sideways, range-bound candlestick series. The headline on the chart reads: In a range a good rule set still loses slowly. — In a range a good rule set still loses slowly. Illustrative chart - not real market data.

A good rule set loses money in the wrong regime. It produces exactly the trades it was written to produce, in conditions where those trades do not work, and each one pays the round trip. That is not a failure of the rules and it looks identical to one.

The second failure is the drifting definition. A rule about “a strong close” means something slightly different each month, and the drift is invisible because nothing is written down.

A third is rules that cannot be violated. A rule set that has never been broken is either automated or too loose to constrain anything.

A fourth is optimisation dressed as refinement. Adjusting a threshold from 20 to 22 after reviewing outcomes is fitting to the sample you have.

And a fifth is having rules for entries and none for size, frequency or stopping, which is the usual shape and the one that leaves the important decisions unwritten.

The original data

On this site’s shared 576-bar history the round-trip cost is 0.0098 price units — 2% of the median bar range of 0.493 and 45% of the smallest bar — and the 10-bar efficiency ratio has a median of 0.34, with only 30% of bars above 0.5. The figures are in research/series-measurements.json, produced by site/measure_series.py.

A 72-bar window of the shared price history, cut short at the decision bar. The headline on the chart reads: The rule says exit and the trade looks fine. Exit? — The rule says exit and the trade looks fine. Exit? Illustrative chart - not real market data.

That efficiency figure is the one to build a rule around. If only three bars in ten sit in conditions where price travels efficiently, then a rule set with no regime filter is trading the other seven — and the filter is a number you can compute rather than a judgement you make. Write one rule that says which conditions you will not trade in, with a threshold in it, and it will do more for a result than any refinement of the entry, because it removes trades rather than improving them.

Trading plan is the document these rules live inside. Discipline is what following them is usually called, and mostly it is design. And trade review is where compliance becomes a number.

What I actually do

The test that fixed my rules was reading them aloud to somebody who does not trade and asking what they would do today. Every time they hesitated, the rule had judgement hidden in it - and the judgement was where my results were actually being decided.

— Michael Whitman

This page is educational, not financial advice. Test every idea on your own charts before risking money.