What goes in a trading plan

Two documents with the same name 

One document says that major currency pairs are traded on the four hour chart when the setup looks right, at sensible size, with a stop below the recent low. The other names the instruments and the ones deliberately excluded, the hours inside which a position may be opened, the condition that has to hold before a position exists, the price level at which the reason for holding it has stopped applying, the fraction of equity a single position may place at risk, the aggregate limit across everything open at once, the conditions under which the position closes in either direction, what is recorded about it afterwards, and the occasion on which the document itself is next examined. Both are called trading plans. Only the second one can be failed.

That is the working test, and it is more useful than any definition. A plan that cannot be failed records nothing, because a statement loose enough to accommodate whatever was done afterwards is satisfied by whatever was done afterwards. The first document is not a weaker version of the second. It belongs to a different category: it describes a disposition, and a disposition adjusts silently to fit the conduct it was supposed to constrain.

What the document specifies 

The components below are the ones a written plan conventionally settles. They are set out as components rather than as recommendations. What belongs inside each of them is a matter for whoever writes the document, and no value, fraction, instrument, hour or condition is put forward here for anybody.

• Scope. Which instruments are traded and, more usefully, which are excluded. Instruments differ in the hours they move, in how far they ordinarily travel and in what they cost to hold, so a document silent on scope leaves the sizing rules underneath it undefined.
• Timing. The sessions or hours inside which positions are opened, and the occasions on which nothing is opened at all, such as around a scheduled economic release or in the opening minutes of a session.
• The condition for opening. What has to be true before a position exists, stated in terms that can be read off a chart, a calendar or an order book rather than inferred from an impression of one.
• Invalidation. The level or condition at which the reason for holding the position no longer applies. This is where the stop rule comes from, and it is the difference between a stop placed where the position stops making sense and one placed where a loss becomes uncomfortable.
• The size derivation. The rule that turns an invalidation distance into a number of units, covered earlier in this module. A plan states the rule rather than the number, because the number is different for every position and the rule is not.
• Aggregate limits. What may be open at once, how much of the account those positions place at risk in total, how correlated exposures are counted, and the stopping rule across a day or a week.
• Exits other than the stop. The conditions under which a position closes while it is ahead: a stated level, a stated condition, an elapsed time, a scaling arrangement. A size decision with no exit stated is incomplete, and the gap gets filled in the moment by whoever is watching the position.
• The record. What is written down about each position and when it is written, which is what makes any later examination of the document possible at all.
• Review. The occasion on which the document is next read, and the circumstances under which it may be changed.

The last two are the ones most often absent, and their absence is the ordinary failure rather than an exotic one. A document with no record clause produces no evidence about itself, so the question of whether its rules were followed has no answer beyond recollection. A document with no review clause is revised anyway, and it gets revised at the only moment anyone reads it closely, which is in the middle of a position that is not going well.

Nothing in that block was decided while the position was being considered. Each row consumes a rule written earlier and hands its result to the next, which is why the components are conventionally described as one document rather than a list of separate good practices. A plan missing the size derivation produces no size at all. A plan missing the aggregate limit cannot answer whether a position is permitted, and that question arrives when several positions are already open and one more looks appealing.

A rule two readers would read the same way 

A written rule constrains anything only if its terms mean the same thing on a bad afternoon as they did on the evening it was written. Phrases like sensible size, a strong trend, decent volatility and a clear level read as content and specify nothing, because the person applying them is also the person who decides what they meant. The conventional test is whether two readers with the same chart and the same page would produce the same answer. Few plans pass it completely, which is an argument about degree rather than a reason to drop the test.

The gap between the two readings is the size of the discretion the sentence quietly left in place, and discretion left in a document is exercised in the moment rather than at the desk where the document was written. This is why plans are conventionally written in numbers, levels, times and named conditions, and why the components in the previous section are stated as quantities wherever a quantity exists.

Why writing it down does anything 

A rule held in memory has a property a written one does not. It can be restated. The restatement is rarely deliberate and almost never feels like one: a rule recalled while a position is open is recalled by someone who already knows which way that position has gone, and recollection is obliging under those conditions. The rule that emerges is the rule that fits, and there is nothing to compare it against, because the earlier version exists only in the same memory that has just revised it.

Writing removes that particular escape and only that one. A written rule was fixed by someone who did not know how the position would turn out, and that ignorance is the qualification rather than a defect: the author of the document had no interest in the outcome of any specific position, because no specific position existed. When the written version and the conduct disagree, the disagreement is visible, dated and available to be examined afterwards. It is the same device that a stopping rule uses, applied to the whole of a method rather than to one quantity.

The limits are worth stating with equal force, because a written plan attracts more credit than it earns. Writing a rule down says nothing whatever about whether the rule is any good, and a carefully specified method can lose money carefully. A document can also be disregarded exactly as easily as a remembered rule can be revised, the difference being that the disregard leaves a trace. The pressure the writing is meant to answer is strongest at the moment the document is least likely to be opened, which is why the record and review components are there rather than being administrative decoration.

A plan, a prediction and a system 

A prediction is a claim about the market: that an instrument will reach a level, that a currency will weaken, that a range will hold. A plan contains no such claim. Its sentences are conditional and their subject is conduct rather than price. If these conditions are observed, a position of this size exists, with invalidation here and an exit there. Nothing in it asserts that the conditions will occur, or how often, or what follows if they do. The two are easy to confuse because both are written in advance, and one test separates them: a prediction can be wrong about the market, while a plan can only be departed from.

A system is a narrower object than a plan and sits inside it. It is the part that could in principle be executed by somebody else, or by a machine, with no question asked: the condition for opening, the size derivation, the invalidation, the exits. Everything a machine has no use for and a person cannot do without belongs to the plan rather than to the system, including the hours a human being is fit to work, what happens after a run of adverse results, what is recorded, and when the document is read again. A system can be examined against historical data, which is the subject of the lesson that follows this one. The conduct wrapped around it cannot be examined that way, which is why a tested system and a followed plan are separate accomplishments.

Revision without abandonment 

A document that can never change is one that eventually gets ignored rather than amended, so revision is a component of a plan rather than a failure of one. What separates a revision from an abandonment is not the size of the change. It is when the change was made and what prompted it. A revision is a dated amendment made at an occasion the document itself nominated, with the reason recorded beside it and the earlier version kept. An abandonment is a change made while a position is open, prompted by that position, and it is recognisable because the amended rule permits exactly what that position needs.

Several conventions exist for the occasion itself, and each carries a cost its own advocates acknowledge. A fixed interval, such as the end of a month or a quarter, is simple and unarguable, and it leaves a rule that has stopped matching conditions in force until the calendar permits otherwise. A stated number of completed positions ties revision to evidence rather than to the calendar, and it can be reached in a week or not for half a year. Changing one term at a time makes the effect of a change legible, and it is slow, and it hides the interaction between two terms that only misbehave together. None of these is put forward here as the right arrangement.

The failure runs in both directions, which is what makes the subject difficult rather than merely procedural. A plan amended after every adverse result is a running commentary on recent results, and it never accumulates enough evidence about any one version of itself to say anything. A plan never amended at all outlives the conditions it was written for, and what usually follows is not a considered replacement but a quiet drift in which the document stays on file and the conduct stops resembling it. Producing the evidence a revision would need, so that it is something other than an opinion about the last few weeks, is what the record component is for and what the journal lesson later in this module examines.

Where practitioners disagree 

The first argument is how mechanical the document has to be. One tradition holds that any term left to judgement will be exercised inconsistently, and that a plan is worth what its most ambiguous sentence is worth. Another holds that a fully mechanical document is a system with a filing cabinet around it, that discretionary traders work from plans specifying conditions rather than triggers, and that over specification produces a document nobody follows on the days that matter. Both objections describe real failures, and neither tradition has evidence that would settle the question for somebody else's method.

The second concerns length. The argument for a single page is that a document consulted under pressure is only useful if it can be read under pressure, and that a manual is a manual precisely because nobody opens it. The argument against is that the components above do not fit on a page without the compression that produces the ambiguous phrases described earlier, and that omitted detail does not disappear. It gets supplied in the moment instead. A common compromise is a short operating page in front of a longer document, which relocates the argument to which parts earn the front page.

The third is whether a plan written before any experience is worth writing. One view is that a first plan is mostly borrowed assumptions, and that its function is to make those assumptions explicit enough to be found wrong. The opposing view is that a document written before a trader knows how they behave under a loss specifies the wrong things, and that early rules acquire an authority they never earned simply by having been written first. Both readings can be held at once, which is roughly what the revision conventions above try to accommodate.