How a trading journal is kept

Two records, written at two moments 

The physical form is unremarkable and deliberately so. A row in a spreadsheet, a page in a notebook, a plain text file per position: the medium carries none of the weight, and no tool is required that a reader does not already have. What carries the weight is that each position produces two entries rather than one, and that the first entry is closed off before the instruction to open is sent. A record written in two sittings, with the market's answer arriving in between, is a different object from a record written in one sitting afterwards. The first can be wrong. The second cannot, because it was composed already knowing what happened.

This is why the timestamp on each half matters more than any field inside it. Memory of a decision is not stored and retrieved. It is reconstructed, and it is reconstructed in the light of everything that has happened since, including the result. A reason recalled after a favourable close arrives sounding clearer and more deliberate than the reason that was actually held; a reason recalled after an adverse close arrives sounding careless, and often was not. Neither recollection is dishonest. Both are simply unavailable as evidence, and writing the reason down first is the only mechanism that makes it available at all.

What is written before the position exists 

The pre trade half records the decision, not the market. Its fields are the ones a plan of the kind described earlier in this module already settles, transcribed for this particular instance, and most of them take a few seconds each because the plan has already done the deciding.

• The instrument, and the condition in the plan that the instrument is currently satisfying.
• The level at which the position is intended to open, written as a level rather than as a description.
• The level at which the reason for holding the position would have stopped applying, and the distance between that level and the intended opening price.
• The fraction of equity the plan permits a single position to place at risk, the money amount that fraction produces, and the size derived from that amount and the distance above.
• The exit arrangement as it stands before anything moves: the level or condition on the favourable side, and whether any part of the size is intended to close early.
• Aggregate exposure at that moment, meaning what else is already open and what the total committed risk becomes once this position is added.
• The reason, in one sentence, in ordinary language, written as though to somebody who will read it without seeing the chart.
• The time, and the session the market is in when the instruction is sent.

Practitioners who keep both halves generally describe this one as the harder to maintain and the one carrying whatever value the exercise has, and the two facts are connected. It is harder because it is written at the only moment in a position's life when attention is fully occupied elsewhere, and because it takes its time before anything has happened rather than after everything has. It is more useful because it is the only half that can turn out to be wrong. A field that can be contradicted later is evidence. A field composed in the knowledge of its own answer is a summary.

The one sentence reason is the field most often skipped and the one that repays the seconds most. Written in advance it is short, sometimes awkward, and occasionally reveals in the writing that there is no reason of the kind the plan names, only a wish to be in the market. Read back a month later alongside eleven others, the sentences group themselves without any analysis being applied to them, and the groups are what a review has to work with. A journal of levels and money amounts records what was done. A journal that also records why records something no platform statement contains.

What is written once it has closed 

The second half records what happened, and most of it can be copied from the platform's own history rather than remembered. The fields worth keeping are the ones that can disagree with the first half.

• The price at which the position actually opened, beside the level the first half intended, so that any difference between them is visible rather than absorbed.
• The price at which it closed, and which instruction closed it: the resting stop, the intended exit, or a manual decision taken while it was open.
• The distance actually travelled in each direction, and the costs charged, spread and commission and any financing adjustment for a position held past the daily cut off.
• The result, as an amount and as a multiple of the amount the first half put at risk.
• Whether the conduct matched the pre trade record, recorded as a plain yes or no before any explanation is attached to it.
• Anything that occurred while the position was open which the first half did not anticipate, stated as an event rather than as an interpretation.

The gap between the intended opening level and the price actually received is worth its own column rather than being folded into the result, because it accumulates quietly and belongs to execution rather than to the decision. The mechanics of that difference, and the conditions under which it widens, are set out in the guide to execution quality.

Two things are conventionally kept out of this half. The first is a verdict on whether the position was a good one, which belongs to the review rather than to the record and which cannot be reached from a single result anyway. The second is an explanation of the market's behaviour, which is available in unlimited quantity after the fact and constrains nothing. A record that says the level was broken by a release at the London open is a note. A record that says the position departed from the plan by half of its permitted size is a fact that can be counted.

How a set of records is examined 

A review reads a set of records, never a single one, and it happens on an occasion fixed in advance rather than whenever the last result makes it feel necessary. The written plan discussed earlier in this module nominates that occasion for exactly this reason. A review held immediately after an adverse result is not a review; it is a reaction with a spreadsheet open, and the amendment it produces will be the amendment that particular result argues for.

The sequence practitioners describe is ordinary, and the ordering is the part that does the work: everything countable is counted before anything is interpreted.

1. Every record from the period is assembled, including the positions that were considered, recorded and then not opened, which are otherwise invisible.
2. The records in which the conduct matched the pre trade record are counted against the records in which it did not, before a word of the reasons has been read.
3. The departures are sorted by what was departed from: size, the invalidation level, the entry condition, the exit, or the aggregate limit across everything open.
4. The one sentence reasons are read as a group, looking only for repetition, since a phrase appearing in nine records out of twelve is describing something the plan has not named.
5. The conditions surrounding the departures are noted, meaning the session, the time of day, the sequence of preceding results and what else was open at the time.
6. The review itself is recorded as a dated entry, with any amendment to the plan written beside its reason and the previous version kept.

What a review produces is a description of conduct over a period, and it is worth being clear that this is all it produces. It says nothing about whether the plan is any good. A run of records showing conduct matching the plan in every instance establishes that the plan was followed and nothing further, and the two testing lessons before this one set out why the number of records a private trader accumulates in a year remains far too small to establish anything about a method regardless of how carefully each one was written.

A process error and an adverse outcome are different objects 

This is the distinction the two halves exist to make, and it is not available from a platform statement, which records only results. A process error is a departure from what was written down before the position existed: a size other than the derived one, an entry taken without the stated condition, an invalidation level moved while the position was open, an aggregate limit exceeded. It is a fact about conduct, it is established by comparing two documents, and it is entirely independent of how the position turned out.

Sorting a set of records by those two questions at once produces four groups rather than two, and the two mixed groups are the ones a results-only record cannot see. A position that followed the record and settled at a debit is the ordinary case, not a fault, since a method whose adverse cases were avoidable would not need a stop. A position that departed from the record and settled at a credit is a process error that happens to have been rewarded, and it is the single most misleading entry any journal contains.

The tendency to judge a decision by its result rather than by the information available when it was taken has a name in the psychology literature, and it is the reason the four groups are worth separating at all. It operates in both directions and it operates silently: a departure that was rewarded is remembered as judgement rather than as a departure, and an adherence that settled at a debit is remembered as an error to be corrected. Correcting it is how a plan quietly becomes a commentary on recent results. The pre trade half is the only defence available, because it fixes the reasoning in writing at a moment when the result cannot yet influence it.

Where practitioners disagree 

The first argument is about volume. One tradition records a handful of fields per position on the grounds that a record kept for years beats a thorough record abandoned in a fortnight, and that most fields are never read again. Another records everything available, including chart images at entry and exit, on the grounds that a question occurring next year cannot be answered from fields chosen this year. The criticism each makes of the other is accurate. Thin records lose the detail that would have explained a pattern once one appears; thick records take long enough per position that they are the first thing dropped in a busy week, and a journal with a gap in it cannot be counted across that gap.

The second argument is whether a state of mind belongs in the record at all. One tradition records it in a fixed vocabulary, on the grounds that departures cluster around states that are recognisable in advance and that a record naming them is the only way that clustering ever becomes visible. Another declines to record it, on two grounds worth taking seriously: the description is written by the same person whose state it describes, so it is the one field in the document that nothing else can be checked against, and a journal that becomes an inventory of personal shortcomings stops being opened, at which point it records nothing at all.

The third argument is what a journal is entitled to conclude, and it is the one where the disagreement is sharpest. One position says the record exists solely to measure adherence to a plan, and that any statement about the plan itself has to come from testing rather than from a few dozen records. Another says a plan meeting conditions it was never written for shows up in the record long before any test could establish it, and that refusing to read the record that way discards the earliest information available. Both are describing the same difficulty from opposite sides: a small set of records is too small to support a conclusion about a method, and it is also the only evidence of that method's contact with a live market that anyone actually has.