The behavioural failure modes

What the record shows 

A record of closed positions is a sequence of timestamps, sizes, instruments and stated reasons. Read down the timestamps and the afternoon of a bad session looks different from the morning: the gaps between entries shorten, the sizes stop matching the rule written at the top of the plan, and the instrument column repeats one name. None of that is a psychological interpretation. It is what the rows say.

That is the first useful fact about the behavioural failure modes. They are not moods, and they are not found by introspection. Each is a departure from a rule that was written down, which makes it a fact about a record rather than a judgement about a person. The names below are conventional labels drawn from behavioural economics and from trading literature, and they are useful mainly because each says where in the record to look. This page makes no claim about how common any of them is.

Four are described here, and they overlap heavily. A single afternoon can contain all of them.

Activity that has come loose from the method 

Overtrading describes activity whose rate is no longer set by the conditions a method requires. A method states when a position is taken. If the count of positions rises while the stated conditions occur no more often than before, the extra positions came from somewhere other than the method, and the usual place is the criterion itself, quietly widened until more things qualify.

In a record that shows up as a count rising without a corresponding change in what the instrument is doing, and as entries that cannot be matched to any written criterion. The rise is not confined to losing sessions: the literature describes activity rising after a run of favourable results as readily as after adverse ones, on the reasoning that a run of either kind changes how the next decision feels rather than what the method says.

The arithmetic that makes frequency expensive was taught in the costs module and does not change here. Cost is charged per position, in the spread and in any commission, and it is charged whether the position closes favourably or adversely. The count is therefore a direct multiplier on the cost of a session, and it is the one term in that product a trader sets exactly.

The fair objection is that additional positions carry their own gross results, so a method taken more often is not automatically a method taken worse. What the block shows is narrower: whatever the gross column does, the cost column grows with the count, and in one direction only.

The position that is about the last position 

Revenge trading describes a position whose reason is the result of the previous one. Its shape in a record is unmistakable. The interval between the close of an adverse position and the opening of the next is shorter than the interval anywhere else in the record. The size is larger than the sizing rule permits for that stop distance. The instrument is frequently the same one that produced the loss, and the direction is frequently the opposite of the one that just failed.

The mechanism conventionally offered comes from prospect theory, the account of decision making under risk set out by Kahneman and Tversky, in which outcomes are evaluated as gains and losses against a reference point rather than as levels of wealth, and a decision maker already behind that point decides differently from one ahead of it. Applied to a session, the reference point is the balance the session opened with, and each decision after it is framed as a distance from that number. The instrument the decision is nominally about does not appear in that framing at all.

The escalation that follows is symmetrical in a way the feeling behind it is not. Doubling size after each adverse result shortens the move a single favourable result needs in order to bring a running total back to where it started, and it doubles the amount at stake on each successive position, so an adverse run grows at the same rate. A loss produced this way is not limited to the amount at stake on the first position, or to the amount deposited.

The two sequences are the same four decisions differing only in size, and the doubling one moves both ends of the range by the same factor. That symmetry is what the framing hides. A run of adverse results is not evidence that the next result will differ, and the size of a position carries no information about the direction of the next move.

Adverse and favourable positions treated differently 

Loss aversion describes the finding, from the same literature, that a loss of a given size registers more strongly than a gain of the same size. In a trading record it produces two symptoms that look like opposites and come from one source: adverse positions held past the level at which they were meant to close, and favourable positions closed before the level at which they were meant to close.

Both are the same move. Closing an adverse position converts an unrealised loss into a realised one, and closing a favourable position converts an unrealised gain into a realised one, while holding either keeps its question open. Whichever action postpones the outcome that would register more strongly is the one taken, and because the two cases are mirror images, the postponement runs in opposite directions. The trading literature calls the pair the disposition effect.

It leaves the clearest fingerprint of the four, because reading it needs no interpretation. The holding time of adverse positions is compared with the holding time of favourable ones, and the realised result of each adverse position with the result the exit level stated before entry would have produced. A gap in either comparison is arithmetic. It is also not the only explanation for one, and the lesson on why stops get hit set out the others: a market that gaps past a level, slippage in fast conditions, a close at a worse level than the one specified.

Both halves are departures from a level written before the position existed, and they point opposite ways, which is why a record can show a long run of small favourable results and still fall behind. Neither half argues that the stated levels were correct. A level decided in advance and a level decided while a position is open are simply two different quantities.

Evidence read in one direction 

Confirmation bias describes the tendency to seek and weight evidence agreeing with a view already held, and to discount evidence that does not. In trading it operates at three points, and only the third leaves a mark that can be audited later.

• Before a position, while the case for it is assembled. The chart is examined on the timeframe that agrees, and the instrument is searched for a pattern already expected rather than examined for what is on it.
• While a position is open and its running result is unfavourable. Contradicting information is reclassified as noise, and being wrong is reclassified as being early.
• After a position closes, in the record itself. Favourable positions are reviewed and adverse ones skipped, or the note explaining an entry is written once the result is already known.

The third matters most here, because it corrupts the instrument the other two are found with. A note written before a position opens records a reason. A note written afterwards is a reconstruction with access to the outcome, so it tends to arrive at a reason consistent with it. That is why the journal lesson made the ordering a rule: the field recording why a position was taken is completed before the result exists.

Why size and frequency change what a departure costs 

None of the four is peculiar to trading, and each is described in settings with no market in them at all. What a leveraged instrument changes is not how likely a departure is, but what follows from one. Profit and loss on a CFD are calculated on the full notional value of the contract while only a percentage of that value is posted as margin, so a position opened outside the sizing rule is measured against the whole contract and not against the money behind it. A loss on that basis can exhaust the margin posted and is not limited to the amount deposited, and a favourable move is measured on exactly the same basis and to exactly the same degree.

Frequency compounds the same effect by a different route. Each of the four is a departure from a rule, and the number of chances to depart from a rule is the number of decisions taken. A method producing few decisions leaves a longer interval between them, during which the state that produced one can pass. A method producing many leaves intervals short enough for the state to carry from one decision to the next. That is a statement about counts, not a ranking of methods: the styles lesson set out what each demands and did not rank them either.

Why a rule written in advance is not the same as an intention 

An intention formed while a position is open is formed by someone with money riding on the answer. A rule written on a quiet evening was written by someone with none. Those are not the same author, and the difference between them is the entire mechanism.

A pre commitment is a decision taken at one time whose function is to remove an option at a later one. Its defining property is that it binds when it is least welcome, which describes how it works rather than a defect in it, because a constraint that applies only when convenient constrains nothing. The loss limit lesson described one instance, and the plan lesson the document the rest live in. What makes any of them enforceable is the same four things.

1. A trigger stated as an observable condition, so that whether it has occurred is a question of fact rather than a matter of judgement at the worst moment to hold one.
2. An action stated as a specific thing that happens when the trigger occurs, since a trigger with no stated action is noticed rather than enforced.
3. A record made at the moment the trigger occurs, because a rule reconstructed afterwards is reconstructed by somebody who knows how it turned out.
4. An enforcement route that does not run through the state the rule exists to constrain: a platform setting, a terminal that is closed, a record another person sees.

The limits of the device matter as much as the device. A pre commitment can be revoked by whoever wrote it, and the moment it binds is precisely the moment revoking it sounds most reasonable, in language resembling analysis rather than a change of mind. Suspending it for this one instrument, applying it from the following session, treating the open position as unrelated to the ones that produced the total: none of those arguments was available before the session began, and all arrive together. A rule is only as good as the distance between the person it constrains and the switch that turns it off.

What a record surfaces before an account does 

An account balance is a lagging and aggregated signal. It nets every decision in a period into a single figure, so a period in which rules were departed from repeatedly and one in which they were followed can end at the same balance. A journal reports one decision at a time, and it reports at the time. It surfaces the patterns above in four comparisons, each between two things already written down.

• The timestamp of each entry against the timestamp of the previous close, alongside the instrument and direction, so that a shortening interval and a repeated name after a loss read as a pattern rather than as unrelated ideas.
• The size of each position against the size the written sizing rule produces for that stop distance. The comparison is arithmetic already taught in this module.
• Whether an entry matched a written criterion, recorded as a criterion rather than as a sentence of reasoning, because a criterion can be checked and a sentence can be rewritten.
• The exit level stated before a position opened against the level at which it actually closed, with the reason for any difference recorded at the time.

None of those is a judgement about a trader, and none requires a view about why anything happened. Each compares a rule stated beforehand with a fact recorded afterwards, and each is available long before the balance has moved far enough for anyone to notice.

Where practitioners disagree 

The first argument is whether these patterns are dispositions or situations. One tradition treats them as habits to be trained out, through review, rehearsal and attention to the state a decision is taken in. Another argues that training is applied by the same person, in the same state, that it is meant to survive, and that what changes behaviour is changing what is available: the rule, the platform setting, the closed terminal. Each position names the other's weakness, since a constraint is also written and enforced by the person it constrains.

The second is whether automation removes the problem. A mechanical system executes without a state, so none of the four has a route into an individual decision. Critics observe that a person switches the system on and off and chooses its parameters, and that adjusting a parameter after an adverse run is the same decision as the escalation above with a configuration screen in front of it. Confirmation bias is arguably easier to exercise on a system than on a position, because a parameter can be adjusted until a record of past prices agrees with it, which is the failure the limits of testing lesson described.

The third is about frequency itself, and it is why the word overtrading is contested. A count that indicates a departure from one method is the ordinary operating rate of another, so no count is diagnostic on its own. Practitioners working at high frequency argue that the meaningful question is whether each position met a stated criterion rather than how many there were. The other side answers that the criterion is exactly the thing that loosens, and that a rising count is the first observable sign of it. Both describe something real, which is why the argument does not resolve.