Reset protocols: halting trading and the conditions for returning
After a run of losses or gains, it is easy to make another decision simply to recover the result or make the most of a winning streak. A reset protocol helps you decide in advance when to stop opening new positions, what to review during the pause and what must happen before you resume trading. A resolution such as “I will return when I have calmed down” is insufficient: you need rules you can follow and subsequently check. This article covers the three phases of a reset, the record in your trading journal (10.8) and signs that the procedure needs changing. It also distinguishes an ordinary losing streak from breaches of the plan, and your subjective state from verifiable return criteria.

Name the event that activates the procedure
A reset starts with an observable event, rather than a promise to feel differently. To audit it, keep the rule that existed before the event and the record that activated it. Otherwise the trigger can quietly change after a profitable trade, a loss or an uncomfortable comparison.
In this hypothetical example, the written plan says that an unapproved increase above the recorded calculator quantity triggers a halt on new entries. At 12:04:10 UTC, a comparison shows an order quantity of 0.18 lot against the recorded 0.12 lot. The order is 1.5 times the recorded quantity, or 50% larger. That identifies a specific mismatch under this rule. It does not establish the trader’s emotional state or diagnose its cause.
The halt is declared at 12:04:12. Keep the earlier activation time and the later response time as separate fields. The quantities here demonstrate a comparison, not a recommended position size or a universal threshold for taking a break.
Check every route that can still create an entry
A desktop button marked “paused” does not establish what a mobile terminal, another device or an automated process can do. The example uses separate controls with recorded confirmations; real services may offer different controls. Do not assume a restriction exists until its scope and behaviour have been checked.
| Entry route | Restriction confirmed at |
|---|---|
| Desktop terminal | 12:04:14 |
| Mobile terminal | 12:04:25 |
| Automated entry process | 12:04:33 |
The record supports coverage of all listed routes only from 12:04:33. It says nothing about an unlisted route. Existing exposure, protective exits and previously submitted pending orders require their own handling under the risk plan. Blocking new entries is not an instruction to remove protective orders or abandon an open position.
Keep one inventory of the routes actually in use and another of outstanding orders. A channel restriction and an order cancellation answer different questions. A request to cancel a pending entry does not prove cancellation; a later fill may need to be reconciled with that request and its acknowledgement.
Count new decisions separately from later fills
Consider the following continuation. An entry submitted at 12:03:50 fills at 12:04:17. A new entry request is submitted from the mobile terminal at 12:04:20. A permitted risk-reducing exit is requested at 12:04:28. Assume the example’s rule permits managing existing exposure during the halt and prohibits new entry requests after the trigger.
There is one prohibited new entry request in that record. There are also two entry fills if the mobile request completes, but the earlier pending entry was submitted before the trigger. Its later execution still changes exposure and must be handled; it is not evidence of a new post-trigger decision. The permitted exit does not become a breach merely because it happened during the halt.
Use separate columns for request time, action type, order identity, fill time and confirmed quantity. A list of completed trades alone cannot reconstruct every decision. Rejecting or cancelling a request also belongs in the record, even when it produces no completed trade.
Make return depend on the missing condition, not the elapsed time
Suppose the prewritten return rule requires verified channel coverage, a reconciled order inventory and a checked next-entry calculation. At the last channel confirmation, the order inventory still contains an unresolved cancellation. Time passing does not complete that missing field. Under this particular procedure, return has not yet been authorised.
When the missing state is resolved, retain its confirmation and rerun any calculation affected by a fill or cost. A fresh entry must satisfy the trading method and risk plan independently. Completing a reset does not establish an edge or turn the next trade into a good opportunity.
Review the procedure itself after the episode. The mobile entry shows that the written halt and actual coverage differed during the transition. Record the gap and whether the chosen safeguard addressed it in later observations. This is a check of execution against a stated rule, not proof that a pause caused improved financial results. The detailed lesson below develops trigger design, relapse branches and the boundary of what a trading procedure can address.
Three exercises with worked answers
The exercise data are hypothetical. Choose an answer, then check the calculation.
From an auditable halt to a complete return procedure
The full lesson retains the original protocol illustrations, counterexample, research discussion and relapse branch. Use them to write trigger, action, return and failure conditions before an episode occurs.
1. What the protocol is meant to stop – and why not at the moment it is needed
A reset protocol primarily concerns orders placed after a process breach. Its purpose is to limit new risk rather than certify that the trader now feels well. You can record your subjective state in the journal as a separate observation, but it should not be the sole basis for returning. Order history, recorded conditions and event timestamps let you check whether the procedure was followed. Persistent health difficulties require a separate assessment, discussed at the end of this article.
Recognising the episode itself – the list of observable violations, the escalation chain and the moment the sequence begins – is the subject of article 10.5 and will not be repeated here. This article starts one step further on, at the question: given that the process is already known to have changed, what exactly should happen then, and how do you tell that you are allowed to come back.
The protocol has a design rationale that can be explained without invoking unmeasured biological mechanisms. People planning their own behaviour in a cool state systematically underestimate how they will behave in an aroused state — the phenomenon described as the hot–cold empathy gap[1]. Loewenstein’s work concerned mainly medical decisions and health behaviour; there were no traders among the participants. Applying this finding to trading is an extrapolation. The design implication is this: you may execute a plan written on Sunday in a different state on Wednesday. It should therefore allow for the possibility that following it will be harder when the trigger occurs.
It helps to state what this argument does not rely on. A popular explanation for protocols invokes willpower being depleted as a limited resource[5]. A preregistered replication of a specific paradigm across 23 laboratories (N = 2,141) found a very small effect, d = 0.04, with a 95% confidence interval from −0.07 to 0.15[6]. It did not confirm the effect in that task; it does not settle every study of self-control or the effects of fatigue. A protocol does not require a “depleted resource” hypothesis. Its design rationale is simpler: record the decision and safeguards in advance, rather than during the episode.
A second, less obvious thing follows. Since the protocol has to work precisely when nobody feels like executing it, it cannot require an assessment of the situation. Every field that reads "assess whether you are ready" moves the decision back to exactly where we wanted to take it from. That criterion runs through the whole article and is the only test I apply to every element of the procedure: can it be settled that the condition has been met, without asking yourself how you feel?
2. The shape of the procedure: four fields instead of a resolution
The procedure’s form — trigger condition, action, end criterion and audit trail — is defined in article 10.7. That article explains its design principles. The reset protocol uses the same four fields. Record a separate failure criterion: how you will recognise that the procedure needs changing. This requirement applies to every rule described in section 10, including those in article 10.7. It is particularly important when halting and resuming trading.
| Element | What it must contain | Correctness test |
|---|---|---|
| Trigger condition | an event readable from the record: a violation from the list, a limit from the plan being exceeded, the detection of an order sized by the amount to be won back | two people looking at the same history export arrive at the same answer |
| Action | what exactly stops being possible, and on which device | it can be carried out without making any decision |
| End criterion | the event that ends the block; not a feeling of readiness alone | decidable without self-assessment |
| Audit trail | what gets recorded: the trigger time, the number of orders after it, the list of violations | it can be compared with the previous episode |
| Failure criterion (a requirement separate from the form) | the observation after which you know the procedure is not changing behaviour | defined before it is needed |
The last row is not a field of the form, and that is why it stands apart: the four fields describe the execution of the procedure, while the failure criterion settles whether it should continue to exist. Without it there is no way to state whether the protocol works. A procedure with no failure criterion cannot be rejected: every run of it can be described as a success, because you can always say things would have been worse without it. A rule that cannot be refuted by your own data is not a rule, it is a belief.
3. A trigger condition written before the episode
The trigger condition is the only element of the protocol that has to fire at the moment the trader is least willing to acknowledge it. That is why its form matters more than its value. The form that has evidence behind it is well known: a plan in the shape "if an observable condition occurs, then I carry out a specific action" performs distinctly better in a meta-analysis of 94 independent tests than a general resolution, with an effect size of d ≈ 0.65[2]. Those studies mostly concerned health and academic goals; trading was not among them – this is an extrapolation, and the design principles for such plans are developed in article 10.7. Below is the same intention written once badly and once well.
| Undecidable wording | Decidable wording | What changed |
|---|---|---|
| "I will stop when I feel I am losing control" | "After the second order whose position size is larger than the calculator result for my recorded risk, I place no further orders until the end of the session" | a condition readable from the history export instead of from a state |
| "I will not revenge trade" | "An order whose size I computed so as to cover a specific earlier loss triggers the block – regardless of whether it was actually placed" | a defined behaviour instead of an intention |
| "I will stop at a larger drawdown" | "Exceeding the daily loss limit recorded in my risk management plan ends the session" | a reference to a value set elsewhere, instead of a number invented on the spot |
| "After a losing streak I take a break" | "After N consecutive losses, where N comes from my own record and is higher than the typical streak of my strategy, I block further orders for a time T" | a parameter explicitly marked as to-be-set, with a rule for setting it |
One caveat about choosing N does belong here, though, because it concerns how the procedure itself behaves. A trigger condition set at a value that is typical for your strategy will fire constantly, and a procedure that fires constantly stops being taken seriously by the trader – it gets bypassed reflexively, like an alarm that goes off for no reason. The calculation that lets you check what is typical for you is in the next section.
4. A counterexample: a streak that is not an episode
Without this section the whole protocol works in the wrong direction: it fires on every rough patch, and the trader learns to work around it. So we need to show a sequence that looks from the outside like account destruction but is not one – and point to how the two are told apart.
Run A (illustrative values). Seven trades across four sessions, six of them losing, a drawdown visible on the equity curve. On each of the seven: an entry condition from a list written before the session, size equal to the calculator result for the recorded risk, a protective order placed before confirmation, an instrument from the list, a gap from the previous close above the trader's own median, and a journal entry made before the result was known.
Run B (illustrative values). Seven trades within a single session, six losing, a similar drawdown. On the first three, everything matches the record. On the fourth, position size is larger than the calculator result. On the fifth, additionally, an instrument outside the list. On the sixth and seventh, no protective order and no entry before the decision.
Both runs have the same outcome. Only the second is an episode. The difference is not visible in the equity curve and cannot be established afterwards from memory. It is visible in the layer of the record drawn up before each decision – and that is the only reason the journal from article 10.8 is a working tool here.
To this we should add a calculation worth doing once, before setting the trigger condition. At a 55% win rate, that is a loss probability of 0.45, a run of at least six consecutive losses over a year containing 250 trades has a probability of:
P(run ≥ 6 losses in 250 trades) = 68.6%
The owner of this calculation is article 10.1, where it is discussed in full together with the consequences for evaluating a strategy; here the number is needed only as a reference point when choosing the trigger condition. The method in one sentence: dynamic programming over states given by the length of the current run, assuming independent trials and a constant win rate; the result was checked by independently recalculating that recurrence. The result depends on the horizon, and that is the most important caveat here: the same run over 200 trades has a probability of 60.2%, over 250 – 68.6%. A run probability quoted without the number of trades it was computed over means nothing.
The second caveat concerns the question this number answers. A run specified in advance, in one particular set of six trades, has a probability of 0.456 = 0.83% – and that is the value intuition offers first. 0.83% and 68.6% are not two versions of the same number; they are answers to two different questions.
The implication for a trigger is direct. Under the stated assumptions, six consecutive losses do not prove a process breach: they occur in most modelled sequences of 250 trades. A behavioural protocol should therefore respond to breaches rather than treat a loss itself as an error. Independently established risk limits still apply: exceeding them may require a halt even when every trade followed the plan. These are two different reasons to stop.
5. Phase one: the halt
Phase one has a single task: to make sure the next order opening a position or increasing risk cannot be placed. Not that the trader resolves not to place it.
- Trigger condition: an event from section three, readable from the record.
- Action: logging out and removing the ability to log back in quickly on every device from which an order can be placed. Not on one – on every one, because the first reflex after the terminal is blocked is to reach for the phone.
- End criterion: the elapsing of the time T recorded beforehand. Not "when I feel I can".
- Audit trail: the trigger time and the number of orders placed after that time. Zero is the only value here that means the phase was carried out.
- Failure criterion (a requirement separate from the form): the record shows even one order after the trigger time, or shows a timestamp of a change to the safeguard configuration between two orders.
The last point matters more than it looks. A safeguard removed mid-session has not failed – it has become a workaround, that is, another element of the same sequence, executed one level up. If that repeats, the problem is not the choice of tool and a harder block will not solve it; I return to this in section eleven.
6. Phase two: the pause and the limits of "reset" techniques
The error in this phase is not that the pause is too short. It is that throughout its whole length nothing changes except that the trader is not trading. A pause after which you return to exactly the same set of rules, the same environment and the same trigger condition is not a reset – it is an interval. A similar episode may recur; the pause alone does not establish that the problem has been resolved.
The pause therefore has content, and its content is a review of the record. Not reliving the session, because that leads to the rumination article 10.8 warns against, but four questions put to the order history export:
- At which order did the first violation appear – not the first loss?
- Had the trigger condition already been met at that point? If so, why did the procedure not fire?
- How many orders came into being after that moment?
- Is there, among them, a timestamp of a settings change, a transfer or a login from another device?
The answers are written into the journal and stay there. Without that, the next episode has nothing to be compared against, and without a comparison no statement about the protocol's effectiveness has any basis.
Phase two written in four fields, the same as phase one:
- Trigger condition: the closing of phase one, that is, the moment the block is already in place on all devices.
- Action: a review of the export following the four questions above, with the answers written into the journal. Trading stays blocked for the whole of that time.
- End criterion: four answers recorded, plus the elapsing of the time T set before the episode. Both conditions at once – time alone, without recorded answers, ends the pause only in appearance.
- Audit trail: the date of the pause, the four answers, and a note of what was changed between episodes: the trigger condition, the location of the barrier, or nothing.
The failure criterion (a requirement separate from the form) for this phase is in the box closing the section: the pause is taken in full, and the number of violations in the next episode stays the same.
What is not known about "reset" techniques
A fair number of practices have grown up around this phase, and in various write-ups they are presented as a way of restoring the initial state: breathing exercises, physical activity, a cold shower, meditation, a walk, a conversation with someone outside the market. There is no reason to forbid them, and some of them have evidence in other applications. There are, however, three things that are not known about them, and it is honest to list them.
First, the literature cited here does not establish how these activities affect retail traders’ decision quality. Nor does it support claims that “ninety seconds of breathing resets the nervous system” or that a pause must last a particular number of hours because hormone levels need that time to return to normal. The measurements discussed here link hormones and behaviour; they do not establish a pause duration for trading or other financial decisions. Deriving such a duration requires additional justification and measurement. A precise number is not evidence: adding a number of hours to this explanation leaves it unsupported.
Second, their role in the procedure is different from the one those write-ups assign them. Within this protocol, their role is to occupy time away from the place where orders are submitted; any change in how you feel is a separate observation. That is reason enough to use them, and it requires no biology at all.
Third, they have a cost nobody mentions: a technique described as effective starts to replace the end condition. "I did the exercise, so I can go back" is exactly the same sentence as "I have calmed down already", only with a prop. The return criterion cannot be whatever the trader did during the pause.
7. Phase three: the return and its criteria
The return condition is written together with the halt condition, at the same moment and on the same piece of paper. Written later, it is created under conditions in which all the parameters are known least well. A rule that ends at the halt is half a rule – article 9.8 says so outright, and it owns the resumption condition on the numbers side.
On the behaviour side one principle applies: the return criterion must be decidable without relying solely on an assessment of your own state. Below are four criteria that pass that test and four that do not.
| Decidable criterion | Apparent criterion |
|---|---|
| The time T recorded before the episode has elapsed | "I feel that I am ready" |
| The review of the episode record is done and written down, with answers to the four questions from section six | "I have thought through what went wrong" |
| The list that settles the sentence "trade consistent with the plan" is written out before the return: a condition from the list, size equal to the calculator result, a protective order before confirmation, an entry before the decision | "My recent trades were consistent with the plan" – with no such list and no check against the export |
| Exposure on the return is smaller than usual, according to a rule recorded beforehand | "I will come back more carefully" |
The third row concerns whether a trade followed the plan. A “yes” or “no” answer requires a list of conditions documented beforehand. The table gives four items: the entry condition checked against the pre-session list; position size checked against the calculator result; a protective order before confirmation; and a journal entry timestamped before the order. Each can be checked in the data. Before returning, you can establish whether the list is ready. Only after returning can you check whether subsequent trades met its conditions. A full set of four conditions across the next M trades is therefore a criterion for the next stage — moving from reduced to normal exposure, as described below. It cannot determine the first resumption because those trades have not yet occurred. Breach codes and process metrics are explained in article 10.8; here we use them as an existing tool.
Phase three in the same form as the two before it:
- Trigger condition: the full set of criteria from the left-hand column of the table – the time T has elapsed, the episode review is recorded, the list that settles plan consistency is written out, and the reduced-exposure rule was recorded before the return. In this phase the same element is sometimes called the resumption condition; the field of the form is identical.
- Action: resuming trading with exposure smaller than usual, according to the rule recorded together with the halt condition.
- End criterion: a full set of four consistency conditions across the next M trades, after which exposure returns to normal. Until that moment the phase continues, regardless of how the results turn out.
- Audit trail: the resumption date, the level of exposure on each of the first M trades, and the violation code on each of them.
The failure criterion (a requirement separate from the form) for the return phase is in the box closing the section.
The direction of the ramp is all this article needs: you come back with exposure smaller than usual and increase it only once a recorded condition is met, never merely once time has passed. Exactly how much "smaller" is, how many stages the ladder has, and what happens when a limit is exceeded in the middle stage, you will compute in article 9.8 for your own parameters. This article gives no multipliers, because any given here would be made up.
8. The audit trail: what can be counted and what cannot
The metric of the protocol is not the financial result. Not because the result does not matter, but because a single result does not tell a good procedure from a bad one – that is settled by article 10.4. The metric is the events the procedure was supposed to change.
Below is an illustrative run: one session, seven orders, the trigger condition met after the third. The values are illustrative and are meant to show how the trace is counted; they are not a reconstruction of anyone's session.
| Quantity | Value | Where it comes from |
|---|---|---|
| Orders in the session | 7 | history export |
| Orders after the trigger condition was met | 4 | history export, compared against the time the condition was met |
| Share of orders after the condition | 57.14% | own calculation: 4 / 7 |
| Violations before the condition | 0 | list of violation codes from the journal |
| Violations after the condition | 9 | list of violation codes from the journal |
| Violations per order after the condition | 2.25 | own calculation: 9 / 4 |
All six quantities share one feature: they can be read from a file without reaching into memory. None requires an answer to the question of how the trader felt at the time. And none of them changes depending on which way the market went afterwards.
The share of orders after the trigger uses the total order count as its denominator, and that total can itself change because of the protocol. The ratio may therefore rise with an unchanged numerator when there were fewer entries overall. Always record both counts and the trigger time. Four out of seven is 57.14%; four out of six is 66.67%. A rise of 9.52 percentage points does not mean that more orders were placed after the trigger. A single episode can establish a breach of the block itself. Lasting process improvement, however, cannot be determined by a universal threshold of two or three episodes: it requires further observations and comparison conditions specified in advance.
An absolute count also needs context. Fewer orders after the trigger may reflect a shorter observation window, lower liquidity or fewer opportunities. Specify the post-trigger window in advance and record entry opportunities and relevant market conditions. Do not discard an episode simply because it contains fewer orders: reducing their number may be the purpose of the procedure. Four out of six versus two out of three means fewer orders and the same share, 66.67%; without information about time and opportunities, you do not know why. Keep every episode in the journal and describe different conditions explicitly rather than removing inconvenient observations.
The chart below only demonstrates how the metric is calculated in two illustrative episodes. The second contains the same four post-trigger orders and fewer orders overall. You can describe that difference, but cannot attribute it to the protocol without further information about execution, the observation window and opportunities. Neither two nor three episodes automatically make a conclusion reliable.
A comparison in which every run confirms the usefulness of the procedure measures nothing – it measures only the selection of examples. A number capable of coming out the inconvenient way is the only one worth watching.
What cannot be counted
The sentence “if I had applied the protocol, that streak would have cost me X rather than Y” does not describe an observed result. You can estimate such a scenario in a historical test if you explicitly specify return times, position sizes, entry conditions, execution and costs. It remains a model result that depends on those assumptions, rather than proof of the exact saving achievable in a live account. Actual order history lets you calculate the result of executed orders; it does not, without an additional model, assign a result to decisions that were never made.
The second family of uncheckable numbers is numbers quoted without a method. The same sequence of losses gives two different results depending on whether the components are added or compounded on running capital – and a drawdown quoted as a single number, without the list of losses that made it up, does not say which of those two results it is. Below is the same calculation written out from components to result, so that every step is visible. Illustrative values.
| Step | Loss on the step | Capital after the step (compounding) | Sum of losses (additive) |
|---|---|---|---|
| start | — | 100.00 | 0% |
| 1 | 1% | 99.00 | 1% |
| 2 | 1% | 98.01 | 2% |
| 3 | 2% | 96.05 | 4% |
| 4 | 3% | 93.17 | 7% |
| Total drawdown | — | −6.83% | −7.00% |
The third column is a product: 100 × 0.99 × 0.99 × 0.98 × 0.97 = 93.168306, that is a drawdown of −6.8317%, rounded to −6.83%. The fourth is plain addition: 1 + 1 + 2 + 3 = 7, that is −7.00%. The difference is 0.17 percentage points and over four steps it is of no practical significance. For a sequence of 1%, 1%, 1%, 2%, 3% and 4%, the same pair of calculations gives −11.45% compounded against −12.00% additive, that is 0.55 percentage points: the divergence depends on both the number of steps and the size of each loss. Own calculation, with the method stated explicitly next to each of the four numbers.
9. The same protocol after a winning streak
A review after a winning streak is easy to overlook because a favourable result may reduce interest in how it was achieved. A loss draws attention to the outcome, but a profit can also accompany a plan breach. Check the record for breaches after a winning streak that call for the same block and review. We do not assume that every trader has them. The export describes your observations; it does not by itself establish their cause or the effectiveness of a future protocol. The order of sections in this article does not determine whether you need a separate procedure after gains.
The mechanism requires no biology here, the record is enough – and it is enough to pose a question instead of a claim. There is no study showing that after a winning streak retail traders show the same violations as after a losing streak, so that sentence does not appear in this article in the indicative mood. There is, however, something anyone can check in their own history export, and it is that which settles the matter individually.
Start with breach codes: list them separately for trades after favourable streaks and after losing streaks. If the lists are similar — position size exceeds the calculator result, entries lack required conditions, or there is no journal entry before the decision — consider applying the same safeguards after gains. The trigger should describe a specific departure from the exposure rule rather than the number of wins; an example is order size increasing while capital is unchanged. If the lists differ, record the observation and review it in the next comparable window rather than automatically adding a second protocol. The file makes this hypothesis checkable, but similar lists do not by themselves establish that the safeguard is effective.
Two phenomena described in the literature are useful here as context, and both have limitations that must be given together with them.
Overconfidence is not one thing. Moore and Healy separated three phenomena that the colloquial name glues into one[3]; that distinction and the full catalogue of cognitive biases are the scope of article 10.3 and will not be developed here. It is a theoretical review paper with laboratory experiments; there is no study of traders in it. For the protocol, one consequence matters: the problem is not high confidence, but the divergence between stated confidence and accuracy – and that is checked by setting previously recorded confidence levels against outcomes across many trades, not after a single good run.
Profits are sometimes booked separately. Thaler and Johnson described a tendency to take more risk after an earlier win – to treat the profit as "house money"[4]. The study was run on students, with hypothetical stakes in some of the experiments. This is an extrapolation to retail trading and I label it as one. The full discussion of mental accounting and the reference point belongs to article 10.6; here the procedural consequence is enough: if your position size depends on how the previous trades turned out, rather than on a previously recorded sizing rule that accounts for capital, risk and the method’s conditions, that is a violation and you record it as one.

The procedure is the same; only the values in the fields change. Below are all four fields from section two filled in on both sides, and beneath them the failure criterion; the last row of the table is a comment on execution and likewise does not belong to the form.
| Protocol element | After a losing streak | After a winning streak |
|---|---|---|
| Trigger condition | violations accumulating within one session | position size exceeding the calculator result across successive sessions, with no single clear starting point |
| Action | the next order stops being possible | position size returns to the calculator result and does not grow until the review is finished |
| End criterion | the time T has elapsed, the episode review is done and recorded | a review of the winning trades' consistency with the record made before the decision is done and recorded |
| Audit trail | the trigger time, the number of orders after it, the list of violation codes | the number of trades in which size exceeded the calculator result, and by how much it exceeded it |
| Failure criterion (a requirement separate from the form) | the count of orders after the condition was met does not fall between successive episodes | the count of trades sized above the calculator result does not fall between successive good runs |
| The hardest moment (a comment, not a field) | carrying out the block | noticing that the condition was met at all |
Both quantities in the “Failure criterion” row are absolute counts. Compare them with the observation window, opportunities and total decision count, because fewer entries may reflect either an effective block or changed conditions. An order that bypasses the block immediately exposes an execution problem. Assessing a persistent change requires more comparable data; three periods illustrate a sequence rather than a scientifically established minimum.
10. When the protocol has failed a second time – a relapse plan
A description of a protocol that ends at the return treats a second episode as though it were proof that the trader did not try hard enough. That assumption is not only unpleasant but also useless: a procedure that does not foresee its own failure has no execution branch at exactly the point where it would be most useful. A relapse is a foreseen state, so it has a separate record, drawn up together with the rest of the protocol.
Telling the cases apart starts with a single question: what went wrong the second time. There are three answers and they lead to three different actions.
| What happened | How it shows in the record | What changes |
|---|---|---|
| The procedure did not fire | the condition was met and there is no trigger time in the record | the condition is undecidable or nobody is checking it – rewrite it as an event readable automatically |
| The procedure fired and was bypassed | a timestamp of a settings change or a login from another device between two orders | the barrier stands on the wrong side – move it to where undoing it takes time and someone else's involvement |
| The procedure was carried out and behaviour did not change | the same number of orders after the condition and the same number of violations as in the previous episode | the trigger condition stands too late – move it to the first violation instead of the Nth loss |
What these three answers have in common is that none of them reads "repeat the same thing with more determination". Repeating the same procedure after it has failed is the weakest response, because it changes none of the parameters that failed.
There is a fourth possibility too, and it has to be listed with the rest: the procedure is too rigid and blocks trading in situations where nothing bad was happening. You can see it in the trigger condition lighting up in sessions with not a single violation, purely because of a losing streak that fits inside the distribution from section four. The cost of excessive rigidity is real and article 10.7 discusses it; here the diagnosis alone is enough: if the protocol fires mainly after losses rather than after violations, it is measuring the wrong thing.
11. What the protocol does not solve
Everything above concerns the execution of a process and can be improved by recording, reviewing and moving the barrier. There is, however, a set of observations for which another version of the procedure is the wrong answer, and it is more honest to say so outright than to propose a harder block.
It is worth adding why this caveat stands here rather than in a footnote at the end. A reset protocol is a tool that in these situations looks apt – the problem, after all, is a failure to stop – and precisely for that reason it can push back the moment of reaching for the right help. Another, harder block looks like progress. But if every successive one is bypassed, the information is not that the block was too weak.
There is nothing exceptional or shameful in this: the course teaches you to construct a decision process, not to solve problems of a different nature that call for different tools. Physical conditions affecting decision-making capacity – sleep, fatigue, session timing – have their own article and their own medical boundary, described in article 10.11.
The protocol in this article does one thing and you should not expect a second from it: it aims to reduce the number of decisions taken after the process stopped being executed as written. It does not improve the strategy, does not create an edge and does not guarantee that a further episode will not occur. The next article deals with the conditions in which an episode begins, before the first order is placed – with informational pressure and a move running away from you: FOMO and information hygiene (10.10).
FAQ – the halt-and-return protocol
After how many consecutive losses should I stop trading?
How long should the pause be?
How am I supposed to know that I am ready to come back?
Do breathing exercises, movement or a cold shower help you reset after a loss?
How will I know that my protocol is not working?
Does a protocol after a winning streak make sense, if everything is going well?
Sources and references
- Loewenstein, G., "Hot-Cold Empathy Gaps and Medical Decision Making", Health Psychology, 24(4S), 2005, pp. S49–S56. A review with experiments: people in a cool state systematically underestimate the influence of an aroused state on their own future behaviour. The context is health and medical decisions, not trading – only the design consequence is carried over to trading. Source.
- Gollwitzer, P.M., Sheeran, P., "Implementation Intentions and Goal Achievement: A Meta-Analysis of Effects and Processes", Advances in Experimental Social Psychology, 38, 2006, pp. 69–119. A meta-analysis of 94 independent tests, d ≈ 0.65. This is a historical pooled estimate across different goals and populations, not the expected improvement in trading. The “if–then” format provides a starting point for procedure design, while its execution and effects must be checked in your own process. Source.
- Moore, D.A., Healy, P.J., "The Trouble with Overconfidence", Psychological Review, 115(2), 2008, pp. 502–517. A separation of three distinct phenomena: overestimation of one's results, belief in one's superiority over others, and excessive precision of estimates. A theoretical review paper with laboratory experiments, not a study of traders. Source.
- Thaler, R.H., Johnson, E.J., "Gambling with the House Money and Trying to Break Even: The Effects of Prior Outcomes on Risky Choice", Management Science, 36(6), 1990, pp. 643–660. Participants: students; hypothetical stakes in some of the experiments. An earlier win was associated with increased risk-taking in the next choice. Replication status: the "house money" effect has sometimes been reproduced, but later results are ambiguous and depend, among other things, on whether the stakes are real – which is why it serves here purely as context and not as the basis of a rule. The extrapolation to the currency market is marked in the text. Source.
- Baumeister, R.F., Bratslavsky, E., Muraven, M., Tice, D.M., "Ego Depletion: Is the Active Self a Limited Resource?", Journal of Personality and Social Psychology, 74(5), 1998, pp. 1252–1265. The original formulation of the self-control depletion model. Cited only together with item [6]. Source.
- Hagger, M.S. et al., "A Multilab Preregistered Replication of the Ego-Depletion Effect", Perspectives on Psychological Science, 11(4), 2016, pp. 546–573. A registered replication report: 23 laboratories, N = 2141, d ≈ 0.04, a confidence interval spanning zero – an effect indistinguishable from zero. That is why the argument for the protocol in this article does not rest on "willpower depletion" but on the hot–cold gap from item [1]. Source.
- Lally, P., van Jaarsveld, C.H.M., Potts, H.W.W., Wardle, J., "How are habits formed: Modelling habit formation in the real world", European Journal of Social Psychology, 40(6), 2010, pp. 998–1009. 96 people recruited, 82 datasets available for modelling, with a good fit for 39. The median of 66 days and range of 18–254 refer to the modelled time to reach 95% of asymptotic automaticity in simple daily actions, not a trader norm. Source.
- Steenbarger, B.N., The Psychology of Trading: Tools and Techniques for Minding the Markets, Wiley, 2003. Practitioner literature – the source for the description of daily limits and of barriers restricting the number of decisions as craft practice, with no measurement of effectiveness. Source.
- World Health Organization, International Classification of Diseases ICD-11, category 6C50 (gambling disorder), version in force since 2022. Cited solely as an indication that a boundary exists between education about decision-making and clinical diagnosis. It is not a tool for self-diagnosis and is not used in this article to assess the reader. Source.