Revenge trading, overtrading and the mechanics of tilt
After a loss, the next order may still follow the plan—or become an attempt to recover money at any cost. The difference appears in the entry condition, position size and protective order, rather than the trader’s mood alone. This article follows how individual departures can form a sequence, increase costs and deepen a loss. We also compare two illustrative paths through the same session: one ends after the third loss, while the other continues outside the rules. Article 10.4 separated decision quality from outcomes; here we examine how departures from the process can be identified in the order record.

The previous loss does not supply the next position size
“I want the account back where it was” describes a goal. It does not specify a signal, an exit or the amount another position can lose. To inspect a post-loss decision, put that sentence beside the rule recorded before the loss. The useful comparison is between two proposed actions with the same market inputs, rather than between a calm-looking and an upset-looking trader.
In our hypothetical record, earlier closed trades total −46 USD net. The already written rule caps the next trade’s scenario loss at 40 USD. That is an invented constraint, not a recommended limit. A new EUR/USD candidate uses 40,000 EUR on a USD account, with one pip equal to 0.0001 USD per EUR. Stipulated complete fills are a buy at Ask 1.2080 and an exit at either Bid 1.2072 or Bid 1.2092. These are calculation assumptions, not promised stop or target fills.
Reconcile the ordinary candidate before changing its volume
| Volume | Pip value | Net loss at protective fill | Net gain at target fill |
|---|---|---|---|
| 40,000 EUR | 4 USD | 34 USD | 46 USD |
| 80,000 EUR | 8 USD | 68 USD | 92 USD |
The protective distance is 8 pips and the target distance is 12 pips. At the original volume, a stipulated 2 USD round-trip commission gives a scenario loss of 8×4+2=34 USD and a target gain of 12×4−2=46 USD. No financing, conversion charge or other adjustment occurs. The fill-based price results already include the quoted sides; there is no second spread deduction.
Now the fictional entry note says “double it to recover faster”. Under unchanged fills and proportional commission, the doubled volume costs 4 USD and gives an 8×8+4=68 USD loss or a 12×8−4=92 USD gain. The original candidate passes this one loss-budget check. The doubled candidate exceeds its 40 USD ceiling by 28 USD. Passing one check does not establish that a real order also fits margin, liquidity or other account constraints.
Keep compliance unchanged when you reveal the outcome
If the doubled position reaches the stipulated target, the combined session result is −46+92=46 USD. If it reaches the stipulated protective fill, the total is −46−68=−114 USD. For comparison, the original candidate produces a session total of zero at its target or −46−34=−80 USD at its protective fill. None of these branches is assigned a probability or an expected return.
A later gain does not remove the budget breach already present when the doubled order was proposed. A later loss does not prove that the compliant candidate had a negative expectation. Cover the outcome column during the first review and ask which stated condition was satisfied or violated. Then examine the cash outcome as a separate part of the record. That order prevents a favourable branch from becoming a substitute for checking the original instruction.
Describe the observable departure before assigning a motive
In this case, the timing of the change and the recorded recovery aim accompany a specific budget departure. In another record, increased volume might follow a different rule written beforehand. The amount or speed alone cannot establish why someone acted. A note about intent is a report, not direct access to a psychological mechanism or a medical diagnosis.
Preserve the earlier rule version, the replacement proposal and the order/deal history. MetaQuotes separates the instruction from resulting deals and positions; an intention to reduce exposure is not the reduction itself. If a pause or review rule was also recorded, assess its execution from confirmations. This example does not choose a universal pause duration or treatment. The detailed lesson follows longer escalation sequences, counterexamples and the points at which a previously designed barrier can act.
Three exercises with worked answers
The exercise data are hypothetical. Choose an answer, then check the calculation.
Continue to complete escalation sequences
The detailed lesson distinguishes revenge trading from a valid post-loss entry, examines frequency and volume changes, and compares whole session paths. It also describes barriers and the limits of what an order record can say about emotional causes.
1. Anatomy of a post-loss sequence — what exactly changes
The label is the least useful thing here, so we start with the record. Below are five trades from a single session on a 10,000 USD account. These are illustrative values: for simplicity, 10 USD per pip per lot is assumed for every pair, and the sequence has been arranged to show the mechanism, not to reproduce anyone's history.
- 13:15 — Long EUR/USD, 0.20 lots, SL 50 pips (1% risk). A setup from the plan, written down before the session. SL hit, loss -100 USD. Day result: -1.00%.
- 13:45 — Long GBP/USD, 0.20 lots, SL 50 pips. The second setup from the plan. SL hit, loss -100 USD. Day result: -2.00%.
- 14:20 — Short EUR/USD, 0.25 lots, SL 40 pips. The size here is exactly by the rule: 0.25 lots with a 40-pip SL is precisely 1% risk, the calculator's output. The violation lies elsewhere, because the entry condition is not on the list from the plan. Loss -100 USD. Day result: -3.00%.
- 14:35 — Long USD/JPY, 0.40 lots. No SL set. Closed manually after 30 pips: -120 USD. Day result: -4.20%.
- 14:52 — Short GBP/JPY, 0.50 lots. An instrument outside the plan, no SL. Closed after 45 pips: -225 USD. Day result: -6.45%.
Five trades in 97 minutes. The last three depart from the rules written before the session, the last two have no SL. The two planned trades cost 200 USD, the remaining three cost 445 USD. That is 2.2 times the whole of the day's planned trading; per trade, the average loss rises from 100 to 148.33 USD, roughly 48%. Those two numbers describe different things and must not be confused: 2.2 is a ratio of sums, 1.48 a ratio of per-trade averages.
Notice where the record changes, rather than how the trader feels. The first two trades follow the plan: size calculated, SL before entry, condition from the list. The third loses the entry condition, although size remains within the rule. The fourth also has no SL, and its 0.40-lot size no longer follows the recorded risk. The fifth adds an instrument outside the plan. Violations accumulate from the third order onwards. Greater exposure increases the effect of an adverse move, but this example alone cannot separate its influence from that of prices or establish an emotional cause.
That last caveat matters. The sequence itself does not establish that the cause was an emotional state. Three losses in a row fall within the distribution of any working strategy, the spread on GBP/JPY is wider than on EUR/USD regardless of who places the order, and afternoon volatility can take out correctly placed stops. The evidence is not the result, it is the change in behaviour. And that can be read from the order history without guessing at what the trader felt.
On top of that comes the execution layer. When market orders go out one after another, some of them may be rejected or requoted. The fill arrives at a less favourable price, and in the middle of a fast sequence that is easy to miss. Slippage only becomes visible in the history export, not on the chart. The same execution error costs more when there are more orders.
2. A definition by behaviour — and a counterexample
Popular definitions call revenge trading an "impulsive attempt to recover losses" and tilt an emotional state in which a trader abandons the system. Such descriptions can organise experience, but they do not establish compliance with a rule. A later account of "I was furious" or "I was calm" may depend on memory and the trade outcome. Keep the state report separately, and assess behaviour against the plan and order history. Remembering an emotion does not by itself let you count violations.
We therefore start with behaviour measurable in the order history against rules written before the session. An order after a loss that breaks a personal rule establishes a process departure. We additionally use revenge trading when the record indicates an aim to recover the earlier loss. Without evidence of that aim, we identify a post-loss violation rather than a proven motive. Anger is not required to assess compliance.
| Violation | How to define it for yourself | Where it shows up |
|---|---|---|
| Shortened interval to the next decision | time below a recorded minimum; an interval below the median of comparable sessions is only a review flag | timestamp difference in the history export |
| Position size outside the rule | lot size larger than the calculator's output for the written risk | volume column compared with the plan |
| Lowered setup threshold | entry without the full set of conditions from the list prepared before the session | day plan compared with the actual entry |
| Increased frequency | entries above a recorded limit or without available planned conditions; frequency above a median is not decisive | histogram of entries through the day |
| Skipped checklist | no record of it being completed before the order was placed | the journal (10.8) |
| Moved or removed SL | a modification of the protective order towards a larger loss that was not permitted by the plan | order modification log |
| Instrument outside the plan | a pair that is not on the list approved before the session | symbol summary |
| Goal defined by an amount to recover | size or SL calculated so that the result covers one specific earlier loss | the recorded reason for entry, written before the decision |
The list requires rules recorded beforehand and comparison of similar conditions. A median from five trades is an unstable description of a small sample, not a ready-made norm. Compare the same instrument, window and signal type in sessions without escalation—including losing ones. Recalculating the median after thirty, forty and fifty entries can reveal sensitivity; these counts are an audit convention, not a scientific threshold. Similar results do not guarantee future stability. Unless a minimum interval has been recorded, a shorter time is a review flag rather than independent proof of a violation.
Tilt is not a point but a period — a stretch of session in which several violations from that list occur at once and persist across successive orders. Exactly how many "several" is cannot be settled universally, and this text does not try. If you need an operational threshold, adopt one explicitly as a convention for yourself — for example two different violations within three consecutive orders — and check after a month whether a threshold set that way catches the episodes you yourself recognise as problematic. The threshold is a parameter of your record, not a fact about traders.
Counterexample: a trade after a loss that is not revenge trading
Without this example the definition has no boundary, and the reader leaves convinced that every order after a loss is suspect. The example is a model case, the values are illustrative, and the "before the decision" layer was recorded before the outcome was known.
- Before the session (08:40): two conditions in the plan — a break of the Asian range on EUR/USD and a return to a demand zone on GBP/USD. Risk per trade written down, size calculated with a calculator, maximum number of entries fixed.
- 10:05: the first condition triggers, entry consistent with the record, SL set before the order was confirmed. The market turns, SL hit. Loss equal to the planned risk.
- 12:30: the second condition triggers, the same one that had been on the list since the morning. Interval since the previous close: 145 minutes, above the trader's own median. Size: identical to the first entry, calculated with the same calculator. SL: set before the order. Instrument: from the list. Recorded reason for entry: condition number two from the plan, with no reference to the day's result.
- Outcome: the second trade also ends in a loss. Day result: twice the planned risk.
There is not a single violation from the table here. This is an ordinary losing day, not a tilt episode. The difference does not depend on whether the trader was irritated at the time. You can be irritated and execute the process correctly; you can be entirely calm and break three rules in a row. The assessment follows the record, not the mood.
A losing streak on its own settles nothing yet. At a 50% win rate, three losses in a row within a triple of trades named in advance has a probability of 0.53 = 12.5%, and across a hundred trades such a streak is practically certain. The calculation with its full assumptions is in article 10.1. A losing streak is not a failure and it is not a symptom. The symptom is what the trader does with it in the next order.
3. Triggers and chasing the day's result
A loss is one possible event preceding a process change, but not the only one. Reviewing only post-loss orders may miss departures after a smaller-than-expected profit or a missed move. The table organises example situations and behaviours to examine; it neither estimates their frequency nor establishes a cause.
| Trigger | The behaviour that reveals it |
|---|---|
| A closed loss consistent with the plan | the next order at an interval shorter than your own norm |
| Slippage or a worse execution price | immediate repeat of the entry "at a better price", outside the condition |
| Your own execution error (wrong direction, wrong size) | attempt to reverse the position without re-analysing the condition |
| A profit smaller than expected | re-entering the same move after exiting, with no new condition |
| A missed move | entry after the fact, with the SL placed "anywhere" so as to get in on time |
| Someone else's result seen mid-session | change of instrument or size with no change in the plan |
| The first loss after a longer winning streak | raising size "because the statistics will even out anyway" |
One aim to check in the record of these orders is returning to breakeven on the current day. A personal session-opening level gives the market no reason to return to it. A financial objective may set working boundaries, but it is not an entry condition by itself. "Breaking even on the day" may become a reference point created by dividing a continuous account into daily compartments. Mental accounting may also make an open loss seem less real than a closed one, although the current valuation already includes it. The reference point and its variants are developed in article 10.6.
The operational consequence is to compare an earlier loss amount or the word "recover" in the rationale with the entry rule. They indicate an aim, but do not yet establish that the condition was absent. Account state may legitimately restrict new risk under the plan. The problem appears when recovering an amount replaces the required signal, changes size outside the rule or justifies an unauthorised SL change. That identifies a specific departure, regardless of the subsequent outcome.
4. Overtrading — when frequency eats the edge
A post-loss violation is not always immediately visible in equity. Costs may accumulate without larger positions or one dramatic session. Before attributing the problem to a trader’s state, compare entry counts with available conditions and execution costs with gross results. The examples below serve that distinction; they do not replace strategy assessment or classify a high trade count alone as overtrading.
Barber and Odean analysed household accounts at a US discount broker over 1991-1996 and showed that the highest-turnover accounts achieved gross results close to the rest but distinctly worse ones net. The difference was eaten by transaction costs[1]. The label on the study matters: these are equity investors in the US, observational data, no random assignment to groups, so alongside costs self-selection is also at work there. Whoever trades more may differ from the rest in several other respects as well. All we carry over to the FX market is the cost channel, not the effect size: the cost structure, the leverage and the horizon are all different. The cost channel itself, however, can be calculated on your own data, and that is what we do below.
Trader A takes 5 trades a week on EUR/USD. Spread 1.2 pips, commission 0, average slippage 0.3 pips, 1.5 pips per trade in total. Per year: 260 trades, 390 pips of cost. Trader B takes 25 trades a week on the same parameters: 1,300 trades and 1,950 pips. On a 10,000 USD account with a 0.10-lot position (1 USD per pip) that is 390 and 1,950 USD respectively, that is 3.9% and 19.5% of capital per year paid out on the cost side, before any trading result is counted at all. All values are illustrative.
| Parameter | Trader A (5/week) | Trader B (25/week) |
|---|---|---|
| Trades / year | 260 | 1,300 |
| Cost / trade (pips) | 1.5 | 1.5 |
| Annual cost (pips) | 390 | 1,950 |
| Annual cost (USD at 0.10 lots) | 390 | 1,950 |
| Result needed to break even | +1.5 pips per trade | +1.5 pips per trade |
| % of capital paid out in costs | 3.9% | 19.5% |
Counterexample: high frequency that is not overtrading
Reading that table as a verdict on anyone who trades often would lead astray. Cost on its own settles nothing. What settles it is comparing that cost with two other numbers: the average gross result per trade, and the number of conditions the strategy actually generated in the period.
Trader C takes the same 25 entries a week as trader B, at the identical 1.5-pip cost and the identical 19.5% of capital paid out per year on the cost side. There are two differences and both are visible in the record. His strategy generates an average of five conditions a day, so 25 entries a week is a number written into the plan, not an excess above it. And his average gross result per trade is clearly higher than the 1.5-pip cost, so the cost is covered. That 19.5% is then the price of running such a process, not a symptom.
The criterion is therefore relative, not absolute: the number of entries compared with the number of conditions generated in the same period. If trader B's strategy generates five conditions a week, as trader A's does, then at 25 entries twenty of them came from somewhere else. Trader C, at the same number of entries, has not a single one. Identical numbers in the export, two different objects. No absolute number of trades per day will settle this for you.
This account distinguishes two illustrative forms of overtrading by their trigger. The impulsive form: the trader treats a condition absent from the morning’s list as a signal. Every candle becomes a candidate for entry; opening a position may interrupt the waiting, but is not the only possible action. The compensatory form: the trader increases entries at unchanged size, with a recorded aim of recovering a loss. Instead of three trades, they take eight. A violation occurs if the additional entries lacked a valid condition or exceeded a limit. Eight orders against an average of three does not establish that alone: the number of valid signals can change. Check actual conditions and the recorded rationale, rather than the count alone.
The cost of frequency depends on position size relative to capital, rather than account size alone. On a 1,000 USD account with a 0.01-lot position, a 1.2-pip spread costs 0.12 USD per trade. At 10 trades a day and 22 working days, that is 26.40 USD a month, or 2.64% of the account, and 316.80 USD a year, or 31.68%. These are illustrative calculations. Profitability depends on comparing costs with the gross result; expected value is covered in section 8. If the gross margin is small, additional slippage, wider spreads or entries outside the condition may exhaust it. Their size must be established from the particular method’s data.
The extreme version of this arithmetic is easy to check for any scalping strategy. A 2,000 USD account, a pair with a 2.5-pip spread, 17 trades a day at 0.05 lots (0.50 USD per pip) gives 21.25 USD of cost a day, 467.50 USD a month, that is 23.4% of the account. After two months that is 46.8% in costs alone. To break even, every trade would have to yield an average of 2.5 pips gross, exactly what it costs to enter it. The arithmetic of costs is responsible for that, not the trader's state: before you decide the problem is tilt, check whether the strategy fits inside the cost at all. If it does not, no psychological procedure will fix it.
5. Escalation under arousal and escalation in cold blood
The sequence from the first section has a repeatable skeleton, named here explicitly because it points to the places where it can be broken:
event → interpretation → state → impulse → procedure violation → new exposure → new outcome
The event may be a hit SL or a worse execution price. Interpretation is the meaning assigned to it: "the market threw me out" and "the condition did not work" are different accounts. Interpretation and reported state can be recorded, but order history alone does not confirm them. A shorter interval is a timing trace, rather than a direct measure of an impulse. Assess a violation by comparing the action with the rule already in force. New exposure and an outcome may then become the next event. The diagram organises a proposed sequence; it does not establish every link or its cause.
The difference between tilt and an ordinary bad day does not lie in the number of losses but in whether the way they are entered changes. Three losses at unchanged size, with an SL before every entry and conditions from the plan, are simply three losses. The same number of losses after which position size grows, the SL disappears and the instrument list widens is a different object, and only the second one calls for a procedural intervention.
The variant under reported calm
Looking only for a strong emotion may make escalation harder to recognise. Adding to a losing position to justify an earlier commitment may occur alongside reported calm; an outburst of frustration need not be assumed. In a non-market experiment, people responsible for an earlier unsuccessful decision allocated more funds to the same course of action than people who inherited it[2]. That study does not establish an absence of arousal or the same effect in FX. Sunk costs, escalation of commitment and additions permitted by a plan are developed in article 10.6. Here we examine the sequence of orders.
The two illustrative variants may leave different traces in the data. Both may involve violations, but their cost and outcome also depend on position size and price movements:
| What you see in the order history | Escalation under arousal | Escalation in cold blood |
|---|---|---|
| Interval between orders | shortened, minutes | unchanged, sometimes longer |
| Number of instruments | grows | one, always the same |
| Direction | variable | constant, consistent with the original thesis |
| Size | grows in jumps | grows gradually or by adding |
| Stop loss | disappears or is moved | moved further out, "because the level changed" |
| Justification in the journal | no entry | an entry exists, but it refers to the entry price |
Maintaining or increasing exposure for longer may increase the effect of an adverse move. Calm alone does not establish an episode’s duration or cost. Not every addition is escalation outside the plan: compare the earlier rule and new information with execution; examples are discussed in article 10.6. A personal entry price used as the only reason for another order is a flag to check whether it replaced the required condition.
If a trader associates the problem only with an outburst of frustration, violations made alongside reported calm may be overlooked. An absence of strong emotion does not establish control, just as its presence does not establish a loss of control. Compare intervals, sizes and instruments with the plan and similar sessions. A historical norm helps flag change, but without a specific rule a difference from the median does not itself establish a violation. A state report can supplement that comparison; it cannot replace it.
A documented case: escalation with concealed positions
A publicly documented example of escalation and ineffective controls comes from the collapse of Barings bank. The chronology follows the official report of the British Board of Banking Supervision[12]: from 1992 losses on futures transactions were booked to an error account numbered 88888 and did not reach reporting, while the position in Nikkei 225 index contracts grew instead of shrinking. After the Kobe earthquake on 17 January 1995 the index fell, and the commitment was increased in anticipation of a rebound that did not come. On 26 February 1995 the bank declared insolvency. The accumulated loss was around 827 million pounds, and the institution was sold for a nominal sum.
The structure is the same as in the arithmetic in section eight, only several orders of magnitude larger: an unfavourable outcome, commitment increased instead of closed, and no barrier capable of stopping it from outside.
6. The night session — when conditions may increase the cost of an error
Night-time may change execution costs, the availability of external oversight and the trader’s state. None of those changes follows automatically from the hour, however. Examine three groups of conditions separately when the same departure may have a larger effect than in a comparable session:
Condition one: the cost of execution. Liquidity and spreads depend on the pair, time, data releases and provider. The European night is not a uniform window of minimum liquidity: some Asian currencies are actively traded then. Separate spread risk from fatigue and check your own instrument’s data. A worse execution price is possible when available liquidity is thin, but it does not follow automatically from the session’s name.
Condition two: a difference between planning and execution. A rule prepared calmly may be executed under loss-related pressure or fatigue, but that does not occur in every session. Cold–hot gap research shows that predictions of one’s choices may depend on the state in which the forecast is made[3]. Applying this to trading is a design hypothesis, not a diagnosis: check whether the rule was executable in the actual conditions rather than assuming a "worse version" of yourself.
Condition three: availability of an external safeguard. If a particular environment has no second-person oversight outside normal hours, one possible barrier is unavailable. That does not describe every trader or every night. A technical block may still operate: check its scope and whether it can be bypassed rather than equating a late hour with an absence of safeguards.
There is also returning to the platform “just to look” after a losing day has ended. In an independent-trial model, with the same setup and fixed costs, the previous outcome does not change the next trial’s expected value. That is a model assumption, not a description of every market: outcomes may provide information about changing conditions or the method. The process issue is another order outside the recorded window, without a valid condition. If such returns recur despite attempts to limit them, consider a doctor or psychotherapist; this course neither determines the cause nor diagnoses it.
| Parameter | Model A: better execution | Model B: worse execution |
|---|---|---|
| Illustrative EUR/USD spread | 0.8-1.2 pips | 1.8-3.0 pips |
| Illustrative GBP/JPY spread | 2.0-3.5 pips | 5-10 pips |
| Liquidity | high | lower — model assumption |
| Slippage on the SL | 0.2-0.5 pips | 1-5 pips |
| Number of earlier decisions in the day | few | usually a full day behind you |
| External brake | present | absent |
| Cost of the same violation | base | higher by the difference in execution |
The table compares two invented sets of execution conditions, rather than European and Asian session averages. Spread and slippage ranges are illustrative assumptions, not measurements or forecasts. Check your own quotes and execution prices. Worse outcomes at a particular time need not have a psychological cause: the instrument, entry condition, volatility and costs may have changed together. Article 10.11 discusses how to separate them.
Check the provider’s rollover window separately. Available liquidity and spreads may change around it; timing depends on the instrument, platform and daylight-saving changes. Do not treat 23:00 Central European Time as a fixed rule for all quotes. Compare the spread before the order with your execution criteria and check prices in the history. The process risk is omitting that check, rather than the window’s name itself.
The countermeasure for this part of the problem is procedural and trivially simple to describe. It is harder to maintain. Trigger: closing the platform at the end of the planned session. Action: no logging back in on the same calendar day, including from the phone. Completion criterion: the start of the next planned session. Auditability: login and order timestamps in the history export. How you will know it is not working: if within a month even one order appears outside the planned window, the barrier is too soft and has to be moved up a level, to the solutions described in section eleven. This is craft practice based on the mechanism of precommitment, not a finding from research on retail traders.
7. A limit set earlier than the moment it has to work
Everything calculated so far leads to one design question: who takes the decision to stop, and when. If the trader takes it in the middle of the sequence, the decision is taken by the same person and the same state that is running that sequence. If it was taken earlier and written down in a form that cannot easily be reversed, the decision is already out of the current moment's reach. How to design such a prior commitment so that it is neither a dummy nor a rule that blocks a sensible change of plan is described in article 10.7. The size of the limits themselves belongs to section 9 (article 9.8) and this article does not set it. What remains is the question proper to this one: at which point in the sequence such a decision must already have been taken.
A limit may be enforced technically outside the trader’s decision: once triggered, the system rejects new orders or closes positions. This is a design example, rather than a feature of every prop firm. In some offerings, crossing a limit only breaches contractual rules and does not trigger an automatic block. Check the calculation basis, open-position treatment, reset times and resistance to changes in the current terms. Risk parameters belong to section 9; individual provider terms belong to section 13.
Such a comparison does not establish that institutional traders remain in the market longer because of limits. Participant selection, oversight, capital source, instruments and horizon also differ. A narrower question is enough here: did the particular safeguard still block unauthorised orders when the trader tried to bypass it? If it works without current consent and cannot easily be removed, it may restrict that step in the sequence. That does not guarantee resistance to every failure or permission change.
For a retail account the same solution comes in several variants differing in their resistance to being reversed: from a limit enforced on the broker's side, through a rule written before the session and executed manually, to a plain interval of time spent away from the desk. Their design, format and the cost of excessive rigidity are the scope of article 10.7. For recognising an episode something else matters. What matters is whether the record shows that the safeguard was bypassed.
Three observations warrant review: more frequent triggers without fewer violations, violations shifting to another instrument or time, and removal of the barrier during a session. They do not automatically have one cause. Check trigger opportunities, the block’s scope and whether a change was permitted. An earlier trigger may be needed; another solution may be better enforcement or coverage of omitted instruments. The decision should follow from the specific bypass, rather than a counter alone.
The limit has one further function, visible only on the scale of weeks: it protects subsequent sessions from the consequences of one. Recovering from a deeper drawdown requires a gain larger than the drawdown itself and, at a realistic recovery rate, stretches across weeks of trading. The recovery ladder is the property of section 9 and that is where it should be calculated for your own parameters. For this article the consequence alone is enough: the bill for one session is paid across the next dozen or so sessions.
8. The cascade: the same session, two endings (with numbers)
Here is the full comparison. Starting session capital: USD 10,000; the base planned-loss limit is 1% of that opening value, or USD 100, rather than 1% of current equity. Strategy assumptions: a 55% win rate and reward-to-risk ratio of 1:1.8. Both paths have identical first three losses. The simplified model assumes a constant USD 10 per pip per lot for every pair; this is not the actual pip value of all the pairs listed. Costs and slippage are omitted. All values are illustrative.
Run A — session ended after the third loss
- Long EUR/USD, 0.20 lots, SL 50 pips. Loss -100 USD. Capital: 9,900.
- Short GBP/USD, 0.20 lots, SL 50 pips. Loss -100 USD. Capital: 9,800.
- Long USD/JPY, 0.13 lots, SL 75 pips. Loss -97.50 USD. Capital: 9,702.50.
- Stop: the threshold written before the session is reached, no new orders until the next session.
Day balance: -297.50 USD, that is -2.98%.
Run B — session continued
- Long EUR/USD, 0.20 lots, SL 50 pips. Loss -100 USD. Capital: 9,900. (per the plan)
- Short GBP/USD, 0.20 lots, SL 50 pips. Loss -100 USD. Capital: 9,800. (per the plan)
- Long USD/JPY, 0.13 lots, SL 75 pips. Loss -97.50 USD. Capital: 9,702.50. (per the plan, third loss in a row)
- Short EUR/USD, 0.40 lots, SL 40 pips. (first violation: size outside the rule) Loss -160 USD. Capital: 9,542.50.
- Long GBP/JPY, 0.50 lots, no SL. (second and third violations: instrument outside the plan, no SL) Closed at -55 pips. Loss -275 USD. Capital: 9,267.50.
- Short AUD/USD, 0.60 lots, no SL. Closed at -40 pips. Loss -240 USD. Capital: 9,027.50.
- Long EUR/USD, 0.80 lots, SL 30 pips. Hit. Loss -240 USD. Capital: 8,787.50.
- Short USD/CHF, 1.00 lots, no SL. Closed at -35 pips. Loss -350 USD. Capital: 8,437.50.
Day balance: -1,562.50 USD, that is -15.63%. The same start to the session, a trough more than five times deeper. How much trading it takes to climb out of each is calculated in section 9; here all that matters is where the difference came from.
The escalation arithmetic needs care. The first three trades cost USD 297.50; the next five cost USD 1,265. That is about 4.25 times as much in total. The average loss per trade rises from USD 99.17 to USD 253, or about 2.55 times. These are different comparisons: the total also depends on the number of orders. USD 1,265 is equivalent to about 12.76 average losses from the first three trades. This describes this particular illustration, rather than the expected loss of another order or proof that compliant trading always costs less.
The key thing here, though, is the structure of that arithmetic, not the multiplier itself. The growth is driven above all by position size, which rises from 0.13 to 1.00 lots, and not by a mysterious decline in decision quality. Size is visible in every statement and can be calculated before the order. That is why size, and not mood, is the proper control point.
At the end of such a sequence there is also the forced close. When the margin level falls below the threshold that forces a close, the order reaches the market as a market order, at whatever price happens to be there. In low-liquidity hours that can mean a fill far away from the level the trader had on screen. That is an additional reason why the size of the exposure at the end of a sequence matters more than at its start.
9. Doubling after a loss — escalation of commitment in its purest form
Doubling the stake after every loss so that one win covers all the previous ones is the oldest and purest example of the mechanism described in section five. The goal of the order here is defined by an amount to be recovered rather than by a market condition, which means it meets the last criterion on the list of violations. All the rest is position-size arithmetic, and that belongs to section 9.
Starting from 1% risk, each successive stake is twice the previous one, so after just a few losses a single order reaches several dozen per cent of the account, and the sequence ends not at a trader's decision but at the lack of free margin. The position size ladder, the risk-of-ruin arithmetic and the comparison with fixed risk belong to section 9 — that is where the numbers stand and where you substitute your own parameters for them. The gentler variant, doubling on every second loss, lengthens the time to capital exhaustion but does not change the structure: position size still depends on the outcome of the previous trade rather than on a market condition. How often a streak occurs that exhausts such a ladder is calculated in article 10.1; the conclusion from there is single and sufficient: this is a probable event, not an exotic one.
For recognising an episode, the structural difference alone is enough. The same system and the same series of outcomes end differently under fixed percentage risk than under doubling, because what decides survival of a streak is position size, not signal quality.
10. After the episode — where the cheapest break point lies
This article ends where the procedure begins. Recognising an episode and identifying the moment at which it could have been broken belongs here. The full break protocol, the criteria for returning, the graduation of position size and the plan for a relapse are the subject of article 10.9, and all numerical limits belong to section 9. What remains is one task that can only be carried out on your own data.
Reviewing the record of the episode. Open the order history export from that session, not the platform. For each trade answer five questions, based solely on what was known before it was placed:
- Was the entry condition on the list written before the session?
- Did the position size follow from calculated risk?
- Was the SL set before the order was confirmed?
- How many minutes had passed since the previous position closed?
- Which violations from the list in section two are present in this trade?
In a typical run the answers form a clear pattern: the first trades have the full set of confirmations, and from a certain point the gaps begin. That point is the answer to the question of where the barrier should stand. Not after three losses, but before the first order at which the first violation appeared. For everyone it falls somewhere different, and only your own record can settle it.
It is worth saying honestly what this review will not give you. It will not say why the violations occurred. It will not distinguish a session in which market conditions genuinely changed from one in which only the trader's acceptance threshold changed. That requires a series of such reviews, not one. And it will not replace the decision about what to do next, because that already belongs to the following articles.
The episode itself is recorded in a form suitable for comparison with the next one. The minimum is the date, the number of orders, the list of violations that occurred and the median interval between orders. The construction of such a record and the process metrics are described in article 10.8. Without a comparison with the previous episode there is no way to establish whether anything is changing.
11. Where in the sequence the barrier has to stand
There is one design consequence: a rule that requires a decision at exactly the moment when decisions tend to get changed is the weakest kind of safeguard. The order of events is responsible for that, not the reader's character. The decision to stop is taken by the same process that departed from the plan a moment earlier. The observations described below are craft practice known from practitioner literature[8]; there is no research measuring their effectiveness on retail traders and this text does not pretend otherwise.
Barriers differ in one respect: how much it costs to take them down. Some you switch off yourself, with one click and mid-session; others require another person's involvement or the passage of time. The literature calls that difference soft and hard precommitment, and the design of the commitments themselves — the format of a rule, the features of a good procedure, the hardness test for a barrier and the cost of excessive rigidity — is developed in article 10.7. That is where what belongs to this article ends. What remains here is the narrower question: at which point in the sequence a barrier has to stand in order to change anything, and how you will know it has been placed too late.
The technical variants differ between platforms and change quickly, so there is no point describing the configuration step by step in a psychology article. That belongs to the section on platforms. From the point of view of this article two questions matter: does the tool require your consent when it activates, and can you immediately remove its block?
When the barrier becomes part of the sequence
A safeguard is sometimes bypassed, and that bypassing is already an element of the episode rather than a break in it. Three events say so outright, and all three leave a trace in the data.
A barrier removed mid-session. If a limit was disabled outside the permitted procedure and this allowed an order it should have stopped, a configuration-level violation occurred. Check the time and nature of the change, authorisation and subsequent orders. A timestamp between two entries alone is not decisive. Circumvention may show that the block enforced its condition but was insufficiently resistant to modification.
Orders placed from another device. Deleting the app from your phone lengthens the path from impulse to order and does nothing more. If orders nonetheless appear in the export outside the planned window, placed from elsewhere, the problem is not the choice of tool. Repeatedly working around your own blocks despite repeated attempts to maintain them is a signal that the matter goes beyond a trading procedure — the right address is then a doctor or a psychotherapist, not one more safeguard.
Topping the account up mid-session. Capital split between a trading account and a bank account works solely through the time a transfer takes. A transfer made mid-session or the same day removes the only active part of that mechanism, and incidentally raises exposure at precisely the moment when the process is already altered. If top-ups during an episode recur, this is an observation from the same category as the previous one and the same note about professional help applies unchanged.
What is visible before the first violation
All the criteria from section two measure orders, so the earliest signal they give arrives once an order has already gone out. This article's thesis requires something earlier. The three observations below have no research behind them and are craft practice, but they have one advantage: they are recorded in the same file as everything else.
- Returning to the platform after the planned session has closed — the moment of logging back in itself, before any order. It is visible in the login log.
- An order prepared and not sent: volume entered, window open, no confirmation. An unsent ticket is not a cancelled broker pending order; it needs a separate record or a local interface log.
- A number of watched instruments or open charts above what the day plan provides for. Easier to count in the evening than during the session.
None of these three observations proves that an episode has begun, and all of them also occur in sessions closed without a single violation. Their value is different: if across the record of several episodes they regularly precede the first violation, you get the place where a barrier costs least. The signal that this observation adds nothing: after three episodes the pattern is different every time, or it appears just as often in sessions without violations.

12. When this is no longer a subject for a trading course
Everything above concerns the execution of a process and can be improved with a record, a review and the construction of barriers. There is, however, a set of observations at which the problem stops being a matter of trading procedure, and it is honest to say so outright rather than to offer another checklist.
There is nothing exceptional or shameful in this: the course teaches how to construct a decision process, not how to solve problems that have a different nature and different tools.
The sequence described in this article begins long before the moment anyone calls it tilt, and that is exactly why recognising it rests on the record rather than on self-assessment. What happens to a loss before any violation occurs at all is covered separately — the reference point, the asymmetry of gains and losses, and decisions about holding positions: loss psychology (10.6). The break-and-return procedure itself, to which this text has referred several times, is the subject of article 10.9.
FAQ — revenge trading, overtrading and tilt
What exactly is the difference between revenge trading and overtrading?
How many trades a day counts as overtrading?
How do I know my safeguard is hard and not just a declaration?
Do sequences of violations also happen to experienced traders?
Can doubling after a loss work at a high win rate?
What should I do if I go back to trading right after a loss despite the barriers?
Does algorithmic trading eliminate this problem?
Sources and references
- Barber B.M., Odean T., "Trading Is Hazardous to Your Wealth: The Common Stock Investment Performance of Individual Investors", The Journal of Finance, 55(2), 2000, pp. 773-806. Observational data from household accounts at a US discount broker (1991-1996), equity investors: high turnover with comparable gross results produced distinctly worse net results. Transfer to the FX market limited to the cost channel.
- Staw B.M., "Knee-deep in the Big Muddy: A study of escalating commitment to a chosen course of action", Organizational Behavior and Human Performance, 16(1), 1976, pp. 27-44. An experiment with budget decisions: people responsible for an earlier unsuccessful decision allocated more funds to the same course of action. A mechanism of escalation of commitment, not a study on traders.
- Loewenstein G., "Hot-Cold Empathy Gaps and Medical Decision Making", Health Psychology, 24(4S), 2005, pp. S49-S56. Systematic underestimation of one's own behaviour in a state of arousal when planning it in a cold state. A medical decision-making context — only the design consequence is carried over to trading.
- 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. Always cited together with item [5].
- 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 = 2,141, d ≈ 0.04, confidence interval spanning zero — an effect indistinguishable from zero.
- Coates J.M., Herbert J., "Endogenous steroids and financial risk taking on a London trading floor", PNAS, 105(16), 2008, pp. 6167-6172. 17 traders from one floor, 8 days, saliva measurements; a correlational result — morning testosterone co-varied with the day's result, cortisol with market volatility and the dispersion of results. Small sample, no transfer to retail trading.
- Tendler J., Carter B., The Mental Game of Poker, Jared Tendler LLC, 2011. Practitioner literature — the source of the concept of tilt and its behavioural description, not of empirical evidence.
- Steenbarger B.N., The Psychology of Trading: Tools and Techniques for Minding the Markets, Wiley, 2003. Practitioner literature — techniques for limiting the number of decisions and daily limits as craft practice, with no measurement of effectiveness.
- European Securities and Markets Authority (ESMA), product intervention decision on contracts for difference, May 2018 — the obligation to disclose the percentage of retail accounts losing money at a given provider. Values depend on the provider and the period; regulatory detail in section 13.
- Bank of England, Financial Stability Report — July 2016, 5 July 2016. Official account of the market response to the referendum result, sterling volatility and the 23 June–1 July period; not a slippage forecast for a retail FX account. Primary source.
- Bank of Japan, "Statement on Monetary Policy", 28 April 2022 — maintenance of existing policy and the announcement of daily fixed-rate bond purchase operations. The market reaction (yen weakening, USD/JPY breaking through 130) per market services from that day.
- Board of Banking Supervision, Report of the Board of Banking Supervision Inquiry into the Circumstances of the Collapse of Barings, HMSO, London, 18 July 1995. Official findings: error account no. 88888, the accumulation of unreported losses on futures contracts, and the chronology and scale of the bank's collapse in February 1995. An institutional case — without transferring the conclusions to a retail account.