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.

Conceptual illustration: successive process violations can increase exposure and costs. A loss alone does not establish revenge trading.

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

The same hypothetical fills at two volumes
VolumePip valueNet loss at protective fillNet gain at target fill
40,000 EUR4 USD34 USD46 USD
80,000 EUR8 USD68 USD92 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.

1. What scenario loss does the doubled candidate produce?

80,000 EUR; pip value 8 USD; protective distance 8 pips; total round-trip commission 4 USD.

Show the answer and explanation

B. Recalculate the proposed quantity from the same assumed fills and costs before comparing it with the rule.

  • A. That omits the separate commission.
  • B. Correct: 8×8+4=68 USD.
  • C. That retains the original volume.
  • D. That is the written ceiling, not the candidate’s loss.
2. What is the session total at the doubled candidate’s protective fill?

Earlier completed net result −46 USD; new net scenario loss 68 USD.

Show the answer and explanation

D. Keep earlier booked results and the new net outcome in separate rows, then add them once.

  • A. That leaves out the earlier completed result.
  • B. That uses the original candidate’s loss.
  • C. That adds a loss as though it were a credit.
  • D. Correct: −46−68=−114 USD.
3. If the doubled position instead wins, what happens to its budget classification?

The proposed loss was above the unchanged written ceiling, but the realised outcome is a gain.

Show the answer and explanation

A. Evaluate the proposed action from the information and constraints available then, before looking at the outcome.

  • A. Correct: a later gain does not change the condition recorded at the decision.
  • B. That substitutes the outcome for the earlier rule check.
  • C. One branch does not estimate an unknown outcome distribution.
  • D. Another trade still requires its own current inputs and rules.

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.

  1. 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%.
  2. 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%.
  3. 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%.
  4. 14:35 — Long USD/JPY, 0.40 lots. No SL set. Closed manually after 30 pips: -120 USD. Day result: -4.20%.
  5. 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.

Five orders: where the violation begins. Article illustration · opening capital USD 10,000
97 minutes. Simplification: USD 10/pip/lot for every pair; costs omitted. Not account data.

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.

How it looks around a data release (illustrative run). On payrolls, re-entering after a stop-out looks harmless for the first half a minute. The market comes back without the trader, so a second order appears — except that it is placed on a widened spread, in worse liquidity, and without re-checking the condition. The direction is sometimes right and the result still comes out negative, because the difference is made by the execution price, not by the market hypothesis.

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.

ViolationHow to define it for yourselfWhere it shows up
Shortened interval to the next decisiontime below a recorded minimum; an interval below the median of comparable sessions is only a review flagtimestamp difference in the history export
Position size outside the rulelot size larger than the calculator's output for the written riskvolume column compared with the plan
Lowered setup thresholdentry without the full set of conditions from the list prepared before the sessionday plan compared with the actual entry
Increased frequencyentries above a recorded limit or without available planned conditions; frequency above a median is not decisivehistogram of entries through the day
Skipped checklistno record of it being completed before the order was placedthe journal (10.8)
Moved or removed SLa modification of the protective order towards a larger loss that was not permitted by the planorder modification log
Instrument outside the plana pair that is not on the list approved before the sessionsymbol summary
Goal defined by an amount to recoversize or SL calculated so that the result covers one specific earlier lossthe 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.

Where the word "tilt" comes from and what we carry over from it. The concept came from poker and from practitioner literature (Tendler[7]), and the related idea of "loss chasing" is described in gambling research. This is an extrapolation and we treat it as one. All we carry over is the structure of the behaviour: an observable departure from one's own process after an unfavourable outcome. We do not carry over effect sizes or clinical conclusions, because there are too many differences. Leverage differs, so does the continuity of quotation, the presence of an opponent at the table, and the existence of a hand as a natural unit after which play ends. The sentence "in trading the mechanics are identical" is not supported by data and does not appear in this article.

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.

What this observation does not prove. The list of violations lets you recognise that the process was changed. It does not let you rule on why. Nor does it say that a run without violations will be profitable, or that a run with violations must end in a loss. It happens that an order placed outside the rules turns a profit, which may reinforce unjustified behaviour and make the violation harder to notice. The criterion of assessment is set out in article 10.4.

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.

TriggerThe behaviour that reveals it
A closed loss consistent with the planthe next order at an interval shorter than your own norm
Slippage or a worse execution priceimmediate 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 expectedre-entering the same move after exiting, with no new condition
A missed moveentry after the fact, with the SL placed "anywhere" so as to get in on time
Someone else's result seen mid-sessionchange of instrument or size with no change in the plan
The first loss after a longer winning streakraising 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.

Frequency multiplies execution cost. 1.2 pips spread + 0.3 pips slippage · USD 1/pip
Illustration: 52 weeks, 0.10 lot EUR/USD, capital USD 10,000, zero commission. Gross P&L is not calculated here.
ParameterTrader A (5/week)Trader B (25/week)
Trades / year2601,300
Cost / trade (pips)1.51.5
Annual cost (pips)3901,950
Annual cost (USD at 0.10 lots)3901,950
Result needed to break even+1.5 pips per trade+1.5 pips per trade
% of capital paid out in costs3.9%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.

The percentage of losing accounts — what that number means and what it does not. Following the ESMA product intervention of 2018, CFD providers in the EU and the UK are required to publish the percentage of retail accounts that lose money — Disclosure requirements differ across jurisdictions; outside the EU framework, check the applicable local rules.[9]. The figure refers to one specific provider and one specific period, so check it in your own broker's disclosure rather than in someone else's article. It is not evidence that the cause of the losses is psychological. It describes the distribution of outcomes, not the mechanism. Regulatory detail is covered in section 13. As checked on 8 October 2026; obtain the current percentage from the provider’s disclosure.

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.

A data review, rather than recollection alone. Trigger: month-end. Action: calculate the average daily trade count from the history export, separately for profitable and losing days. Completion criterion: record both figures and the median interval between trades in the journal. A higher count on losing days is a review flag, not proof of compensatory overtrading. Check available signals, volatility, instruments and any recorded aim of recovering a loss; the whole-day outcome may itself depend on those additional entries. Limitation: varying results after three months may reflect a small sample, changing conditions or no stable association. Also check whether the day definition and observation window changed during the series.

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.

Same gross result, different costs. Conditions and inferential limits are part of the diagram.
Illustrative cost assumptions as in the preceding graphic; future strategy outcomes are not assumed identical.

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.

The same setup, a different window. The feasibility of an entry differs ten minutes before a data release and two hours after it. Spread, slippage and the rejection rate are calculated separately for main-session hours and separately for data windows. The scale of the cost: a trader taking 5 trades a week who enters at a 2.5-pip spread instead of 1.0 pays 260 × 1.5 = 390 pips more over a year. At 0.10 lots (1 USD per pip) that is 390 USD; at 1 lot (10 USD per pip) — 3,900 USD. The same market hypothesis, a different bill for execution.

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.

Four stages of the sequence — how you recognise each one in the record 1. TRIGGER Loss, slippage, error Process still unchanged 2. DEVIATION A review flag Condition loosened 3. ESCALATION Size outside the rule No SL before the order 4. SEQUENCE Instrument outside the plan No recorded reason for entry BREAK POINT Early rule check The order of stages is illustrative — not every episode passes through all of them Violations accumulate in this example Here, a later break leaves more exposure
Four stages of the sequence described purely by observable behaviour: the trigger, the first deviation from the rule, escalation of size and of the missing SL, the full sequence of violations. The diagram orders the stages; it does not predict them.

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 historyEscalation under arousalEscalation in cold blood
Interval between ordersshortened, minutesunchanged, sometimes longer
Number of instrumentsgrowsone, always the same
Directionvariableconstant, consistent with the original thesis
Sizegrows in jumpsgrows gradually or by adding
Stop lossdisappears or is movedmoved further out, "because the level changed"
Justification in the journalno entryan 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.

Similar violations, different reported states. Conditions and inferential limits are part of the diagram.
Conceptual diagram. It does not establish motive, cause or financial outcome.

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.

What this case does not prove. It says nothing about the internal state of the person involved — the report examines decisions, accounting records and control mechanisms, not emotions, and this text does not add them. Nor does it transfer 1:1 to a retail account: scale, someone else's capital, access to the books and the ability to hide positions from supervision are conditions that do not exist on a retail account. Finally, it does not prove that escalation always ends this way — this is a case that ended badly and for exactly that reason ended up in a report, while cases that ended otherwise leave no documents behind. One thing carries over from it: the sequence ran on for as long as nobody other than the person running it stopped it.
Why there are no neurochemical explanations here. Popular accounts of tilt work with cortisol, the prefrontal cortex and the amygdala. The problem is that in none of those accounts did anyone measure anything in the trader being described. The study that actually did measure hormones in traders covered 17 people from one London trading floor over 8 days; it was correlational and linked morning testosterone to the day's result, and cortisol to market volatility and the dispersion of results[6]. That is an interesting finding and far too little to rule on anything about a single retail trader, and certainly too little to derive procedures from. Everything this article needs can be described at the level of behaviour that can be counted.

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.

A note on "willpower depletion". The Baumeister et al. (1998) model[4] does not by itself justify explaining evening errors as fuel consumption. A preregistered report from 23 laboratories, N = 2141, found d ≈ 0.04 with a confidence interval covering zero[5]. An effect was not demonstrated in that protocol; this does not establish that no effect exists in every task. Later projects yielded different results, discussed in article 10.7. The argument for an auditable barrier here does not require the willpower-muscle metaphor or attribution of an error to a particular psychological cause.
Session time does not explain everything. Conditions and inferential limits are part of the diagram.
No universally “worst session”. Check spreads and liquidity for the specific pair and provider.

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.

ParameterModel A: better executionModel B: worse execution
Illustrative EUR/USD spread0.8-1.2 pips1.8-3.0 pips
Illustrative GBP/JPY spread2.0-3.5 pips5-10 pips
Liquidityhighlower — model assumption
Slippage on the SL0.2-0.5 pips1-5 pips
Number of earlier decisions in the dayfewusually a full day behind you
External brakepresentabsent
Cost of the same violationbasehigher 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.

A violent move on a thin book (illustrative run). Night sessions can stand still for hours and then show in a few minutes what liquidity is really worth. On yen pairs, violent moves have happened precisely when most of Europe was asleep. A position without an SL looks safe in such a market right up to the moment when closing at the planned price stops being possible. In the morning there is no longer a question about the quality of the entry condition. There is a question about the size of the exposure left to close.

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.

A weekend gap and a move with the market open. A position held over a weekend cannot be managed through ordinary order execution while the market is closed, and an SL does not guarantee a post-gap price. The pound’s fall following the Brexit referendum result began on Friday 24 June 2016 while the market was open—it was not a post-weekend gap. The Bank of England documented the strong market response and increased volatility[10]. The example distinguishes risk timings, but does not establish that one window always gives worse execution than another.

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.

Terminal lockout (illustrative run). A trader on a desk reaches the daily drawdown limit. The system locks the terminal until the next session, regardless of what the trader thinks about it. The key point is what you must not conclude from this example: the value of the lockout does not depend on which way the market went afterwards. Had the market turned back, the same lockout would have looked expensive, and it would have been exactly as justified. A procedure is assessed on what was known before it fired, that is, that in this state and this session the rules had already been broken repeatedly. Judging it by the later price move is the same error that article 10.4 dismantles.

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.

A limit: trigger, enforcement and reversal. Conditions and inferential limits are part of the diagram.
Structural example, not a description of every prop firm. Check contractual rules and technical enforcement separately.

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.

What belongs to this article. Where the threshold is enforced and who can take it down are design questions — they are posed and settled by article 10.7. The value of the threshold itself, the reference point (balance, current equity or a trailing version) and the way it is calculated belong to section 9. What remains here is a weekly review: how many times you came close to the threshold, and whether the number of violations in the record is falling.

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

  1. Long EUR/USD, 0.20 lots, SL 50 pips. Loss -100 USD. Capital: 9,900.
  2. Short GBP/USD, 0.20 lots, SL 50 pips. Loss -100 USD. Capital: 9,800.
  3. Long USD/JPY, 0.13 lots, SL 75 pips. Loss -97.50 USD. Capital: 9,702.50.
  4. 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

  1. Long EUR/USD, 0.20 lots, SL 50 pips. Loss -100 USD. Capital: 9,900. (per the plan)
  2. Short GBP/USD, 0.20 lots, SL 50 pips. Loss -100 USD. Capital: 9,800. (per the plan)
  3. 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)
  4. Short EUR/USD, 0.40 lots, SL 40 pips. (first violation: size outside the rule) Loss -160 USD. Capital: 9,542.50.
  5. 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.
  6. Short AUD/USD, 0.60 lots, no SL. Closed at -40 pips. Loss -240 USD. Capital: 9,027.50.
  7. Long EUR/USD, 0.80 lots, SL 30 pips. Hit. Loss -240 USD. Capital: 8,787.50.
  8. 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.

One session, two endingsIdentical first three losses · the continuation is illustrative Model USD 10/pip/lot, no costs. Third loss: 0.13 × 75 × 10 = USD 97.50. Stopping does not guarantee better future returns. FOREX MECHANICS · PSYCHOLOGY One session, two endingsIdentical first three losses · the continuation is illustrative USD8 4008 8009 2009 60010 000012345678Order number (0 = start)USD 9,702.50 · −2.98%USD 8,437.50 · −15.63%Model USD 10/pip/lot, no costs. Third loss: 0.13 × 75 × 10 = USD 97.50. Stopping does not guarantee better future returns.
Illustrative capital paths within one session. The vertical axis is linear, USD 8,400–10,000. The third loss is USD 97.50; stopping does not guarantee a better future outcome.

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.

What this run does not prove. First, it does not prove that the cause of the violations was an emotional state. It only shows that the violations occurred and what they cost. Second, it does not measure the win rate within an episode: to say how far it fell you would need a large sample of such sessions with an unambiguous definition of an episode, not one sequence. Popular figures of the "win rate drops from 55% to 33% on tilt" kind have no source and do not appear in this article. Third, it does not prove that stopping always saves money. In a different run the fourth trade might have been profitable, which would not have made it consistent with the plan. Illustrative values show an arithmetic mechanism, not an empirical regularity.

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.

Margin call and stop-out are different events. Conditions and inferential limits are part of the diagram.
No universal thresholds are given. Gaps and slippage can change the actual execution price.
One session, two endings. Identical first three losses · the continuation is illustrative
Model USD 10/pip/lot, no costs. Third loss: 0.13 × 75 × 10 = USD 97.50. Stopping does not guarantee better future returns.

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.

The event such systems end on. Algorithms that average a position against the move look safe for as long as the market retraces. They end on a move that does not retrace. An example with a date: on 28 April 2022 the Bank of Japan maintained its existing policy and announced daily fixed-rate bond purchase operations in defence of its yield ceiling; the yen weakened sharply and USD/JPY broke through 130, the highest in two decades[11]. For an algorithm that adds further positions against such a move with growing volume, direction is irrelevant. What matters is the absence of the retracement the construction counts on. We do not quote the size of the move in pips or the result of a hypothetical account here, because that cannot be verified in the source.

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.

When size depends on the previous result. Conditions and inferential limits are part of the diagram.
A rule schematic, not an account simulation or proof that a given streak will occur at a particular time.
How to check doubling in someone else’s algorithm. Review three points before buying or launching it: (1) a declared win rate above 90% in a backtest warrants checking the calculation and open losses, but does not establish doubling; (2) a smooth curve requires distinguishing balance from equity and checking sample length and conditions — it does not by itself establish an impossible result; (3) trace whether volume rises after a loss and how the multiplier, grid and simultaneous-order limit operate. Parameter names alone do not settle it: rules or a history of actual sizes are needed. Risk-of-ruin consequences are covered in the article on the mathematics of losses.

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.

Three reflexes that come after an episode. The first: topping the account up immediately in order to return to the previous state, which is precisely the same mechanism that drove the sequence, only one level higher. If the top-up comes from money earmarked for other purposes or from debt, this is already an observation from section twelve: the right address is then a doctor or a psychotherapist, not another trading rule. The second: changing the system. A sequence of violations is not evidence that the strategy has stopped working, and swapping strategies after a losing streak is the textbook example of judging a decision by its outcome, described in article 10.3. The third: looking for the cause on the broker's side. Checking slippage and spread in the export makes sense, because those are concrete numbers, but the check has to settle the question rather than confirm an answer decided in advance.

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.

Without a history export there is nothing to discuss. Every criterion in this article (intervals, sizes, instrument counts, slippage) is measurable only in the data, not from memory. Exporting the order history and splitting the cost into spread, slippage and swap points is the operational minimum; the tools for that are described in the section on tools.
Four example barriers: a server-side limit, a time window, removing the phone app and separating capital.
Examples of restricting access to the next order. A barrier's strength depends on how it can be removed; none guarantees an outcome or an execution price.

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.

Observations beyond the scope of this course. Repeated loss of control over trading despite repeated attempts to regain it. Hiding the scale of trading or of losses from those close to you. Financing trading with debt or with funds earmarked for other purposes. Persistent insomnia. Persistently lowered mood or anxiety. These are matters for a doctor or a psychotherapist, not for an article about trading procedures. This text diagnoses nobody, judges nobody and indicates no method of treatment. It only indicates the boundary of its own competence. Loss chasing is, incidentally, the subject of a separate body of research literature on gambling, which says something about the seriousness of the phenomenon and nothing about any particular person.

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?
Revenge trading links an order to an attempt to recover a previous loss; the record may show both that aim and a rule departure. A post-loss violation alone does not establish the motive. Overtrading means entries beyond what the process permits: above a limit or without available conditions, rather than simply many trades. It can occur after a loss or otherwise, including at unchanged size. Check both against history and pre-decision records. A short interval or one session’s outcome is not decisive alone.
How many trades a day counts as overtrading?
There is no universal entry count. If the plan permitted three satisfied conditions on a particular day, while history shows eight entries including five without a valid condition, those five were outside the process. An average of three signals per day does not establish the same thing: one day may have more. Compare entries with actual conditions, limits and costs in the review. Higher frequency on losing days is a flag; first check volatility, instruments and available signals.
How do I know my safeguard is hard and not just a declaration?
What matters is enforcement and the ability to remove the block, rather than the strength of a resolution. The test and safeguard design are covered in article 10.7, and the threshold in section 9. A configuration-change timestamp between orders is something to examine, not independent proof of a violation. Compare the change with the plan: was it permitted, did it weaken the barrier and allow an order that should have been blocked? Unauthorised circumvention may also reveal a weakness in the safeguard’s design.
Do sequences of violations also happen to experienced traders?
There are no data here to quantify such episodes by a trader’s experience; an invented percentage explains nothing. A structural difference can be identified: some professional environments enforce limits outside the trader’s immediate decision. If a system actually rejects new orders and the trader cannot immediately remove the block, it prevents a particular step in the sequence. A server or contractual rule alone does not guarantee this. It does not establish greater personal resilience; it demonstrates a specific safeguard whose scope and susceptibility to circumvention must be checked.
Can doubling after a loss work at a high win rate?
A high win rate lengthens the time to a streak that exhausts the capital, but it does not eliminate it. At an 80% win rate, seven losses in a row within a group of seven named in advance has a probability of 0.2^7 = 0.00128%, that is, one sequence in 78,125. At a thousand trades a year the probability that a streak of at least seven losses occurs anywhere within the year is about 1% — so the expected waiting time is on the order of a hundred years. The problem is that under doubling one such streak is enough, and its position in time is random: it may just as well fall in the first year. Calculated by dynamic programming over streak-length states, assuming independent trials.
What should I do if I go back to trading right after a loss despite the barriers?
First check in the data whether the barrier was hard at all, and move it up a level if it could be taken down mid-session. Then compare the records of successive episodes: is the number of violations falling, or are they only moving to a different instrument or a different time of day. The full break procedure and the criteria for returning are the subject of article 10.9. If, however, the loss of control recurs despite repeated attempts, and there is hiding of the scale of trading from those close to you, financing trading with debt, persistent insomnia or lowered mood — that is a matter for a doctor or a psychotherapist, not for another trading procedure. This course does not diagnose and does not indicate a method of treatment.
Does algorithmic trading eliminate this problem?
It moves it up a level. An algorithm does not change position size in response to a result unless that is in the code, but the person supervising it can switch the system off mid-drawdown, change parameters halfway through a streak or move the stop threshold. The violation here is any configuration change made outside the planned review window, and it is recorded in exactly the same way. Teams working with systems apply rules such as "no parameter changes for a set number of trades after launch" or a requirement for sign-off by a second person. This is craft practice, based on the same precommitment mechanism, and not on research. It is worth adding that the value of such a rule is not decided by which way the market went afterwards.

Sources and references

  1. 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.
  2. 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.
  3. 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.
  4. 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].
  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.
  6. 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.
  7. 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.
  8. 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.
  9. 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.
  10. 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.
  11. 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.
  12. 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.

Your learning progress

Finished this lesson? Mark it yourself. Opening or scrolling a page never marks it complete.

Progress requires JavaScript and browser storage.

Saved only in this browser. No account or cross-device sync. Clearing site data can erase your progress.

Jarosław Wasiński

Editor-in-chief, ForexMechanics.com • Macro & FX analyst

Independent analyst and practitioner with over 20 years of experience in financial markets. Actively involved in Forex since 2007, with a focus on fundamental analysis, OTC market structure, and disciplined capital risk management.

  • Creator and editor-in-chief of MyBank.pl, providing reliable financial education since 2004.
  • Author of hundreds of in-depth market commentaries, structural analyses, and educational materials for traders and investors.
  • Advocate for market transparency, promoting education grounded in hard data and institutional reports (BIS, ESMA, CFTC).

All content is educational and informational only – it does not constitute investment advice or a recommendation. Remember: leveraged trading (Forex/CFD) carries a high risk of rapid capital loss.