Loss psychology: the reference point, loss aversion and the disposition effect

The account shows USD 11,000: a thousand above starting capital, but USD 1,400 below its earlier peak. The same situation can therefore look like a gain or a loss, depending on the reference point. That comparison may influence whether a position is closed, held or increased, although it does not change market evidence. Article 10.5 examined process violations after an adverse outcome. We now consider how an outcome is evaluated, what research actually says about loss aversion, and how to distinguish behaviour consistent with a plan from a justification added afterwards.

The same account value can be a gain relative to one point and a loss relative to another. This illustration does not show market data.

A smaller red number is not necessarily a better account result

A position showing a loss disappears when it is closed. Reopening the same exposure creates a fresh entry price and a much smaller open loss. Before treating that change of display as an improvement, reconcile equity. A screen label and the resources remaining in the account are different objects.

Our hypothetical USD account has a settled balance of 5,000 USD and a long position of 30,000 EUR bought at 1.1180. The balance already includes all previously booked charges, including the original entry commission. Current EUR/USD is Bid 1.1150 and Ask 1.1152. A pip is 0.0001 USD per EUR. Assume complete immediate closing and reopening at these unchanged quotes, a commission of 0.90 USD on each new deal, and no other fees, financing, taxes or currency conversion. This is an accounting example, not a proposed way to manage a losing position.

Carry the realised loss into balance before opening the new row

The same hypothetical account through the reset
CheckpointBalanceOpen price resultEquity
Before closing5,000.00 USD−90.00 USD4,910.00 USD
After closing, before reopening4,909.10 USD0.00 USD4,909.10 USD
After reopening4,908.20 USD−6.00 USD4,902.20 USD

The original open price result is (1.1150−1.1180)×30,000=−90 USD. Equity before the operation is 5,000−90=4,910 USD. Closing realises that loss and charges the new exit commission: balance becomes 5,000−90−0.90=4,909.10 USD. The original entry commission is not deducted again.

Reopening buys the same 30,000 EUR at Ask 1.1152 and charges another 0.90 USD, leaving balance of 4,908.20 USD. At the unchanged Bid, the new position immediately shows (1.1150−1.1152)×30,000=−6 USD. Its displayed open loss is smaller, but the earlier loss is now in balance. Equity is 4,908.20−6=4,902.20 USD.

Identify what the change of reference actually cost

Compared with leaving the position untouched at the initial checkpoint, the operation lowers immediate equity by 4,910−4,902.20=7.80 USD. The new entry crosses a 2-pip spread, worth 6 USD at this volume, and the two new commissions total 1.80 USD. Their sum is the same 7.80 USD. Do not add the old 90 USD again to that difference; both alternatives already contain the earlier marked loss.

MetaQuotes describes buying at Ask and selling at Bid. The arithmetic here follows those stipulated deal sides. An actual operation may also involve changed quotes, a temporary flat interval, different contract conditions or tax effects. Those omitted features would require additional records. The calculation establishes only the stated immediate comparison, not that closing or holding is generally preferable.

Separate a change in the display from an explanation of behaviour

After reopening, the position’s entry reference is different, while the account still carries the previous loss and the extra costs. The same currency quantity is again exposed to future price moves. Changing the label does not supply new market evidence or an expected return. An exit rule, revised hypothesis or capital constraint still needs its own justification.

Imas’s research distinguishes realised and paper losses in experimental risk-taking. It does not show that paying to reset an entry improves a particular trader’s performance. Nor can this ledger diagnose loss aversion or identify a motive. Record what was said beforehand, the actual action and the resulting statement separately. The detailed lesson then examines reference points, research findings and counterexamples; this example first keeps the economic result visible when an old red row is replaced by a new one.

Three exercises with worked answers

The exercise data are hypothetical. Choose an answer, then check the calculation.

1. What is equity before the close-and-reopen operation?

Settled balance 5,000 USD; current open price result −90 USD; no other adjustments.

Show the answer and explanation

A. Use the balance and open result at the same checkpoint. Previously booked entry costs are already in that balance.

  • A. Correct: 5,000−90=4,910 USD.
  • B. That leaves out the current open loss.
  • C. That is equity after the complete operation.
  • D. That is the checkpoint after closing and its new commission.
2. What is equity immediately after reopening at the stipulated quotes?

Balance 4,908.20 USD; new open price result −6 USD.

Show the answer and explanation

C. The new open row belongs in equity even though the older loss has moved to balance.

  • A. That reports balance without the new open loss.
  • B. That treats the spread loss as a gain.
  • C. Correct: 4,908.20−6=4,902.20 USD.
  • D. That assumes the operation did not incur the additional costs.
3. What explains the immediate difference from leaving the original position untouched?

All quotes and quantities are as stipulated; the old marked loss exists in both alternatives.

Show the answer and explanation

D. Compare like checkpoints and subtract only differences between the alternatives, while retaining their common history.

  • A. Realisation moves the old result into balance; it does not remove it.
  • B. A ledger does not establish an intervention’s effectiveness.
  • C. That counts an amount common to both alternatives again.
  • D. Correct: 6+1.80=7.80 USD.

Continue to reference points and the research

The full lesson distinguishes loss aversion, risk aversion, the disposition effect and mental accounting. It explains the populations and tasks studied, why estimated coefficients are not universal constants, and what a pre-decision record can establish.

1. One number, five reference points

Kahneman and Tversky's prospect theory (1979) rests on an observation that sounds trivial until you see its consequences: people evaluate outcomes as changes relative to some point, not as states of wealth[1]. To nobody is an account holding 11,000 USD simply "an account holding 11,000 USD". It is a thousand better or four thousand worse – depending on what happens to be sitting in the background.

The 1979 paper is descriptive, not normative. Kahneman and Tversky showed how people decide in a laboratory, with gambles of known probability, most of them hypothetical. They did not write how one ought to decide, and they did not study traders. That distinction will matter for every number in this article.

The same number against four different reference points starting capital 10,000 against it: +1,000 this morning 11,300 against it: −300 peak three weeks ago 12,400 against it: −1,400 own target 15,000 against it: −4,000 account: 11,000 USD axis: account value in USD, linear scale from 9,800 to 15,400; illustrative values
Illustrative values. The market knows none of the four reference points – all of them belong to the trader. Different answers to "how am I doing" may change the assessment of the same position; they do not determine a decision.

The reference point moves, and rarely is there just one. There are at least five candidates: the capital deposited at the start, the account value at the beginning of the day or week, the highest value the account has ever shown, the entry price of an open position, and the figure the trader has set for themselves as a target. Each produces a different description of the same situation, and the description governs the decision.

Example: the same position, two reference points

Illustrative example; nominal values, with no reference to any specific account. A trader has an open position showing a profit of roughly half of what the plan written before entry assumed. The exit condition in that plan read: I close when price touches the planned level or when the level invalidating the scenario is breached.

  • Reference point = entry price. The result is positive, so the question "am I about to lose what I already have" appears. Decision: close now, before the market gives it back.
  • Reference point = the condition from the plan. Neither exit condition has occurred, although the financial result remains positive. Decision: do nothing and wait for a condition.

The same market data, the same setup, opposite actions. The difference does not lie in the chart analysis, but in what the result is being measured against. And one more thing: you cannot settle which decision was better by looking at what price did afterwards. That is an assessment of process, not of outcome – developed in the article on trade outcome versus process quality.

What can be recorded before the decision. The reference point, exit condition and declared state can all be recorded beforehand, but carry different evidential status. An emotion self-rating is not an objective reading, and an account statement cannot validate it. The sentence “I assess this position against the recorded invalidation level, rather than the entry price” can be timestamped and compared with subsequent execution. The record tests compliance, rather than proving a psychological mechanism.
Recovering a drawdown: a nonlinear relation. Conditions and inferential limits are part of the diagram.
Mathematical identity, not an emotion model. −20% needs +25%; −50% needs +100%. Risk details: section 9.

2. Loss aversion: a range, not a constant

Loss aversion means that a loss carries greater subjective weight than a comparable gain relative to a given reference point. It is a model concept and a pattern found in some decision tasks, rather than a universal law of experienced pain. The estimated effect’s scope and size depend on task design, method and population, so the result must be read alongside those conditions.

The popular coefficient of 2.25 does not come from the 1979 paper, though it is often attributed to it. It appeared in the later, cumulative version of prospect theory (Tversky and Kahneman, 1992), as a parameter fitted to the results of a small group of subjects[2]. The meta-analysis by Brown, Imai, Vieider and Camerer (2024), covering several hundred estimates across different studies, methods and populations, gives a value of 1.955 with a 95% Bayesian interval of [1.820, 2.102][3]. That is a pooled estimate, not a measurement of a particular trader’s parameter. Individual preferences can be estimated in separate tasks, but this average must not automatically be assigned to them.

There is also criticism of the phenomenon’s scope. Gal and Rucker (2018) argued that some classic findings can be explained by choice inertia, switching costs or question design[4]. Yechiam and Zeif’s re-meta-analysis (2025) found a coefficient close to one when tasks were restricted to symmetric gains and losses with unordered presentation[12]. This does not settle every context, but it rules out treating one pooled average as a law or a personal measure of every trader.

The value function – an illustrative diagram, not a measurement change relative to the reference point subjective weight of the outcome 0 gains losses a gain of size x a loss of size x plotted from: value of a gain = x to the power of 0.88; value of a loss = −1.955 times x to the power of 0.88 the vertical axis units are nominal and are not account percentages
A diagram, not a measurement. Both branches flatten out with distance from the reference point – that is diminishing sensitivity. The coefficient 1.955 describes the relation between values on the vertical axis, not a conversion of account losses into "percentages felt".
Value function: schematic, not an individual measure. Conditions and inferential limits are part of the diagram.
Qualitative axes. The shape illustrates a model; it neither calibrates a trader nor converts account percentages into feelings.
What must not be done with this number. The sentence "a 20% loss feels like a 39% loss" cannot be derived from any version of this theory. The coefficient is a ratio of subjective values at comparable magnitudes, not a multiplier for account percentages. What is more, diminishing sensitivity – the same theory, the same chart – says outright that sensitivity falls with distance from the reference point. Linear extrapolation therefore contradicts the very claim it supposedly invokes.

In the model shown, an additional step of the same size carries less subjective weight further from the reference point. "I am already deep in it; one more position changes nothing" may fit that interpretation, but does not establish it. Neither reduced care nor adding a position reconstructs the trader’s value function. The record can test the earlier reference point, condition and change in action; new information, costs and beliefs about the next move must also be considered. This tests a hypothesis rather than measuring a curve from one decision.

3. Loss aversion versus risk aversion

These two concepts are used interchangeably in trading texts, and that is a mistake which changes the conclusions.

Risk aversion means preferring a certain outcome to a lottery with the same expected value: in this example, the person chooses a certain 100 rather than equal chances of 200 or zero. The concept does not require classifying gains and losses relative to a reference point. It does not imply that one person will be equally cautious at every stake and in every setting.

Loss aversion describes an asymmetry around a reference point: an outcome labelled as a loss weighs more than a comparable outcome labelled as a gain. Without a reference point the concept loses meaning, because there is no telling what counts as a loss.

Prospect theory can describe a person choosing a more certain outcome in the gain domain and a riskier possibility in the loss domain. A position history may resemble this through earlier realisation of gains and longer retention of losses, but that pattern does not identify the model or mechanism. Probabilities, stakes, exit conditions and new information also matter. "Risk-averse" or "risk-loving" is too broad a label if it leaves out those conditions and the reference point.

What is not loss aversion. Reducing exposure after a run of losses can be a rational response to greater uncertainty about your own estimate of your effectiveness, rather than a symptom of fear. Conversely: holding the same exposure unchanged despite changed conditions does not prove "mental toughness". What settles it is whether the change was written down beforehand as a condition or invented along the way. The risk numbers and thresholds themselves belong to the risk management plan in section 9, not to psychology.

4. The disposition effect: what Odean actually counted

The concept was introduced by Shefrin and Statman (1985): the tendency to sell winning positions too readily and hold losing ones too long[5]. The original paper contained both theory and evidence. Terrance Odean later examined the pattern in a large account sample.

Study card. Odean (1998), Journal of Finance. Data: roughly 10 thousand randomly selected accounts at a US discount brokerage, the years 1987–1993, individual investors in the equity market[6]. Method: for every day on which an investor sold anything, count how many of the available winning positions were realised, and do the same for the losing ones. From that come two proportions: the proportion of gains realised and the proportion of losses realised.

What investors were more willing to realise Odean (1998), approx. 10 thousand brokerage accounts, USA, 1987–1993 0.00 0.04 0.08 0.12 0.16 0.148 0.098 gains realised share of available positions losses realised share of available positions share of opportunities realised
Vertical axis scale from 0.00 to 0.16, gridded every 0.04. The ratio of the two proportions is about 1.5 – which does not mean every investor in the sample behaved this way. It is a difference in frequency across the whole group.

The difference in frequency alone says only that something in the choices was systematic. The stronger part of the result concerns what happened afterwards: over the following twelve months the winning positions that were sold delivered an excess return of +2.35%, and the losing ones that were kept −1.06%. The difference is 3.4 percentage points against what the investors were holding on to.

What this result does not say. It does not say that had those investors acted the other way round they would have earned 3.4 percentage points more – reversing the decisions would have changed the portfolio, the costs and the entry timing, and the counterfactual outcome was not measured. Nor does it say why they did it: some trades had a tax motivation, some followed from beliefs about price reverting to the mean. Odean showed a regularity of behaviour and its correlate in returns, not a psychological mechanism operating inside every participant's head.

Extrapolation to the currency market – flagged as such. These were equity accounts, without leverage in today's sense, in an environment where a position can be held for years and where US rules on realising tax losses applied. Retail FX trading has a different structure: leverage, a cost of carrying a position, protective orders as standard, and a horizon counted in hours or days. The direction of the effect is credible as a hypothesis to test against your own records. The magnitude does not carry over.

It is also worth knowing that the picture of "investors reluctant to realise losses" was later complicated. Ben-David and Hirshleifer (2012) analysed the propensity to sell as a function of a position's return so far and found a V shape: it rises both at large gains and at large losses[7]. That does not fit a simple reluctance to close losing positions, and fits rather a change in beliefs about the next move. The disposition effect is therefore the name of a pattern in data, not proof that loss aversion was at work in every single case.

Counterexample: holding a loss that is not a mistake

Illustrative example. Before entry the trader wrote: the scenario is invalidated only once a level is breached that sits clearly further away than a typical swing on this instrument; until then the position stays open regardless of the current result. The position is now under water, but that level has not been breached. The trader does nothing.

From the outside, there is an open losing position and no response. The pre-decision record distinguishes compliance. If the condition existed beforehand and remains unbreached, inaction executes the plan. If the record is missing, the decision’s basis is unknown; if a condition was breached, a departure from the plan is identified. None of these facts alone establishes an emotional cause or the disposition effect in a particular person.

The criterion works in the other direction too, and this is the part that is easy to miss: closing a losing position before the condition from the plan is also a departure, even though it looks like discipline. A trader who cuts a loss "to get some peace", ahead of the invalidation level, changed the rules exactly as much as one who failed to honour them the other way.

5. Paper losses and closed losses: mental accounting

Thaler (1985) described mental accounting: people assign outcomes to separate accounts in their heads and settle them separately, instead of treating wealth as one pool[8]. A mental account closes at the moment of realisation – and that is where the belief comes from that as long as a position is open, "nothing has happened yet".

Account equity includes open positions at their current marked-to-market value. An unrealised loss does not disappear because the position remains open. Closing it realises the result and removes further exposure; it may also involve execution costs and tax consequences. It therefore does more than change mental bookkeeping. The psychological distinction concerns excluding an existing result from the account assessment, rather than economic identity between an open and a closed position.

What this looks like in behaviour rather than in declarations:

  • a weekly summary calculated from balance rather than equity, with open positions under water;
  • the sentence "this week was flat" spoken with a losing position still open;
  • a separate "experimental account" whose results do not enter any summary;
  • treating transaction costs as something outside the result, because "that is not a loss on the market".

What this should not be confused with. Valuing a position by the condition from the plan rather than by the current deviation is not mental accounting – it is simply assessment against a previously adopted criterion. Likewise, splitting capital across two accounts at a broker can be a deliberate limitation on freedom of decision rather than self-deception; the difference lies in whether the split was designed beforehand and whether both accounts enter a common summary. Designing such constraints is the subject of the course article on checklists over willpower (10.7).

Holding a loss: behaviour does not reveal cause. Conditions and inferential limits are part of the diagram.
Conceptual diagram. It does not establish motive, cause or financial outcome.

6. Breaking even and "gambling with the house money"

Thaler and Johnson (1990) studied how a prior outcome changes willingness to accept the next bet[9]. The study was laboratory-based, the participants were students, some bets were hypothetical and some were played for real, small stakes. Two effects came out of it which trading texts tend to confuse.

The house money effect: after a prior gain, willingness to accept a risky bet rises, because the result is booked as "playing with winnings" rather than with one's own wealth. The break-even effect: after a prior loss, acceptance rises for a bet that offers a chance of returning to the starting point – even if under normal conditions it would be rejected.

Carrying this across to the currency market is an extrapolation and that has to be said plainly: students in a laboratory made single decisions about small amounts, with no weeks of history behind them, no costs and no chance to learn from their own records. What carries over is the direction of the mechanism and its link to the reference point. What does not carry over is any number whatsoever.

Breaking even can be a particularly salient reference point: it allows a mental account to close without a loss. The study does not establish that this level always has more influence than every other one. The market does not know a personal entry price or recovery target. When returning to that value begins to change successive orders outside the plan, the sequence becomes the subject of the article on revenge trading and tilt. Wanting to break even does not alone establish a violation.

One sentence on the arithmetic of recovery. The deeper the drawdown, the larger the percentage gain needed to come back – and this is a non-linear relation, not a matter of attitude. The arithmetic, the thresholds and the consequences for position size belong to section 9: the math of losses and risk of ruin. In this instalment the psychological conclusion is enough: the further from the reference point, the more tempting the shortcut becomes, and the less reliable is any assessment of your own situation made at that same moment.

7. Sunk costs and escalation of commitment in cold blood

Arkes and Blumer (1985) showed in a series of experiments that costs already incurred and irrecoverable influence decisions about continuing, although from the point of view of future outcomes they should not[10]. Their most convincing piece was a field experiment: buyers of theatre season tickets were randomly granted discounts, and those who had paid the full price subsequently attended more performances. The ticket price did not change the value of any single show – it changed the sense of how much had already been invested.

Staw (1976) described a related phenomenon, and a more important one for trading: escalation of commitment. In his study, participants making a simulated financial decision allocated more funds to a course of action they were themselves responsible for, and did so particularly after feedback about failure[11]. The participants were business students and the decision was a simulation – but the mechanism concerns responsibility for an earlier choice, not arousal.

The distinction matters. Adding to a losing position does not require a visible outburst of frustration. It may accompany reported calm and an elaborate rationale, but that establishes neither an absence of emotion nor greater difficulty recognising it. The checkable fact is further commitment without a previously permitted condition or recorded new information. "I was upset" does not settle compliance; neither does "I was calm".

What this should not be confused with. Information that came into being after the position was opened can genuinely matter for the future outcome – that is not a sunk cost. Scaling a position written into the plan before entry is not escalation either, even if the next tranche goes in at a worse place. A sunk cost concerns only what has already been spent and cannot be recovered: money, learning time, months spent on one approach.

Where this stops being a subject for a course. There are behaviours for which the question about a reference point is already beside the point: topping up an account with funds that were not meant for it, including from a credit card or a loan; hiding the scale of trading from people close to you; opening another account right after closing the previous one; circumventing blocks you set for yourself earlier. These are not symptoms of a "weak psyche" nor material for another procedure – and this article makes no diagnosis, because it has neither the data nor the standing for one. If you recognise any of these behaviours in yourself, the right address is a family doctor, a psychologist or a psychotherapist, and, where available, a local behavioural addiction treatment service. Classifications such as ICD-11 are invoked here solely to mark where the competence of a course about decision-making ends.
Comparison of past learning and decision costs with current evidence relevant to the next order.
A cost already incurred does not change future outcomes. New market information may matter; previously planned scaling is not in itself escalation of a mistake.

8. Behaviour, mechanism and the simpler explanation

Naming a bias after the fact proves nothing. Before pinning a label from this article onto a behaviour, you have to rule out explanations that require no psychology at all: market conditions, the instrument, the time of day, volatility, the way the platform behaves. The table below organises that distinction – the fourth column shows what settles the dispute between the second column and the third.

Observable behaviourMechanism from this articleSimpler explanation to be ruled outWhat settles it
Closing a winning position before the condition from the planReference point = entry price; disposition effectA previously written condition for partial realisation; a news item changing the scenario's assumptionsThe record of the exit condition made before entry
No reaction to a position under waterReluctance to book the lossThe invalidation level sat further away than you remember; the order was rejected or never arrivedThe recorded invalidation level and the order log
A further commitment in the same direction after a lossEscalation of commitment, sunk costsScaling written into the plan before the first entryA scaling plan dated earlier than the first order
Rejecting a signal that met the conditions right after a lossAsymmetry around the reference pointThe signal failed one of the conditions; the volatility filter worked correctlyA checklist completed before the decision, not after
A weekly summary without open positions under waterMental accountingThe report shows balance by default, not equityThe report setting and the field definition in the journal
Increasing commitment as the account approaches its starting valueThe break-even effectMarket conditions genuinely improved against the criteria in the planEntry criteria written down before the session

No row in this table is settled by the outcome of a trade. A position closed earlier than the plan foresaw may have ended better than the plan – and it was still a departure. That distinction matters enough to have its own instalment in the course: outcome versus the decision process.

9. How to record the reference point so it can be checked

Everything above comes down to one practical problem: the mechanism operates before the decision and is recognised afterwards. The only way out is to record the part of the situation that is checkable – namely, what the result is being judged against.

Below is the full form of the procedure. It is not a journal template nor a set of risk rules; each of those has its own place in the course (the trading journal and the risk management plan).

FieldContent
Trigger conditionEvery new order and every decision to change an existing position.
ActionOne sentence before the order is placed: what you are judging this position against (the invalidation level and the exit condition) and which account value you are treating as the reference point this week.
Completion criterionThe entry closes at the moment the order is placed. It is not edited once the outcome is known – after-the-fact commentary goes into a separate field.
Audit trailThe timestamp of the entry earlier than the timestamp of the order in the broker's history. Without that, the entry is indistinguishable from a rationalisation.
Failure criterionAfter thirty closed positions, count in how many cases the entry lets you settle whether the decision matched the condition. If that share is not rising, the field is badly designed – usually too general. That is a flaw in the tool, not in your character.

Two notes on what this procedure does not give you. It does not remove the mechanism: recording the reference point does not stop it operating, because awareness of a bias is no protection against it. What it does give you is data on which, after a few dozen decisions, you can check whether departures form a regularity or are noise. And it does not replace an assessment of the strategy itself – if the method has no foundation, the best-recorded reference point will fix nothing.

The nearest test is in fact sooner than it seems: at the first position that goes against you in a week in which the account is already below its peak. That is when it turns out whether the morning entry contains a condition or only a hope. The article on checklists over willpower (10.7) deals with how to build such conditions so that they survive the moment when their author is no longer the same calm person who wrote them.

Reference point recorded before the decision. Conditions and inferential limits are part of the diagram.
Failure review: can the record establish compliance? No increase alone is not proof of poor design.

FAQ – the reference point, loss aversion, the disposition effect

Does a loss really hurt twice as much as an equal gain pleases?
Not as a universal conversion factor. Brown et al. (2024) estimated a mean coefficient of 1.955 with a 95% Bayesian interval of [1.820, 2.102]; 2.25 comes from a 1992 paper. These are model estimates in the tasks studied, rather than measures of an individual trader’s pain. Critical reviews and the 2025 re-meta-analysis show that the finding also depends on task design. Do not multiply it by an account-loss percentage or assume equal asymmetry in every situation.
I am holding a losing position. Is that the disposition effect?
Holding a position or its later outcome is not decisive. If the invalidation level was recorded beforehand and remains unbreached, inaction may execute the plan. A missing condition means the basis for assessing compliance is missing, rather than automatically proving loss aversion. Breaching an existing condition is checkable; its cause needs separate evidence and simpler explanations must be ruled out. The disposition effect describes a pattern across decisions, not a diagnosis of one position.
Is an unclosed loss already a loss?
Yes. Account equity includes open positions, because a position is worth what the market pays for it now. The only thing that is "on paper" is the accounting inside your head: the mental account closes only at realisation, so until then the result tends to be left out of summaries. A simple test: count your week once from balance and once from equity. If those two numbers differ, you already know which of them you usually quote to yourself.
Are loss aversion and caution the same thing?
No. Risk aversion is a preference for a certain outcome over a lottery with the same expected value, and it needs no reference point. Loss aversion is an asymmetry around a reference point. The same person can close gains early while accepting worse bets when under water – prospect theory expects exactly that. Describing such behaviour as "caution" or "risk appetite" loses what is actually happening.
After what size of loss should I stop trading?
That is a question about numbers, and numbers do not belong to psychology. Thresholds, limits and position size are the subject of the section on risk management, while the moment and the manner of breaking off a session are described in the article on reset protocols. This instalment answers the earlier question: what are you counting that loss against, given that the account has several possible reference points at once and the market knows none of them.
Is topping up the account after a big loss always a mistake?
The deposit itself is neither a mistake nor proof of one – what matters is when it was planned and where the funds come from. A deposit written down in advance as part of a funding schedule is a different thing from a deposit whose only reason is the size of the loss so far; the second is a textbook sunk cost, because money already spent does not change expectations about future outcomes. If, however, the top-up comes from funds not earmarked for the purpose, including a loan or a credit card, this stops being a question about a reference point – and then the right address is a doctor, a psychologist or a behavioural addiction clinic.

Sources and references

  1. Kahneman D., Tversky A., "Prospect Theory: An Analysis of Decision under Risk," Econometrica, 47(2), 1979, pp. 263–291. The reference point, diminishing sensitivity, the asymmetry of gains and losses. A descriptive paper based on mostly hypothetical gambles; it contains no loss aversion coefficient.
  2. Tversky A., Kahneman D., "Advances in Prospect Theory: Cumulative Representation of Uncertainty," Journal of Risk and Uncertainty, 5(4), 1992, pp. 297–323. Source of the 2.25 parameter, fitted to the results of a small group of subjects.
  3. Brown A.L., Imai T., Vieider F.M., Camerer C.F., "Meta-Analysis of Empirical Estimates of Loss Aversion," Journal of Economic Literature, 62(2), 2024. Pooled estimate of the coefficient: 1.955, 95% Bayesian interval [1.820, 2.102].
  4. Gal D., Rucker D.D., "The Loss of Loss Aversion: Will It Loom Larger Than Its Gain?," Journal of Consumer Psychology, 28(3), 2018, pp. 497–516. A critical review of the evidence for the generality of loss aversion.
  5. Shefrin H., Statman M., "The Disposition to Sell Winners Too Early and Ride Losers Too Long: Theory and Evidence," The Journal of Finance, 40(3), 1985, pp. 777–790. Introduction of the disposition effect.
  6. Odean T., "Are Investors Reluctant to Realize Their Losses?," The Journal of Finance, 53(5), 1998, pp. 1775–1798. Roughly 10 thousand accounts at a US discount brokerage, 1987–1993; proportions of gains and losses realised 0.148 against 0.098; excess returns over the following 12 months +2.35% against −1.06%.
  7. Ben-David I., Hirshleifer D., "Are Investors Really Reluctant to Realize Their Losses? Trading Responses to Past Returns and the Disposition Effect," The Review of Financial Studies, 25(8), 2012, pp. 2485–2532. A V-shaped propensity to sell as a function of a position's return so far.
  8. Thaler R., "Mental Accounting and Consumer Choice," Marketing Science, 4(3), 1985, pp. 199–214. Mental accounting and the moment a mental account is closed.
  9. Thaler R.H., Johnson E.J., "Gambling with the House Money and Trying to Break Even: The Effects of Prior Outcomes on Risky Choice," Management Science, 36(6), 1990, pp. 643–660. A laboratory study on students, stakes partly hypothetical; the house money effect and the break-even effect.
  10. Arkes H.R., Blumer C., "The Psychology of Sunk Cost," Organizational Behavior and Human Decision Processes, 35(1), 1985, pp. 124–140. A series of experiments together with a field experiment on theatre season-ticket buyers.
  11. 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. A simulated financial decision, business students; escalation of commitment after feedback about failure.
  12. Yechiam E., Zeif D., “Loss Aversion Is Not Robust: A Re-Meta-Analysis”, Journal of Economic Psychology, 107, 2025, 102801. Re-analysis; the estimated effect depends on amount symmetry and presentation order. Authors’ full text.

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.