The disposition effect is a tendency to realize gains more readily than losses; sunk-cost reasoning uses money, time or effort already spent to justify continuing a commitment. In crypto trading, the two can overlap when someone sells a winner while keeping a loser because they want their original money back. Review the current exposure, available alternatives and future costs separately from the purchase-price label.
Two questions that should stay separate
The disposition effect describes a selling pattern. Sunk-cost reasoning describes a justification for continuing. A trader could show either one without the other. Someone might quickly sell profitable positions for the satisfaction of booking gains without mentioning past effort. Another person might keep funding an unused trading tool because configuring it took weeks, even when no winning or losing position is involved.
In Terrance Odean's original brokerage study, the source records covered 10,000 accounts from 1987–1993. Of those, 6,380 accounts traded stocks covered by the study's price database. The analysis found a preference for realizing gains over losses. That is evidence about the studied stock accounts, rather than a rule that every losing crypto position should be sold.
Arkes and Blumer's original sunk-cost research examined how prior investment of money, effort or time influenced continuation decisions, using questionnaire experiments and a theater-ticket field study. Applying that distinction to a trading review is a proposed decision aid; the experiments did not test this worksheet or a crypto exit strategy.
An entry price remains necessary for accounting. Depending on the situation, historical transactions can also affect future taxes or contractual obligations. Those consequences belong in the comparison. “I have already spent so much” by itself supplies no evidence about what happens next.
A profitable sale can leave the account down
Consider a hypothetical unleveraged spot account holding two fictional assets. All amounts are USD. Initial purchase fees are excluded; each sale has an assumed 0.20% fee deducted from its cash proceeds. Prices fill exactly as shown. Taxes, slippage and interest are excluded. These assumptions are illustrative, with no claim about a venue's actual fee schedule.
| Holding | Purchase cost | Current marked value | Unrealized result |
|---|---|---|---|
| 10 units of A, bought at $100, now $80 | $1,000 | $800 | −$200 |
| 10 units of B, bought at $50, now $60 | $500 | $600 | +$100 |
| Total | $1,500 | $1,400 | −$100 |
The account is already $100 below its purchase cost at current marks. Compare four choices at this same instant, starting with no cash:
| Choice | Cash after sale fees | Remaining holdings at current marks | Total equity |
|---|---|---|---|
| Hold both | $0 | A: $800; B: $600 | $1,400 |
| Sell all B; keep A | $598.80 | A: $800 | $1,398.80 |
| Sell all A; keep B | $798.40 | B: $600 | $1,398.40 |
| Sell both | $1,397.20 | None | $1,397.20 |
Selling B produces $600 gross proceeds and a $1.20 fee. Its realized gain after that fee is $98.80. A still carries a $200 unrealized loss. Combined, the account is down $101.20, matching $1,398.80 of equity against the original $1,500.
Selling A instead realizes a $201.60 loss after its $1.60 fee, while B retains a $100 unrealized gain. The account is down $101.60. The $0.40 difference between the two sale choices comes from the larger sale notional and fee. Moving a loss from the unrealized column into the realized column does not create that price loss anew.
This is where a disposition preference can mislead a review: a positive closed-trade result looks reassuring while most of the remaining exposure sits in the losing asset. But the ledger alone does not show which sale is justified. Selling B might meet a planned exposure limit; retaining A might have a defensible forward rationale. Those reasons need their own evidence.
Compare what each choice still exposes you to
The next table continues the same hypothetical choices under two deliberately opposite future scenarios. Cash stays constant. Remaining assets are marked at the new prices without further sales, so no future liquidation fees are included. Neither scenario has an assigned probability.
| Choice made today | Later A = $72, B = $66 | Later A = $88, B = $54 |
|---|---|---|
| Hold both | $1,380 | $1,420 |
| Sell B; keep A | $1,318.80 | $1,478.80 |
| Sell A; keep B | $1,458.40 | $1,338.40 |
| Sell both | $1,397.20 | $1,397.20 |
For example, selling B leaves $598.80 cash and ten A units. If A later falls to $72, equity becomes $598.80 + $720 = $1,318.80. If A rises to $88, it becomes $1,478.80. The earlier realized gain on B does not protect those A units.
These choices retain different amounts and types of exposure: $800 in A, $600 in B, $1,400 across both, or cash alone. They are not matched-risk strategies, and the table cannot rank their expected returns. It makes the consequences visible before a choice is justified using a green or red P&L label.
The same care applies to “waiting to break even.” From $80, A would need a 25% price increase to return to its $100 purchase price before sale costs. The purchase price establishes that arithmetic reference; it does not establish a reason the increase will occur.
Test the reason for continuing
Use a paired review with the same current position and different fictional histories. One record says ten A units cost $1,000; another says they cost $600. In both, A now trades at $80, the market evidence is identical, and holding limits, future costs and obligations are identical. Exclude tax differences for this exercise.
Each record has an $800 current asset value and $798.40 available from the assumed immediate sale. Only the historical gain or loss changes. Write the chosen future exposure before comparing the entry prices. If the preferred action changes solely because one record shows a loss, investigate that difference. It is a prompt for review, not proof of a psychological condition.
Be precise about sunk costs here. A's current $800 market value is recoverable through a sale before costs; it is not all sunk. Historical expenditure cannot be undone, while the decision still controls whether that current value remains exposed. Similarly, six hours spent researching cannot be recovered, but useful information learned during those hours can still matter. Separate the effort from the evidence it produced.
Ask what new evidence, current constraint or future consequence supports continuing. “The thesis still passes its specified tests” is reviewable. “Selling would waste my research” is an appeal to past effort. A planned purchase schedule should also be assessed against its written purpose and budget; a price below average entry alone does not make another purchase necessary.
Count opportunities as well as completed sales
A journal containing only closed trades cannot show which profitable and losing positions were left open. Odean's realization-rate methodology compares realized gains with realized plus paper gains, and does the equivalent for losses. The original study counts paper positions on sale days; a personal review must declare its own observation rules.
For a simplified hypothetical review, preselect decision timestamps and use purchase cost as the reference. Here, prices stay unchanged during each decision, gains and losses exclude fees, and the sample includes only full exits or untouched positions. Record partial exits separately and exclude exact break-even cases. Count each eligible position once per timestamp. Suppose the resulting observations are:
| Classification | Sold | Kept open | Proportion realized |
|---|---|---|---|
| Gain | 6 | 9 | 6 ÷ 15 = 40% |
| Loss | 2 | 8 | 2 ÷ 10 = 20% |
The gap is 20 percentage points. This snapshot measure is not a replication of the brokerage study, a significance test or a diagnosis. Repeated observations of the same holding are dependent. Ineligible assets, missing purchase records, different strategies and cash needs can also distort a comparison. Retain the action reasons alongside the counts.
Crypto evidence needs similar restraint. Schatzmann and Haslhofer's Bitcoin study used transfers to exchanges as selling proxies and found a disposition pattern that varied across periods. The authors identify unseen off-chain trades as a limitation. An exchange transfer does not establish an individual trader's eventual sale or motive.
A decision ledger for the next review
Attach this short record to the relevant crypto trading plan. The purpose is to make the reason inspectable before the outcome is known.
| Field | What to write |
|---|---|
| Current state | Units, market value, available cash, working orders and timestamp |
| Proposed action | Exact quantity to retain, reduce or add; estimated future costs |
| Alternatives | Keep, reduce, close or defer; resulting exposure for each |
| Forward reason | Current evidence, constraint or obligation supporting the choice |
| History check | Would the choice change with a different entry price or less prior effort? Why? |
| Review trigger | Evidence that would invalidate the reason, plus the next review time |
Complete the forward-reason field before reading an AI-generated recommendation. Then compare its evidence with the recorded criteria. The automation-bias review method explains how an approval can become a rubber stamp even when a human clicks it.
If the decision changes an automated position, translate it through the DCA manual-intervention handoff. A sound reason does not reconcile orders or prevent two controllers from acting on the same units. Keep the decision record free of credentials; the environment-variable versus vault guide addresses the separate lifecycle of trading secrets.
Finally, compare the recorded reason with the information available at the time, not only the subsequent return. A recovery does not validate “I had already spent too much,” and a further decline does not prove that every hold was irrational. Before the next change, write the exposure that remains and the evidence for accepting it. If the only justification is the money or effort already spent, the decision ledger still needs a forward reason.