Sunk Cost Fallacy: Why You Keep Holding Losers Long After You Should Exit

You're down 20% on a trade. The thesis you originally had no longer holds. The structural reason for the trade has been invalidated. You should exit. But you don't, because exiting means realizing the loss, and realizing it means accepting that the time, attention, and capital you invested in the trade is gone. So you hold. The price drops further. The sunk-cost fallacy has just turned a contained loss into an uncontained one.

What sunk cost actually is

A "sunk cost" is anything you've already paid that you can't get back. The capital in a losing trade is sunk: whether you exit at -20% or -40%, the capital that's gone will be gone either way. The decision about whether to hold or exit should be based purely on the trade's forward expected value, not on what you've already invested.

This is the rational rule. It's also psychologically extremely hard to follow. Humans evolved to be loss-averse, the pain of loss is roughly 2x the pleasure of equivalent gain. So exiting at -20% feels far worse than continuing to sit in a position that might recover. The "might" is the trap.

The trading examples

The held-too-long loser. You bought ETH at $3,000 with a stop at $2,800, a 6% risk on your account. ETH falls to $2,800. You don't take the stop; "it's just a wick, will recover." ETH falls to $2,500. Now you're at -17%. You think "I can't sell here, I'm too far underwater, has to bounce." ETH falls to $2,200. You sell finally at -27%, more than 4x your planned 1R loss. Sunk cost fallacy converted -1R into -4R.

The averaging-down spiral. You're long an altcoin that's down 30%. You think "if I add here, my average is much better." You add. The token drops another 30%. Your "improved average" is now -45% on a much larger position. You add again. Each add increases the position size and pushes the recovery requirement further out. You're throwing good capital after bad to avoid admitting the original trade was wrong.

The bag-holder convert. You bought a token in 2021 at $50. It's now $2. You're down 96%. You tell yourself "I'll just hold long-term, it'll come back." It probably won't, but you can't sell because that would mean admitting the entire investment was a mistake. The token sits in your portfolio for years, occupying attention and capital that could be deployed elsewhere.

In each case, the reasoning about the trade has stopped being about forward expected value and started being about the prior commitment.

The psychological dynamics

Several specific patterns drive sunk-cost behavior:

Loss aversion. Realizing a loss feels much worse than not realizing it. Holding keeps the loss "unrealized" and therefore psychologically smaller, even though it's financially identical.

Identity protection. Exiting a losing trade is admitting you were wrong. For traders who tie self-image to being right, the exit becomes an identity threat. The brain prefers to maintain the fiction that the trade is still "working", even at the cost of further capital.

The recovery hope. "If it just gets back to break-even, I'll sell." This frames exit as conditional on recovery, which sounds reasonable but ignores that the same logic that drove the price down may continue to drive it down.

Effort justification. You did research. You wrote up the thesis. You committed to the position publicly (twitter, chat, friends). All of that effort is sunk cost, but the mind doesn't want to invalidate it by admitting the trade was wrong.

Endowment effect. You feel ownership of the position. The asset is "yours." Selling feels like loss in a way that not- buying never would. This is why investors who own a stock at $100 won't sell it at $80 but also wouldn't buy the same stock at $80 if they didn't already own it.

These dynamics combine. Each one alone might be manageable. Together they create powerful inertia against exit.

The right rule

The decision rule that defeats sunk cost: at any moment, ask yourself, would I open this position now, at the current price, with the current evidence, if I had no prior commitment? If the answer is no, exit. If the answer is yes, hold.

This single question reframes the decision from "acknowledge the loss" to "evaluate the trade fresh." Most sunk-cost held losers fail this test instantly. You wouldn't buy the position now, you're holding it because of what you've already invested, not because of forward expected value.

The question is hard to ask honestly. The bias resists. But asking it explicitly, in writing, with evidence, is a structural defense that bypasses the emotional dynamics.

Structural defenses

Pre-committed stops. The single strongest defense. If your stop is in place and reduce-only, the exchange executes the exit before sunk-cost reasoning can take over. Bracket orders on entry. No manual moving of stops once placed.

Hard rules against averaging losers. "I do not add to losing positions, ever." Make it identity-level. Adds happen only on winning positions, with their own R-based sizing. This single rule eliminates the averaging-down spiral.

Periodic position review. Once a week, ask the "would I open this now?" question for every open position. Document the answer. Positions that consistently fail the test get closed in the next session, regardless of unrealized loss.

Maximum drawdown per position. Even with stops, build a hard cap: "no individual position drawdown exceeds X% of account." If you're somehow past your stop and approaching the cap, the position is closed. Period.

Disclosure tracking. If you publicly share trades (twitter, journal, chats), explicitly disclose exits. The act of publishing the exit forces you to confront the sunk-cost defense. It also keeps you accountable to the same standards you set for yourself privately.

A common mistake: confusing patience with sunk-cost

Patience is a virtue in trading, letting trades develop, not exiting on every wiggle. Sunk cost is the opposite: not exiting despite the trade being invalidated. They feel similar from the inside.

Distinction: patience holds because the thesis is intact and the trade is moving as expected (or hasn't moved much). Sunk cost holds because the trade is failing and the holder can't bring themselves to exit.

Test: "Is the original thesis still valid given current evidence?" Yes → patience. No → sunk cost. Honest answer required.

A common mistake: re-justifying with new theses

You bought BTC because of an expected breakout. The breakout failed. You should exit. Instead, you "find" a new reason to hold, "the long-term fundamentals are still strong" or "the on-chain accumulation is continuing." The new reason wasn't your reason for buying; it's a post-hoc rationalization to avoid the exit.

The honest move: each trade has one reason for being open, and that reason was set at entry. If the entry reason is invalidated, the trade exits. New reasons might justify new trades, but they don't justify keeping a current trade open under different theses.

A common mistake: cycling the asset

You exit a losing trade. A few days later, you re-enter the same asset because "now it looks better." If the new entry has a real, independent setup, that's fine. If you're re-entering because you "owe" yourself a recovery on this specific asset, that's sunk cost re-asserting itself with a fresh wrapper.

Test: would you take this entry on a different asset at the same probability? If you'd skip it on a different asset and only take it because it's the asset you previously lost on, it's sunk cost in disguise.

Mental model, sunk cost as the casino's mistake

Imagine a gambler who's lost $1,000 at the blackjack table. They keep playing because they "can't leave on a loss." Each new hand they play is an independent hand with independent odds. The $1,000 is gone whether they play 100 more hands or 0. Continuing to play because of the prior loss is the sunk-cost fallacy.

Now imagine a casino doing the same thing, running a game that's losing money because "we've already invested in the equipment." The casino would (correctly) shut down the game and reallocate the floor space to a profitable game. They don't honor sunk costs because they treat capital deployment as forward-looking. You should too.

Why this matters for trading

Sunk-cost fallacy is the mechanism that converts most planned 1R losses into 3R-and-larger losses. Bracket orders eliminate the most common version (failure to honor a pre-set stop) by mechanically executing the exit. Weekly position reviews eliminate the slower version (held-too-long winners that quietly turned into losers). Hex37's journal page makes visible patterns in your trading, including the duration and depth of held losers, so you can confront the data.

Takeaway

Sunk cost is the bias toward continuing a trade because of prior investment, even when current expected value says exit. Loss aversion, identity protection, and effort justification combine to produce powerful resistance to realizing losses. The defense is the question: "Would I open this position now, at this price, with this evidence?" If no, exit. Structural defenses include hard stops, no averaging losers, periodic review, max drawdown caps, and disclosure tracking. The capital you've already lost is gone either way, only the forward decision matters.