Anchoring Bias in Trading: Signs and a Practical Check

Anchoring bias in trading is the tendency for a salient number a trader has already seen — an entry price, a prior high or low, a round number, or the first price quoted — to exert more influence on a current decision than its actual evidentiary relevance justifies. The reference number itself is not the problem: an entry price can be a legitimate risk input, and a prior high can reflect real technical structure. Anchoring is what happens when the weight a number receives outruns what it currently supports — not only when a number has become entirely irrelevant, but also when it remains partly relevant and still gets more say in the decision than that partial relevance earns. The number does not have to represent a loss. A trader can anchor on a price while sitting on an open gain, while deciding whether to re-enter a name they already sold, or while judging whether an unrelated instrument looks “cheap” or “expensive” relative to a level that has no bearing on its current value.

That is a narrower and different claim than “losses feel worse than gains.” Loss aversion is about the asymmetric weight a loss carries relative to an equivalent gain, evaluated against a reference point — it does not require the loss to be realized, only possible relative to that reference. Anchoring operates through a fixed reference value itself, regardless of whether accepting the current price would produce a gain or a loss. A trader can anchor while making money, and can anchor on a price that was never their own entry at all.

What is anchoring bias?

Anchoring is one of three heuristics Amos Tversky and Daniel Kahneman described in their original account of judgment under uncertainty: numerical estimates tend to stay close to an initial value — the anchor — and adjustments away from it tend to be insufficient, even when the anchor is arbitrary. In their classic demonstration, participants who watched a rigged wheel of fortune land on a number, then estimated an unrelated quantity, gave systematically higher estimates when the wheel showed a high number than when it showed a low one — despite knowing the wheel was random.¹

That experiment used an artificial, admittedly irrelevant anchor and still moved judgments. A trading anchor is rarely that arbitrary — an entry price or a prior swing high at least once meant something — which is part of why it is harder to recognize as a distortion in the moment. What matters for diagnosing a specific case is not whether the anchor was meaningful when it was set, but a narrower, diagnostic question: does the reference value receive influence on the current judgment that is justified by its relevance to that judgment right now?

Later research narrowed the explanation for why adjustment tends to fall short, but for one specific case: anchors a person generates themselves, by starting from a related value they already know and adjusting away from it. For that self-generated case, adjustment tends to stop once a value that feels plausible is reached, rather than continuing to the value the evidence actually supports — the search ends at “good enough,” not at a checked best estimate.² That stopping-at-plausible mechanism was demonstrated for self-generated anchors specifically; it should not be assumed to explain every case where an externally supplied anchor — a quoted price, an analyst target, a number someone else mentions — shapes a judgment, since the adjustment process for an anchor a person did not generate is not the same experiment. Applied to a chart, a trader does not need to consciously decide “I will use my entry price as a target.” An anchor can shape where a target, an exit, or a “cheap versus expensive” judgment lands without the trader ever treating it as an explicit input — but not every anchor in trading works through the same mechanism, and this account comes from laboratory estimation tasks rather than a demonstrated explanation of any specific trading decision.

Anchoring, loss aversion, and the disposition effect are not the same mechanism

These three terms get used interchangeably in trading discussion, but they describe different distortions with different ownership boundaries here.

Mechanism	What drives the distortion	Does it require an existing, realized loss?	Owning article
Anchoring	A fixed reference value — any salient number — keeps pulling a current judgment toward it	No. An anchor can bias a decision on a profitable position or on an instrument the trader has never held	This article
Loss aversion	A loss carries more psychological weight than an equivalent gain, evaluated relative to a reference point	No. The loss can be anticipated rather than realized — a trader does not need an open losing position to weigh a possible loss more heavily than an equivalent possible gain	Loss aversion in trading
Disposition effect	An observed pattern of selling winners sooner and holding losers longer	Describes an outcome pattern across trades, not a single judgment, and is not on its own conclusive evidence of loss aversion or any other single mechanism	Covered inside loss aversion in trading
Recency bias	The most recent outcome is given more evidentiary weight than it deserves when scoring the next setup	No, but the distortion is about a recent event, not a fixed price	Recency bias after a trading loss

Anchoring and loss aversion are not mutually exclusive, and neither can be diagnosed from a single decision or its outcome. A reference point — an entry price, a prior high — can function as a loss-aversion trigger and an anchor on the same trade at the same time. Anchoring can also show up on its own in a winning trade, a re-entry decision, or a read on a symbol the trader has never owned, and loss aversion can operate on a decision where no position is even open yet, such as sizing down a new entry because the possible loss looms larger than the equally possible gain. A pattern that happens to recur only around decisions where a loss is possible is not, by itself, proof that loss aversion rather than anchoring is the mechanism — a reference point that is driving an anchoring effect can also sit near a loss threshold. Review can only work from what is suggestive, not what is definitive: an unsupported reliance on a numerical reference can suggest anchoring, while a disproportionate weighting of equivalent gains and losses can suggest loss aversion. Contemporaneous decision notes are evidence toward one reading or the other, but notes alone cannot establish which psychological mechanism actually produced a given decision — anchoring can shape a judgment without the trader ever recognizing or writing down that a reference number was involved, and a record’s silence on the point proves nothing either way. Both mechanisms can also be present in the same decision at once, in which case review should say so rather than forcing a single label.

Where trading anchors come from

An anchor does not have to be a number the trader chose deliberately. Common sources include:

Anchor type	Example judgment it distorts	Circumstances that can make it misleading as a current input
Own entry price	”I’ll sell when it gets back to what I paid”	Break-even is a personal cost basis; treated as a market signal about where price should go next, rather than as a risk or tax input, it carries no such signal
A prior high or low	”It’s cheap — it’s way off its high”	A past extreme may reflect real technical structure, but distance from it alone says nothing about current supply, demand, or catalysts
Round numbers	Orders and mental targets cluster at levels like $50 or $100	Round-number clustering is a documented market effect, not just a mental one, but being psychologically salient does not by itself make a specific round level valid for a specific trade
Price at a prior sale	”I already sold this once at $40, I won’t buy it back above that”	The prior sale price reflects a past decision under past information; treated as a ceiling on current value rather than as one input among others, it can misstate current value
First number seen	An analyst target, a headline price, or the quote when the trader first looked at the chart	Being first is not the same as being accurate or current
Session or account starting balance	Treating the day’s opening equity as the reference point for “up” or “down”	Opening equity may legitimately define a daily drawdown limit under a documented risk policy; absent that policy, the print itself has no evidentiary claim on what “down” requires as a response

None of these anchors are inherently forbidden to notice — a prior high can be genuine technical structure, a round number can coincide with real order concentration, and an entry price is a legitimate input to a risk calculation. A number becomes a problem when the influence it has on the decision exceeds its actual evidentiary relevance — when it substitutes for an evaluation of current conditions rather than informing one. Noticing the number is not the distortion; a trader still needs an independent reason, tied to their own method, for assigning it a role.

Can anchoring help explain the 52-week-high effect?

Anchoring is not only a laboratory finding about arbitrary numbers, and one line of market research asks whether it shows up in prices themselves. George and Hwang examined a large historical sample of U.S. equities and found that a stock’s nearness to its 52-week high explains a large share of the profitability of momentum strategies in that sample — stocks trading close to their 52-week high tended to keep outperforming those farther from theirs.³ That empirical association is what the data show. The authors’ proposed explanation, not a directly observed fact, is that investors use the 52-week high as a reference point and are slow to incorporate new information as price nears it — a market-level anchoring story. The study cannot establish that any individual trader is anchoring on a given day, and it describes a specific historical U.S. equity sample rather than a universal law across markets, asset classes, or time periods. Treat it as evidence that the anchoring mechanism can plausibly show up beyond contrived experimental tasks, not as a proven, ongoing drag on price discovery or a contemporary trading edge.

A check before treating a reference price as a signal

When a judgment leans on “it’s near its old high” or “I won’t buy it back above what I sold it for,” the task is not to predict whether the anchor will hold. It is to test whether the anchor is actually doing evidentiary work. This is a narrower, decision-point version of the same question how to test for cognitive bias in a trading review applies afterward to a completed review classification: is the verdict resting on something it wasn’t supposed to use?

Ask:

What is the reference number, and where did it come from? Name it specifically — entry price, a prior high, a round number, a headline figure — rather than leaving it as an unexamined feeling that the price is “high” or “low.”
Does the trading plan assign this number a defined role, such as a technical level with a stated invalidation, or is it being used only because it is familiar?
Would the judgment change if the reference number were different, with every other current condition unchanged — and if so, is that because the number is doing unsupported work, or because it is a legitimate input, such as a technical level, a defined risk threshold, or a cost-basis figure, that is supposed to move the judgment when its value changes? This is a prompt to look closer, not a controlled experiment; a shift in judgment alone does not prove the reference was acting as an unsupported anchor.
What does the current evidence say on its own, evaluated as if this were the first time the trader had ever looked at the instrument?
Is the reference number actually relevant to the decision being made now, such as a real cost-basis or tax input, or is it only psychologically salient?

These questions test whether a number belongs in the decision, not whether the decision is correct. A prior high can be legitimate resistance under a trader’s own method, and a round number can coincide with real order concentration; the check exists to separate that defensible use from the number being followed simply because it is memorable. Answering the five questions once is not a diagnosis — it is a prompt to look, not proof of what the trader finds.

A three-point workflow around the decision

Turning the check into a repeatable process means attaching it to three points, not only the moment something feels off:

Before the decision. When a reference price first enters the plan — a target, a stop, a level to watch — write down which number it is and what role it is meant to play, before the position is open or the decision is due.
At the decision point. Evaluate the current setup, risk constraints, and invalidation criteria on their own terms — independent of historical cost basis or a remembered high/low where those have no defined role in this decision — then compare that read against the role the reference price was assigned in the first step.
After the decision. Compare the actual, stated rationale against the original plan and note whether an unsupported reference price appears to have shaped the outcome, using the review categories described below.

This sequence is a process for reducing decisions where an unexamined number substitutes for a trader’s own criteria. It does not eliminate anchoring and does not prove any single decision was biased; it gives the trader a consistent place to record the judgment so a pattern, if one exists, becomes visible over time.

A worked example: two anchors, one trader

Consider a hypothetical trader who bought a stock at $60. It rallies to $95, and along the way the trader adopts an informal target of $100 — a round number with no defined technical basis — without first checking whether $100 corresponded to anything the strategy’s own exit rule would actually have produced. Price stalls at $95, short of $100, then reverses to $80. The trader describes this as “giving back the move to $100,” even though price never reached $100 and no rule had ever defined it as the exit.

Two separate things are tangled in that reaction. The $100 figure was a missed, informal target, never reached and never rule-based — its absence is not a loss of anything the plan promised. The retracement from $95 to $80, by contrast, is a real, measurable give-back of unrealized profit, and whether holding through it was a deviation depends on what the trader’s actual exit rule specified at $95 and at $80 — not on the fact that the outcome was worse than selling earlier would have been. If the strategy’s own invalidation condition was reached at $80 and the trader exited there, that is a rule-followed exit that happens to look bad in hindsight; if the rule would have triggered earlier and the trader held anyway while anchored on $100, that is a rule-inconsistent exit. The outcome alone does not distinguish these two cases.

Later, the same stock falls to $55, below the trader’s original $60 entry. The setup that originally justified buying it reappears, cleanly, under the trader’s own criteria. The trader passes, reasoning that buying back below where they first bought “feels wrong.” For this hypothetical, assume the new setup meets the strategy’s predefined entry criteria and that no independently relevant information or risk constraint invalidates it. Under those stated conditions, the $60 entry price has no independent bearing on whether the current setup is valid; a different read could be reasonable if some other consideration tied to that earlier entry, such as an unresolved reason the position was originally closed, actually applied here.

The two anchors come from different sources, not a shared origin: the $100 figure is a salient round number with no connection to the trader’s own history in the stock, while the $60 figure is the trader’s actual historical cost basis. What they have in common is not where they came from but what they did — neither figure came from the trading plan’s own criteria, yet each still shaped a decision. A rule-based version of the same two moments would ask what the exit rule and the entry rule actually specify, independent of what the stock once cost this particular trader.

Review reference-price decisions over a sample, not one trade

One instance of leaning on an old price is not evidence of a pattern, and a quick read of a trade note cannot reliably tell rule-based reference use apart from unsupported anchoring — both can mention the same number. A useful review works at the level of the individual decision event, not the trade as a whole: a single trade can contain a rule-supported entry and a separately anchored target revision, or a defensible exit followed by an anchored re-entry decision.

The population under review should be every target, exit, entry, and re-entry decision in the sample — not only the decisions where the trader happened to mention a reference price. Selecting only decisions that name an old price biases the sample toward finding anchoring; the absence of an explicitly mentioned anchor is not itself evidence that no anchoring occurred, since an anchor can shape a decision without being named.

For each decision event — a target set or revised, an exit, an entry, or a re-entry — record:

Field	What it captures
Decision event	The specific choice under review (target set, exit taken, re-entry declined, and so on)
Reference price and its origin	The number involved and where it came from — entry price, a prior high or low, a round number, a first-seen quote
Predefined rule or justification	What the trading plan says about this type of decision, if anything, ideally recorded before the fact
Evidence available at decision time	What the setup, risk conditions, and market information actually supported at that moment
Actual decision and stated rationale	What the trader did and the reason recorded at the time, not reconstructed afterward
Review classification	One of the categories below

Classify each decision using categories that allow for ambiguity instead of forcing a yes/no call:

Justified reference use — the reference price maps to a role the trading plan actually defines (a technical level, a risk calculation, a documented invalidation), and the decision followed it.
Suspected unsupported reference influence — a reference price appears to have driven the decision with no rule assigning it that role, and no other evidence justifies its use.
Mixed reference use — both justified and unsupported reference-price influence appear in the same decision.
Ambiguous reasoning — the record does not make clear whether the influence on the decision was justified or unsupported.
No identified reference influence — the decision does not appear to involve a reference price at all. This is a legitimate outcome and should not be read as evidence that anchoring is absent from the trader’s broader pattern, only from this decision.
Insufficient information — the record does not contain enough detail to classify the decision into any of the categories above. This is a legitimate outcome, not a failure of the framework, and a reason to record more detail at the decision point going forward.

If a numerical rate is useful for tracking change over a sample, define its population, numerator, and denominator explicitly rather than treating any mention of an old price as disqualifying. Every decision event must land in exactly one of the six classification categories above — the categories are mutually exclusive, so no event is double-counted across them.

The full review population is every eligible decision event in the sample, as defined earlier — not only decisions that mention a reference price. Report that full population broken out by category before computing any rate:

Total eligible decision events
Classifiable reference-involving events (justified, suspected unsupported, or mixed)
No-identified-reference events
Ambiguous events
Insufficient-information events

No-identified-reference, ambiguous, and insufficient-information events are excluded from the rate below because none of them resolve to a comparable justified-versus-unsupported judgment. Excluding them narrows what the rate can speak to — report the exclusion counts alongside the rate rather than dropping them silently.

suspected unsupported reference-use share =
  (suspected unsupported reference influence + mixed reference use)
  / (justified reference use
     + suspected unsupported reference influence
     + mixed reference use)

Mixed reference use appears in both the numerator and the denominator because each such decision is classifiable and contains some suspected unsupported influence, even though it also contains justified use. If the denominator is zero — no classifiable, reference-involving decisions in the sample — report the share as undefined, not as zero.

This share measures the proportion of classifiable, reference-involving decisions that contain suspected unsupported influence. It does not measure how often anchoring occurs across all trading decisions in the sample, since decisions with no identified reference, ambiguous decisions, and insufficient-information decisions are excluded from it by construction — that is why the full population must still be reported separately. The share is a descriptive review indicator for a trader’s own record over time, not a validated psychological instrument and not a causal measure, and its value depends on how completely the record was kept. A high share does not by itself establish that anchoring caused any specific outcome; some flagged decisions may still have been reasonable for unrelated reasons. There is no universal threshold above which a share proves a problem — meaningful comparison requires comparable decisions and a consistent classification standard applied by the same reviewer across a meaningful sample, not one trade read in isolation.

Common anchoring failures

Treating a personal cost basis as a market signal

An entry price reflects what one trader paid. Cost basis alone does not establish what a fair price is now, though it remains relevant to accounting, tax, and risk decisions. A target or exit tied only to getting back to entry, rather than to the strategy’s own criteria, is a cost-basis anchor wearing a technical-analysis costume.

Confusing “off its high” with “cheap”

Distance from a 52-week or all-time high says nothing about current earnings, catalysts, or structure on its own. A stock can be justifiably far from an old high and still expensive, or near one and still cheap, depending on what changed.

Refusing to re-enter above a prior exit

Having sold a position once at a given price does not make that price a ceiling on future value. Treating it as one can mean missing a setup that would otherwise qualify cleanly under the trader’s own method.

Letting a round number’s salience substitute for a defined level

Round figures like $50 or $100 are not automatically meaningless. A study using Taiwan Stock Exchange limit-order data from September 2005 through May 2006 found that orders — especially from individual investors — cluster at round and even prices more than chance would predict, and identified price-barrier effects in that market and sample.⁴ That finding describes the market and period the authors studied; it should not be generalized into a universal claim about every market, instrument, price level, or present-day order book. Within that qualification, it means a round level can coincide with genuine order concentration and support/resistance behavior in markets with similar structure. It does not mean a round number is a valid level simply because it is round: the trader still needs an independent reason — order-flow evidence, a repeated reaction at that level, a rule in their own method — to assign it a role. A target or stop set at a round number only because it is memorable borrows the number’s psychological weight without the evidence to justify it, and clustering at a level does not by itself make that level reliable or tradable for a specific setup.

Where Costante fits

Costante supports self-defined guardrails set before a session starts — such as a daily loss limit, a per-trade risk limit, and a reentry limit for a given instrument — along with trade logging and after-trade review through structured tags covering setup type, execution discipline, mistakes, and outcome quality. Traders can use that recorded trade history alongside their own documented plans to examine whether a remembered price, rather than the plan’s own criteria, appears to have shaped a decision.

Costante does not identify which number a trader is anchored to, calculate a suspected unsupported reference-use share automatically, validate whether a technical level is meaningful, or determine the correct target, stop, or entry. The trader defines the plan’s actual criteria, decides which reference prices deserve a role in it, and remains responsible for every decision.

Frequently asked questions

Is anchoring bias the same as loss aversion?

No. Loss aversion is about the asymmetric weight given to a loss relative to an equivalent gain, evaluated against a reference point — the loss can be anticipated rather than an existing open position, so a trader does not need to already be underwater to exhibit it. Anchoring is about a fixed reference number continuing to influence a judgment regardless of whether the current decision involves a gain or a loss, or whether a loss is a live possibility at all. The two are not mutually exclusive and can occur together on the same trade — an entry price can anchor a target while also sitting at the reference point a trader is loss-averse around. A trader can also anchor with no loss anywhere in the picture, such as refusing to buy back above a prior sale price on a profitable round trip.

Why do traders keep referencing their entry price?

An entry price is genuinely useful for calculating risk and position size. The distortion is not using it for that purpose — it is letting it also function as an implicit target, support level, or “fair value” judgment the trading plan never assigned it.

Is a 52-week high a legitimate level to watch?

It can be, if a trader’s method explicitly treats it as a technical level with a defined role. Research on a historical sample of U.S. equities found nearness to the 52-week high associated with momentum-strategy returns in that sample, which the authors read as evidence of investors anchoring on it — a separate claim from whether it is the correct level for a specific trade, market, or strategy today.

Can anchoring happen on a stock a trader has never owned?

Yes. The first price a trader happens to see — a headline number, an analyst target, the quote at the moment they started watching a chart — can function as an anchor even with no position and no prior trade in that name.

How can a trader reduce anchoring-driven decisions?

Name the reference number explicitly and check whether the trading plan assigns it a defined role. A judgment changing when that number changes is not, by itself, evidence of bias — a technical level, an accounting or tax input, or a risk constraint can also legitimately change the judgment when its value changes. Reviewing a sample of decisions for recurring old numbers, rather than judging one trade in isolation, is what turns an occasional lapse into an identifiable pattern worth fixing.

Sources

Costante provides educational workflow tools, not financial advice. Trading involves risk.

Tversky, A., & Kahneman, D. (1974). Judgment under Uncertainty: Heuristics and Biases. Science, 185(4157), 1124–1131. ↩
Epley, N., & Gilovich, T. (2006). The Anchoring-and-Adjustment Heuristic: Why the Adjustments Are Insufficient. Psychological Science, 17(4), 311–318. ↩
George, T. J., & Hwang, C.-Y. (2004). The 52-Week High and Momentum Investing. The Journal of Finance, 59(5), 2145–2176. ↩
Chiao, C., & Wang, Z.-M. (2009). Price Clustering: Evidence Using Comprehensive Limit-Order Data. The Financial Review, 44(1), 1–29. ↩