Confirmation Bias in Trading: Build a Process That Can Prove You Wrong

Confirmation Bias in Trading: Build a Process That Can Prove You Wrong

বাংলায় পড়ুন

Learning objective

By the end of this guide, you will be able to recognise confirmation bias at the research, entry, trade-management and review stages, write a falsifiable thesis and use a disconfirmation checklist that gives opposing evidence the same standard of attention as supporting evidence.

Confirmation bias is the tendency to seek, notice or interpret information in ways that support an existing belief while discounting information that challenges it. In trading, it can turn a tentative idea into a defended identity: once a trader becomes “bullish gold” or “bearish Bitcoin”, every new fact is forced into that story.

The problem is not having a view. Trading requires decisions under uncertainty. The problem begins when the evidence required to enter is weaker than the evidence required to admit the view is wrong.

How confirmation bias appears in a trading process

During research

The trader searches for bullish commentary after deciding to buy, follows accounts that share the same view and stops reading once enough supporting points have been collected. Negative evidence is labelled uninformed, temporary or already priced in without a consistent test.

At entry

A weak signal is accepted because it supports the thesis. The same signal would have been rejected in another instrument. A brief move above resistance becomes a “confirmed breakout”, while poor breadth or weak volume is ignored.

During the trade

Good news is treated as fundamental confirmation. Bad news becomes “noise”. The stop is widened because the long-term story remains attractive, even though the original short-term setup has failed.

During review

A winning trade is credited to analysis even when the plan was broken. A losing trade is blamed on manipulation or bad luck without checking whether contrary evidence was visible before entry.

These patterns matter because more information does not automatically correct bias. A classic 1979 study by Charles Lord, Lee Ross and Mark Lepper found that people with opposing prior views assessed mixed evidence in ways that strengthened their original positions. The subject was not finance, but the mechanism is directly relevant: evidence can increase confidence when it is processed asymmetrically.

Step 1: write the thesis before searching for more evidence

Capture the idea in a fixed template:

• Observation: what happened in price, fundamentals or macro data?
• Interpretation: why might it matter?
• Forecast: what observable outcome is expected, by when?
• Invalidation: what evidence would make the thesis wrong?
• Action: what entry, size and exit rules follow?

For example:

Observation: the index closed above a three-month range. Interpretation: demand may be broadening. Forecast: price should hold above the range for three daily closes. Invalidation: a daily close back inside the range with deteriorating breadth. Action: enter only after the confirmation rule, with pre-defined risk.

Writing before further research prevents the thesis from changing invisibly as new facts arrive.

Step 2: define a falsifier, not a flexible concern

An invalidation must be capable of proving the thesis wrong. “I will reconsider if conditions change” is too flexible. It allows every adverse development to be reinterpreted.

A useful falsifier includes:

• a measurable condition;
• a deadline or review point;
• the action triggered if it occurs.

Examples include a close back below the breakout level, guidance below a stated threshold, a catalyst failing to occur by a date, or a relationship between two assets breaking beyond the range used in the strategy test.

Price invalidation and thesis invalidation can differ. A stop may limit loss before the full economic thesis is disproved. Conversely, a company may cut guidance while the share price initially remains high. The process should say which one controls the trade.

Step 3: keep a two-column evidence ledger

Before entry, create two columns:

Supports the thesis	Challenges the thesis
Fact, source, timestamp	Fact, source, timestamp
Relevance to forecast	Relevance to forecast
Confidence and limitation	Confidence and limitation

Do not force equal numbers of items. Evidence should be weighted by quality and relevance, not counted like votes. One audited profit warning can matter more than ten optimistic social-media posts.

Apply the same questions to both columns:

• Is the source primary or repeating another report?
• Is the information current?
• Does it address the forecast or merely sound related?
• What would make the evidence unreliable?
• Was the information already known when price moved?

Symmetric standards are the defence. Seeking one bearish article after reading twenty bullish ones is not balanced research if the bearish source is dismissed automatically.

Step 4: check source independence

Ten headlines may trace back to one anonymous claim. Treating them as ten confirmations creates false confidence.

Map important evidence to its origin:

• company filing or official release;
• exchange or regulator;
• central bank or statistical agency;
• named first-hand reporting;
• analysis based on already public data;
• unattributed social-media repetition.

Count independent sources, not links. Also separate facts from opinions. “The company cut full-year guidance by 8%” is a fact if taken from the release. “The cut is temporary” is an interpretation requiring additional evidence.

Step 5: build the strongest alternative explanation

Ask:

“If my preferred thesis is wrong, what is the strongest explanation of the same evidence?”

A breakout might represent new demand, short covering or a thin auction. Higher gold might reflect falling real yields, currency weakness or temporary hedging. A stock rally after earnings might reflect expectations that were even worse, despite weak absolute results.

The alternative should be plausible enough to change the decision. A deliberately foolish counterargument only protects the original view.

Then write the observation that would distinguish the two explanations. For a breakout, that might be follow-through, breadth and a close outside the range. For earnings, it might be the direction of guidance and margin rather than the headline revenue beat.

Step 6: use a red-team question before entry

The red-team role is to challenge the plan, not to predict the opposite direction. If working alone, write the answer yourself. If working with others, ask someone who has not helped build the thesis.

Useful questions include:

• What fact would make this trade obviously unattractive?
• Which source in the thesis is least independent?
• What am I assuming about liquidity, volatility or execution?
• Would I accept the same evidence for a trade in the opposite direction?
• Am I using a new reason to protect an old position?
• If I had no position, would I enter now at this price?

Record the answer before the order. A challenge performed only after a loss becomes hindsight.

Worked example: a hypothetical earnings breakout

Assume a trader is studying a fictional UK-listed company trading at £48 before results. The plan is to buy only after a daily close above £50.

The written bullish thesis requires:

• revenue growth of at least 10%;
• gross margin of at least 30%;
• full-year guidance maintained or raised;
• a close above £50 after the release.

The thesis is invalidated if guidance is cut or the post-release session closes below £49. The trader's evidence ledger looks like this:

Supporting evidence	Challenging evidence
Sector demand has improved	A major peer cut guidance last week
Revenue consensus has risen	Input costs increased during the quarter
Price is testing £50	Options imply a large move, so expectations are already high

Results arrive. Revenue grows 12%, but gross margin falls to 27% and management cuts guidance. The share spikes to £51 early, then closes at £47.80.

Confirmation bias would focus on the 12% revenue growth and the brief £50 breakout. The pre-written process reaches a different conclusion:

• revenue supported the thesis;
• margin failed the required threshold;
• guidance triggered explicit invalidation;
• the closing-price condition failed.

The planned long entry is rejected. This does not prove the share must keep falling. It means the specified setup did not qualify.

Now change the outcome: suppose margin is 32%, guidance is raised and price closes at £50.60. The setup qualifies, but normal sizing and execution checks still apply. A balanced process can approve a trade; its purpose is not permanent scepticism.

Step 7: separate thesis changes from position changes

Create a rule for new information:

1. Record the fact and timestamp.
2. State whether it supports, challenges or does not affect the thesis.
3. Update the probability or scenario only if the evidence was not already included.
4. Apply the pre-defined position rule.

Do not reverse this order by deciding to hold and then inventing a new thesis. If the timeframe changes from a two-day breakout to a five-year investment after the trade falls, that is a new decision and should be evaluated from zero.

Step 8: score the process after the outcome

Review the trade independently of profit:

• Was the thesis written before entry?
• Was disconfirming evidence collected?
• Were supporting and opposing sources judged by the same standard?
• Did the trade obey the invalidation?
• Was the final outcome caused by a known risk or a new one?
• Would the same process be acceptable if the trade had lost?

A good process can lose because markets are uncertain. A biased process can win through luck. Judge repeated decisions, not one outcome.

Common mistakes

Treating every contrary opinion as equally valuable

Balance does not mean giving weak claims the same weight as verified primary data. Use symmetric quality standards.

Becoming automatically contrarian

Rejecting popular evidence simply because it is popular is another shortcut. The goal is accurate assessment, not permanent opposition.

Moving the invalidation

If invalidation changes after price moves, save the original and write a new decision. Do not overwrite the audit trail.

Searching until the desired answer appears

More search results can amplify bias. Use a fixed research checklist and stopping rule.

Confusing confidence with evidence

A confident analyst, large following or prestigious employer does not replace the underlying data and assumptions.

Reviewing only losing trades

Winning trades can contain serious bias and rule-breaking. Review a consistent sample of wins and losses.

Practical checklist

• Write observation, interpretation, forecast and invalidation separately.
• Define a measurable falsifier and deadline.
• Keep supporting and challenging evidence in parallel columns.
• Trace repeated reports back to their original source.
• Apply identical quality tests to evidence on both sides.
• Write the strongest alternative explanation.
• Ask what observation would distinguish the two theses.
• Use a red-team question before entry.
• Record thesis changes before changing the position.
• Review process quality separately from profit or loss.

Sources

The evidence-processing mechanism is described in Lord, Ross and Lepper's 1979 study, Biased Assimilation and Attitude Polarization. Practical investing examples and guardrails are also discussed by Singapore's official financial-education programme, MoneySense, and the Nova Scotia Securities Commission.

This tutorial was prepared with AI-assisted research and writing under The Money Man editorial workflow. It is educational information, not personalised investment advice. The worked example is hypothetical, and no research checklist can guarantee a profitable outcome.