The strategy that lies to you: confirmation bias in live trading and in backtesting

A backtest can produce a beautiful equity curve and still give a trader more confidence than the evidence deserves.

The problem does not have to be fake data or an obvious calculation error.

Sometimes the distortion comes from something much less visible: which evidence the trader chooses to notice, which results receive the most attention, and which uncomfortable observations are explained away.

This is where confirmation bias becomes especially relevant.

Confirmation bias is the tendency to favour information that supports an existing belief while giving less weight to information that challenges it.

For traders, that can influence both sides of the process: the decisions made while a position is live and the supposedly objective testing used to decide whether the strategy deserves to be traded in the first place.

How Confirmation Bias Distorts Live Decisions

The bias can begin influencing a decision before the trader consciously realises anything has changed.

Instead of asking:

“What does the available evidence currently show?”

The question quietly becomes:

“What evidence supports the position I already want to be right?”

Once that shift happens, analysis can start becoming justification.

Defending an Existing Position

Confirmation bias can become particularly dangerous after a position has already been opened.

A trader who is long may pay greater attention to bullish information while minimising evidence that the original setup has weakened.

A trader who is short may do exactly the opposite.

This may show up as:

Ignoring a predefined invalidation condition.

Searching for new reasons to remain in a losing position.

Adding indicators until one supports the original view.

Changing the interpretation of market structure after entry.

Looking only for commentary that agrees with the position.

The issue is not simply having conviction.

A trader needs enough confidence to execute a tested strategy.

The problem begins when conviction prevents new information from being evaluated according to the rules that were established before the trade began.

Conviction and Rigidity Are Not the Same Thing

A trading plan should not require the trader to change their opinion every time price moves against them.

Normal market movement can occur without invalidating a setup.

The important distinction is between:

Remaining with a trade because the original conditions are still valid.

And:

Remaining with a trade because accepting contradictory evidence has become psychologically difficult.

A good strategy should define what information actually invalidates the trade.

That allows the trader to distinguish between ordinary adverse movement and evidence that the original thesis no longer meets the strategy’s rules.

The Slower Cost: Protecting Beliefs Instead of Testing Them

Confirmation bias can also affect learning over a much longer period.

A trader develops an idea about how a particular setup works.

They then begin noticing examples that support the idea more readily than examples that challenge it.

Over time, the strategy may appear stronger simply because contradictory evidence receives less attention.

This is particularly dangerous because the trader may genuinely believe they are researching the strategy carefully.

They are collecting evidence.

The problem is that the evidence is being filtered.

A useful research process therefore needs to ask not only:

“When does this strategy work?”

But also:

“When does it fail?”

“Under what conditions does performance deteriorate?”

“What evidence would convince me that my original idea is wrong?”

Where Confirmation Bias Can Enter Backtesting

Backtesting can feel more objective because the trader is working with historical data rather than a live position.

But the person designing, interpreting and modifying the test can still introduce bias.

Confirmation bias can enter the process at several stages.

1. Focusing on Favourable Periods

A trader may give disproportionate attention to periods where the strategy performed particularly well.

A strong run can feel like proof that the concept works.

But good performance during one period may reflect conditions that were unusually favourable to that particular strategy.

For example, a trend-following method may look exceptional during a persistent directional market and considerably weaker during prolonged ranging conditions.

The correct question is not simply:

“Did the strategy make money here?”

It is:

“How did it behave across different conditions, including the ones least favourable to it?”

2. Rationalising Losing Periods

Drawdowns deserve investigation.

But a trader who strongly believes in a strategy may begin explaining each difficult period away.

For example:

“That loss was because of unusual news.”

“That month doesn’t really count.”

“The market was behaving strangely.”

“I wouldn’t have taken that trade live.”

Some exclusions may be legitimate if they were defined by the strategy beforehand.

The danger appears when exclusions are created only after the outcome is known.

If losing trades are repeatedly removed after the fact while profitable trades remain untouched, the backtest can gradually become a description of the strategy the trader wishes existed rather than the rules that could actually have been executed.

3. Cherry-Picking the Historical Window

The choice of testing period can also affect the result.

If a trader selects a historical window because they already know the strategy performed well during that period, the test provides less useful evidence than one designed without reference to the desired outcome.

A strategy should ideally be examined across multiple environments rather than one convenient stretch of history.

That can include periods characterised by:

Trending markets

Ranging markets

Higher volatility

Lower volatility

Different economic environments

The objective is not to demand that every strategy performs equally well under every condition.

It is to understand where the strategy’s strengths and weaknesses actually appear.

4. Changing the Rules After Seeing the Results

Another problem appears when strategy parameters are repeatedly adjusted after historical results have already been seen.

A trader notices that a 20-period setting performed poorly, but a 17-period setting would have avoided several losses.

Then another parameter is adjusted.

Then another filter is added.

Eventually the historical equity curve becomes extremely attractive.

But each improvement was discovered with knowledge of the same historical data.

This can create an overfitted strategy: one that explains the past extremely well but may generalise poorly to unseen data.

This problem is broader than confirmation bias alone, but confirmation bias can encourage it when every modification is judged mainly by whether it improves the desired result.

Define the Test Before Looking at the Answer

One of the strongest protections against this problem is to define important parts of the test in advance.

Before examining the results, specify:

The exact strategy rules

The instruments being tested

The historical period

The minimum sample size

Transaction-cost assumptions

The performance measures that matter

The conditions that would cause the strategy to be rejected or reviewed

This reduces the opportunity to keep changing the question until the historical data produces the answer you wanted.

Separate Development Data From Validation Data

Where practical, another useful technique is to avoid using every piece of historical data to develop the strategy.

A trader can use one sample to develop or refine the rules and then examine the strategy on data that was not used during that development process.

This does not remove every source of bias.

But it provides a stronger test than repeatedly modifying a strategy against the same historical sample and then treating performance on that same sample as independent validation.

Review Winning and Losing Periods With Equal Curiosity

A good backtest review should investigate both sides of the result.

For profitable periods, ask:

What conditions helped the strategy?

Were profits concentrated in a small number of trades?

Was the result unusually dependent on one market regime?

For losing periods, ask:

Were the losses consistent with expected strategy behaviour?

Did market conditions change?

Was the strategy repeatedly exposed to a known weakness?

Would the rules genuinely have excluded these trades at the time?

The purpose is not to prove the strategy good or bad.

It is to understand what the evidence actually says.

Countering Confirmation Bias in Live Trading

The same discipline can be brought into real-time decisions.

Define the Invalidation Condition Before Entry

Before taking the trade, answer:

“What would prove this setup no longer meets my rules?”

Write that condition down if necessary.

Once the position is open, the trader has a reference point created before the psychological attachment to the outcome became stronger.

Actively Consider the Counterargument

Before adding risk, extending a target or changing the original trade management, ask:

“What is the strongest evidence against my current position?”

The goal is not to manufacture reasons to exit every trade.

It is to make sure contradictory evidence receives deliberate consideration rather than being filtered out automatically.

Use Predefined Information Sources

Continuously searching for more market opinions can sometimes increase confirmation bias rather than reduce it.

A trader who wants reassurance can usually find someone online who agrees with almost any market direction.

A stronger process is to decide in advance what information is relevant to the strategy.

For example:

Price structure

Specified timeframes

Scheduled economic events

Defined volatility measures

Other indicators explicitly included in the trading rules

This limits the temptation to keep searching until supportive evidence appears.

Use a Checklist

A short checklist can create a pause before a biased interpretation becomes an action.

For example:

What was my original reason for entering?

Is that reason still valid?

What evidence currently challenges the position?

Has a predefined invalidation condition appeared?

Am I changing the trade because the strategy requires it or because I want to avoid accepting a loss?

Would I make the same decision if I did not already have this position open?

Use the Journal to Compare Before and After

Confirmation bias becomes easier to investigate when the trader records their reasoning before the outcome is known.

For each trade, consider recording:

Why the trade was entered

What evidence supported it

What evidence argued against it

What would invalidate the setup

What changes were made after entry

Why those changes were made

After the trade closes, compare the original reasoning with the story being told afterwards.

This can help distinguish between:

“The evidence changed.”

And:

“My explanation changed because I already knew the outcome.”

The Objective Is Falsification, Not Self-Attack

Actively challenging a strategy does not mean trying to prove that everything you believe is wrong.

The goal is to create a process in which ideas are allowed to fail.

A useful strategy should survive reasonable attempts to challenge it.

If a trading idea only looks convincing when unfavourable periods are excluded, parameters are repeatedly adjusted after the fact and contradictory evidence is ignored, that weakness deserves attention before real capital is exposed.

The strongest evidence for a strategy is not that the trader can construct a persuasive argument in its favour.

It is that clearly defined rules continue to produce acceptable results when tested against data and conditions that were not selected simply because they made the strategy look good.

What This Comes Down To

Confirmation bias affects research as well as live trading. Traders can favour evidence that supports both an open position and a strategy they want to believe in.

Backtesting is not automatically objective. Test windows, exclusions, parameter changes and interpretation all involve human choices.

Define rules before seeing results. Predefined parameters and evaluation criteria reduce the freedom to rewrite the test after the answer is known.

Study losses as seriously as wins. Difficult periods may contain important information about strategy risk and market conditions.

Challenge the thesis deliberately. Ask what evidence would show that the position or strategy is wrong.

Do not search endlessly for agreement. More sources can create more opportunities to find confirmation. Define which information is genuinely relevant to the strategy.

Use unseen data where practical. Testing a developed strategy on data not used to build it provides a stronger challenge than repeatedly testing against the same sample.

Keep a before-and-after record. Journaling the original reasoning helps prevent hindsight from rewriting what you believed before the outcome was known.

Conclusion

Confirmation bias is difficult precisely because it rarely feels like bias while it is happening.

In live trading, it can make a trader defend a position long after their predefined conditions have changed.

In backtesting, it can make a weak strategy appear stronger through selective periods, after-the-fact explanations or repeated adjustments to the same historical data.

The solution is not simply to promise yourself that you will be more objective.

It is to build objectivity into the process.

Define the rules before the result.

Record the evidence for and against the idea.

Test difficult periods as well as favourable ones.

Allow a strategy to fail its test.

And when a position is live, keep asking whether the current evidence still supports the original rules — not whether you can find another reason to remain convinced.

A useful backtest should challenge what you hope to find, not simply confirm it.