Confirmation Bias in Technical Analysis: Why Seeking the Other Side Does Not Work

By Sofia Harchich | Trading Psychologist & Behavioral Finance Writer | thewealthmirror.com

Confirmation bias in technical analysis is unusual among the biases, in that the standard remedy for it is not merely ineffective. Under a live position it can make the problem worse.

The remedy is familiar enough to recite. Seek disconfirming evidence. Argue the other side. Find the strongest bear case for your long. It appears in every serious book on the subject, it is taught in every risk course, and it is genuinely good advice for a scientist evaluating a hypothesis on a Wednesday afternoon with nothing at stake.

A trader holding size in a position that has started moving against them is not in that situation, and the difference is not one of degree.

What the standard advice assumes

Three assumptions sit underneath the standard remedy for confirmation bias in technical analysis, and specifically underneath look for evidence against your thesis, and all three are reasonable in the setting the advice came from.

It assumes you can inspect your own filtering. It assumes the search for counter-evidence is conducted by the same person, in the same state, as the search for supporting evidence. And it assumes that having considered the opposing case, you will weight it accurately.

Confirmation bias is standardly defined as the tendency to notice, seek and give greater credence to evidence aligned with what we already believe. Notice what that definition implies about the remedy: the instrument you would use to audit the filtering is the filter.

Why it fails, three ways

It is performed by a mobilised nervous system. The moment the position turns, physiological state changes, and what degrades first is exactly the capacity the exercise requires — holding two competing accounts in mind and weighing them without collapsing into one. The counter-argument gets generated. It gets generated by someone currently unable to hold it steady.

It is aimed at the wrong step. Installing a change between trigger and behaviour has an order: detect, regulate, decide, execute, enforce. Consider the other side is a decision-stage instruction. When the failure is at detection — you have not yet noticed that the analysis went sideways — a decision-stage instruction has nothing to attach to, because from the inside there is no decision in progress. There is only continued correct analysis.

It converts into validation. This is the failure that makes the standard treatment of confirmation bias in technical analysis actively harmful. A trader who deliberately seeks the bear case, finds it, and rejects it now holds a position that has survived scrutiny. The rejection was performed by the filter, but the feeling it produces is the feeling of having been rigorous. Considered-and-dismissed is a stronger psychological position than never-considered, and the size on the next entry frequently reflects it.

A counter-argument you generated and rejected often becomes the last thing that made the position feel safe.

Confirmation bias in technical analysis is a timing problem

The reframe that makes this tractable: confirmation bias in technical analysis is not primarily a problem of what you believe. It is a problem of when you were willing to believe it.

Almost every trader eventually acknowledges the evidence that invalidated their thesis. The acknowledgment happens. It happens forty minutes late, or two sessions late, or after the stop, and by then it registers as hindsight rather than as bias — which is why the pattern is so hard to see in a review, and so easy to see in a timestamp.

This makes it measurable in a way that was I being objective never will be. The measurable question is how long the gap ran between the evidence appearing on the chart and you treating it as real.

Technical analysis is especially exposed to this because it supplies a continuous stream of interpretable material at every timeframe. There is always a lower timeframe where the thesis still holds, always an indicator that has not yet crossed. The market provides as much confirming evidence as a trader is willing to drop down a timeframe to find, and dropping a timeframe does not feel like bias. It feels like diligence.

Calling it a timing problem places it inside the instability loop: once a position is open, the position itself is the trigger, and the evidence that arrives afterwards is filtered by a state that already wants to be right.

What replaces it: invalidation written in advance

The replacement for confirmation bias in technical analysis is unglamorous and takes about ninety seconds per trade.

Before entry, while the state is still regulated and nothing is at stake, write the single observable event that would make the thesis wrong. Observable is the operative word — not if it starts looking weak, which is a judgement to be made later by a compromised judge, but if it closes below 2,340 on the hourly, which is an event that either happens or does not.

This works for one reason: it moves the decision out of the moment when your capacity to make it is lowest, and into the moment when it is highest. Nothing is being asked of the trader under pressure except recognition of a fact. That is a step-one task, and step one is the only one still fully operational.

Four conditions make the difference between an invalidation note and a decoration:

1. It names an event on a stated timeframe, not a feeling or a general condition
2. It is written before entry, not after the position has started moving
3. It has no escape clause — no unless volume is light, no unless the higher timeframe still supports it
4. It is timestamped, so the gap between the event and your acknowledgment of it becomes visible afterwards

The fourth is the one traders skip, and it is the one that turns a rule into data.

The measure that replaces “was I objective?”

Log one number per closed trade: the gap between the invalidation event occurring and you acting on it.

Not whether you eventually acted. The interval. Ten trades gives you a distribution, and the distribution says something was I objective never can — because it is a behavioural anchor rather than a self-assessment. How many minutes produces data. How honest was I with myself produces mood, and mood tracks the outcome of the trade rather than the quality of the process.

What the distribution typically shows is that the gap is patterned: longer on larger positions, longer after a loss, and longer on the instrument that carries the most personal weight — which is often gold rather than an index, because gold carries symbolic loading around scarcity and permanence that a broad index does not. A gap that holds at two minutes everywhere except one instrument is the most useful line in the record.

Where being wrong stops being informational

Underneath the timing problem sits a layer the standard advice never reaches.

Analytical rigour has almost nothing to do with the delay. What produces it is that acknowledging the invalidation means acknowledging having been wrong, and for most traders being wrong about a market has become entangled with something larger — competence, judgement, the reason for doing this at all. The pain of a loss carries roughly twice the psychological weight of an equivalent gain, and identity attaches to the loss rather than to the position.

That is what happens when a person invests real attention in a skill, and it is the same mechanism that produced the discipline to learn technical analysis properly in the first place. The task is to separate two claims — this thesis was wrong and my judgement is unreliable — which are treated as one claim by almost everyone and are not the same claim at all.

The second remedy that fails: journal it and review at the weekend

The other standard prescription deserves the same audit, because it fails for a related reason and is even more widely followed.

Keep a journal, the advice goes, and review it periodically to spot your patterns. The trouble is that a journal entry written after the position closes is written by the same filter that ran the position, working with material that has already been reorganised by knowing the outcome. What gets recorded is what a person who already knows how it ended remembers believing at 10:40, which is a different document from what you actually believed at 10:40.

Weekend review compounds it in three ways.

• The trades that get reviewed are the memorable ones, which are the large and the recent — the same selection bias the journal was meant to detect
• Reading entries produces a narrative, and narratives resolve. A pattern that would be obvious in a column of numbers becomes, in prose, a series of individually reasonable decisions
• By the weekend, the physiological state that produced the behaviour is unavailable to inspection, so the record documents conclusions rather than conditions

None of this makes journalling worthless. It makes the standard form of it a step-six intervention aimed at a step-one failure, which is the same structural error as consider the other side.

The adjustment is small. Timestamp the invalidation event and the acknowledgment while the position is open — two entries, four words each, no narrative. Then review by the number rather than by reading: sort ten trades by the gap and look only at the longest three. Narrative comes afterwards, applied to a pattern the numbers already found, rather than being the instrument of discovery.

Where the gap runs longest

Ten trades of timestamped gaps will usually show the same three conditions, and knowing them in advance is worth more than any resolution to be objective.

On the largest positions. The gap scales with size, which is the opposite of what a trader would design. The position that most needs a fast acknowledgment gets the slowest one, because the cost of being wrong is highest exactly where the willingness to be wrong is lowest.

After a previous loss. The second position of a losing session carries a longer gap than the first, often by a factor of two or three. Something other than the chart is being defended by then.

On the instrument that carries personal weight. Discipline can hold across an entire book and collapse on one symbol. Gold does this more often than an index, because it carries loading around scarcity and permanence that a broad basket of equities does not, and a thesis about gold is rarely only a thesis about gold.

A trader who knows these three conditions can do something no amount of general objectivity allows: pre-commit to a mechanical exit on trades that meet two of them, and leave discretion intact everywhere else. That is a narrower intervention than trying to be less biased, and unlike trying to be less biased, it can be checked afterwards.

What the advice was right about.

One part of the conventional treatment of confirmation bias in technical analysis survives the audit, and it is worth keeping.

Generating the opposing case is genuinely valuable — before entry, in writing, while nothing is at stake. At that point it supplies the material from which the invalidation condition gets written, rather than serving as an audit run by the filter itself, and it is why the ninety seconds before the trade is worth more than the forty minutes after it.

For the wider set of biases this sits inside, start with behavioural finance for traders, and with recency bias trading for the bias most likely to be sizing the position you are about to defend. If you do not yet know which emotional pattern your behaviour under pressure fits, the pattern quiz is the shorter route in.

---

About the Author

Sofia Harchich is a Trading Psychologist and Behavioural Finance Writer with a Master’s in Psychology. She works at the intersection of Jungian shadow work, neuroscience, and market behaviour — helping traders understand the psychology driving their decisions, not just the strategy.

Read more at thewealthmirror.com/about