Confirmation bias on charts and how to disagree with yourself
A chart doesn't tell a story until you decide what story you want — then it tells that story, perfectly, in your favour, with surprising consistency. The candlesticks are the same. The moving averages are the same. The volume bars are the same. The narrative laid over them changes entirely depending on whether you're looking for a reason to buy or a reason to skip.
This isn't a flaw in your discipline. It's a specific and well-documented cognitive bias — confirmation bias — and trading is one of the cleanest places to watch it work in real time, because the test is fast: you take the trade, and within a few hours you find out whether your read was right or whether you saw what you wanted to see. It's also one of the cleanest sub-causes of the broader pattern of how retail loses money — bias-confirmed entries tend to be late entries.
The interactive widget below shows the effect on a single fictional chart. Same prices. One toggle. Watch what happens to the story.
The original finding
The classical reference is Peter Wason's 1960 selection-task experiment, published in the Quarterly Journal of Experimental Psychology. Wason gave participants a simple rule — "the next number is larger than the previous one" — and a starting triplet (2, 4, 6). Their job was to figure out the rule by testing other triplets. He'd tell them whether each new triplet fit the rule or not.
Almost every participant tested triplets that confirmed the rule they suspected (8, 10, 12 — yes; 100, 200, 300 — yes; 1, 2, 3 — yes). Almost no one tested triplets that could disconfirm it (10, 5, 1 — no; 6, 4, 2 — no). When they were asked to commit to a rule, most picked something narrow ("ascending even numbers" or "each number is double the last") because all their evidence pointed that way. The actual rule, "any three ascending numbers", was never the favourite, even though it would have been suggested by literally any successful test that broke their initial hypothesis.
The study has been replicated dozens of times, in dozens of variations. It's one of the most stable findings in cognitive psychology. The brain doesn't naturally test the hypothesis it holds — it confirms it, gathers evidence in favour, and stops. Nickerson's 1998 review in the Review of General Psychology is the standard summary; if you want one paper that covers the whole literature, that's the one.
For trading, the implication is simple: if the brain naturally seeks confirmation rather than disconfirmation, the moment you decide a chart is bullish, you'll find five reasons it's bullish before you find one reason it isn't. The bias does exactly the work the original Wason experiment showed it doing — fast, automatic, and invisible to the person doing it.
How it shows up at the chart specifically
Three patterns are common. Once you know to look for them, they're hard to un-see in your own behaviour.
Pattern 1: indicator shopping. You're considering a long entry. You look at the RSI — it's neutral. You look at the MACD — it's borderline. You look at the moving averages — they're trending sideways. You keep flipping through indicators until you find one that says "bullish", then you commit. The signal that pushed you over the edge is the one that confirmed what you already wanted to do; the indicators that disagreed get treated as noise. Importantly, you'd have done the exact same scan with the indicators that actually confirmed your existing read, regardless of which indicators those were. The order of arrival is the giveaway.
Pattern 2: news-after-the-thesis. You've decided to short. You scroll through Twitter or your news source of choice. The bullish takes look like marketing copy, written by people who don't understand the chart. The bearish takes look like clear-eyed analysis. Same words written by other people are now being read with completely different filters. Lord, Ross and Lepper's classic 1979 study at Stanford found this effect in a non-trading context — they showed pro and anti capital-punishment partisans the same set of mixed-evidence studies, and both groups came away more confident in their original positions. The mechanism is the same; the speed is faster.
Pattern 3: candlestick pareidolia. This is the trader-specific version of seeing faces in clouds. You scroll back over the past few hours of price action looking for the pattern that supports your thesis. There are always candles you can string together to make almost any pattern. Hammer at the bottom of the down-leg? It's a reversal. Long upper wick on the rally? It's distribution. Doji at a key level? Indecision. Doji after a doji? Now it's a trend. Each of these patterns has some validity in some contexts, but the pattern your brain is reaching for is almost always the one that fits the thesis you walked in with. The named patterns have precise definitions on paper — the engulfing pattern checker walks through one of them — and most "patterns" people see in the wild fail those definitions when actually checked.
The common thread is that the chart isn't doing anything new — your eyes are doing different work depending on what conclusion they're being asked to defend.
See the bias in one chart
Same chart · pick a story
Both narratives are coherent. Both pick out features that are genuinely there in the chart. Neither is "wrong" — and that's the trap. The chart supports both stories about equally well, and which one you walk away with depends almost entirely on which one you walked in with. The bias isn't that you saw something that wasn't there. The bias is that you didn't bother to check whether the opposite story was equally well-supported.
The disconfirming-view technique
The fix is borrowed from intelligence analysis, where it's been a standard practice for forty years. The CIA's Psychology of Intelligence Analysis (Heuer, 1999) — a free public document — devotes a full chapter to "Analysis of Competing Hypotheses". The protocol is simple enough to use at the chart in real time:
- Write the bull case in one sentence. Specifically. Not "looks bullish" — what about it is bullish. "Higher lows holding through three retests, breakout above prior local high."
- Write the bear case in one sentence. Specifically. "Each rally smaller than the last, range narrowing, recent retest of the level rejected."
- Pick the trade only if one case is meaningfully stronger than the other. Both equally plausible? No trade. The bias has done its work; the chart is genuinely ambiguous and you don't have an edge.
- While the trade is open, the only question is whether the disconfirming case has gotten stronger. Not weaker. Stronger. If yes, exit. If no, hold to your stop or target.
This is mechanical, not intuitive, and that's the point. The brain will try to interpret incoming information as confirming the trade you already took. The protocol forces you to check the opposite, in writing, in advance, where the bias has the least leverage.
Phil Tetlock's research on forecasting accuracy (compiled in Superforecasting, Crown 2015, with Dan Gardner) found that the single biggest discriminator between accurate and inaccurate forecasters wasn't IQ, training, or domain expertise. It was the willingness to actively seek out reasons the forecast might be wrong. The accurate forecasters were systematically less attached to their initial reads. They would update faster, in both directions, when new information arrived — because they'd already done the work of imagining what disconfirming information would look like.
This generalises to charts almost word-for-word. The traders who hold up over hundreds of trades aren't the ones with the cleverest reads. They're the ones who notice fastest when the read isn't holding up.
What the chart can't tell you
There's a deeper version of this problem worth naming. The chart can show you what price has done. The chart cannot show you what price will do, and it especially cannot show you which scenario — bull or bear — is more likely. Both stories that the widget above offers are post-hoc rationalisations: they explain the past but predict nothing about the future.
This is a hard thing to internalise because the visual language of charts looks predictive. A trendline ending at the right edge of the screen feels like an arrow. The brain extrapolates automatically. It takes deliberate effort to remember that the line is just where price has been, and that the future portion of that chart will be drawn by aggregate behaviour you cannot see — order flow, liquidity, news, macro positioning, central-bank statements, the actions of people with information you don't have.
The honest position to hold while looking at any chart is: this is what's happened so far. I have a thesis about what's likely next. The chart does not validate the thesis any more strongly than its opposite. The validation will come from price action over the next N hours/days, against my pre-defined exit levels.
Charts are diagnostic, not predictive. The bias makes them feel predictive. The bias is wrong. The same mechanism powers noticed-the-chart-first FOMO — once your eye locks onto a setup, the brain spends the next sixty seconds rationalising why the entry is "right now" rather than asking whether it's an entry at all.
Why "look at it on a higher timeframe" doesn't fix it
A common piece of advice for confirmation-biased traders is to "zoom out and look at the bigger picture." This is intuitive but doesn't actually fix the bias — it just changes the chart you're biased about. The same neurology that found bullish features on the 5-minute chart will find bullish features on the 4-hour chart, then on the daily chart, then on the weekly chart. Higher timeframes have larger and more obvious patterns, which is sometimes useful for context, but the underlying bias process is identical at every scale.
The actual fix is the disconfirming-view exercise applied at whatever timeframe you're trading on. Multiple timeframes are useful as additional inputs, not as bias insurance. A trader who runs the bull-case-vs-bear-case discipline at the 5-minute timeframe will outperform a trader who looks at five timeframes but only writes down the case that supports the trade.
For traders running our signal feed, the framework is built into how the entries are filtered: the system has to find evidence for both directions before it'll fire a signal in one. Setups that look strong from one angle but weak from another don't generate alerts. That's not a feature of the strategy as much as a structural application of the same disconfirming-view rule — applied at scale, in code, where the bias can't reach.
FAQ
If both cases are coherent, doesn't that mean charts are useless?
No — it means charts are useful for context but not for unambiguous direction calls. Two-thirds of the price action on most timeframes is genuinely indeterminate, and treating it as such is the right reading. The remaining third is where one case is meaningfully stronger than the other; those are the setups worth taking. The skill is recognising which third you're in, not generating a directional read on every chart you see.
Doesn't the disconfirming-view exercise just mean I'll always have a bear case for every long?
That's the point. The exercise isn't to decide that every long is wrong; it's to make sure you've actually evaluated the bear case before committing. Often the bull case wins clearly when you compare them honestly. When it doesn't, you skip the trade — and the cost of skipping a borderline setup is far lower than the cost of taking one that gets you stopped out.
Why does this hit traders harder than people in other domains?
Three things. First, fast feedback (you find out within hours, not months) creates strong reinforcement loops where the bias gets rewarded by short-term wins. Second, the financial stake makes the emotional charge much higher than in low-stakes domains, which amplifies the bias. Third, charts are visually compelling — they look like data, but they're really pattern Rorschachs that the bias paints over. Most other domains don't have such a clean visual surface for the bias to attach to.
Is "Analysis of Competing Hypotheses" overkill for retail trading?
The full ACH protocol is — it's designed for multi-week intelligence assessments. But the core mechanic (write each hypothesis explicitly, evaluate evidence against each separately, pick the one with the most surviving evidence) is the right shape for trading and takes about 90 seconds per trade. That's a small price for a meaningful reduction in bias, and far less than the time most traders spend hunting for confirming indicators.
Won't this slow down my trading too much?
It will slow you down. That is part of the value. Most retail traders trade too much, not too little — typical retail flow takes more setups than the underlying edge supports because the bias is producing more trade-it candidates than it should. Slowing down by even 30 seconds per setup, with a forced bear-case write-up, removes the worst trades disproportionately. Net P&L often improves while trade count drops.
Can a trading journal help with this?
Yes, but only if the journal records both the bull case and the bear case at entry, not just the trade. A journal of "I bought BTC at 75k because of higher lows" tells you nothing about how you'd evaluate the bias. A journal that records "bull case = X, bear case = Y, picked bull because it's stronger on point Z" gives you a feedback loop: when the trade fails, you can check whether the bear case's reasons were actually stronger than you credited at the time. Over many trades, that's the most direct way to retrain the bias-prone parts of your reading.
How does this connect to other biases like loss aversion or anchoring?
Confirmation bias is upstream of most of them. Once a trader takes a position, confirmation bias keeps them committed (filters incoming information through the position they hold), which makes loss aversion (don't close at a loss) and anchoring (the entry price stays meaningful) much more powerful. Closing the confirmation-bias door at entry — by writing the bear case and being ready to act on it — limits how far the downstream biases can run when the trade goes against you. The five mental traps post covers the others; this one is the upstream one.
If you read this and noticed yourself nodding along thinking "yes, other people do this" — that's confirmation bias in action one layer up. Try the exercise on your last three trades. Write the bull case and the bear case for each, in one sentence each, as honestly as you can recall. Most traders find at least one of three was a close call where they only wrote down the case that supported their action. That trade is the one to study.
Want a system that runs the disconfirm check for you?
You're in, dashboard's ready
Our trade alerts only fire when both directions have been evaluated and one case is meaningfully stronger. The disconfirming-view exercise, applied in code, on every signal. Join the waitlist and you'll know the day the feed goes live.
You're signed in. Open your dashboard for the live track record + your trade log so you can audit your own bull-vs-bear reads over time.