The Overconfidence Effect: What the Evidence Shows About Feeling Sure.
Most of us feel surer than our track record justifies, and experts are no exception. Studies of doctors, investors and national security officials found the same gap, and in trials, short training and simple habits narrowed it a little. Here is what the science actually says, and what to do with it.
01Confidence and Accuracy
Check Confidence Against Outcomes
Why do you feel more sure than you should, and how big is the gap? Start with a quick test. Pick something you feel confident about: a date, a figure, how next quarter will go. Now put a number on how sure you are. Confidence can be checked against how often similar judgements prove correct.[8] Researchers call the match between the two calibration. If you are well calibrated, the things you feel seven-in-ten sure of happen about seven times in ten.
In one study, nearly nine in ten licensed American students rated themselves safer drivers than the median participant.[9] Only half of any group can sit above its middle. That is overconfidence bias: feeling surer of yourself, or of your judgement, than the facts allow.
That matters because experts are not spared. One study followed political and economic experts, such as professors, government advisers and think-tank analysts, for twenty years. Across political and economic forecasting questions, they often did poorly against simple benchmarks, with results varying by question, time horizon and thinking style.[4] The experts who appeared most often in the media, and who spoke with the most certainty, did worst of all.[5]
How big is the gap? It shows up in very different jobs. In medicine, studies compared diagnoses with what autopsies later found. In one study of patients who died in intensive care and underwent autopsy, diagnoses made with complete certainty were wrong about two times in five.[20] That figure does not describe medical judgement in general, and calibration is better where doctors get quick feedback.[20] In finance, surveys asked senior finance executives, most of them chief financial officers, for ranges they were eight-in-ten sure would contain the annual stock-index return. The ranges did so only about a third of the time.[41] In national security, a large recent study looked at officials in advanced military education programmes.[26] They judged probabilities about current facts and future events.[26] Statements they gave a nine-in-ten chance of being true were true less than six times in ten.[26]
These are not isolated cases, though the size of the link varies. A pooled analysis of more than thirty studies found a small association between overconfidence and financial decisions, strongest for investing, then trading and innovation.[24]
Decades of research have found the pattern many times over.[1][6] So the useful questions are why the mind produces it and what helps.
Why does it happen? This article makes a simple case. The overconfidence effect is less a flaw of character or education than a by-product of how the mind works. Your brain produces a feeling of certainty much as your stomach produces hunger. It does so automatically, without being asked, and from inner signals that may have little to do with whether you are right.[2][18]
Four kinds of evidence support this case. Decades of studies have compared people's confidence with how often they were right.[8] An influential review sorted overconfidence into three kinds.[6] A large brain-scan study, using a reasoning task, compared the brain activity linked to feeling sure with the activity linked to judging your own accuracy well.[15] And randomised trials found that different kinds of training can improve forecasts or reduce specific biases.[31][33] The chapters below take each in turn. Then they answer the questions most readers bring: who is affected, what it costs, and which habits help.
02The Mechanism
Three Types of Overconfidence and the Brain Signals Linked to Them
For decades, studies of overconfidence seemed to disagree. Some found that people greatly overrated their abilities. Others found that people underrated themselves on easy tasks. An influential review showed that researchers had been using one label for three different things.[6]
The first is overestimation: thinking you did better than you really did. The second is overplacement: thinking you rank higher than others when you do not, as the student drivers did.[9] The third is overprecision: being too sure that what you believe is right, which shows up as ranges that are far too narrow.[6] The review found that on easy tasks, people underrated how well they did but still ranked themselves above others. On hard tasks, the pattern reversed. Because the three kinds respond differently to how hard a task is, studies that lumped them together got mixed results.
Overprecision is the most stubborn of the three. In the studies reviewed, it appeared on easy and hard tasks, in different groups and in different fields.[6][21] A review of research on experts found them overprecise across levels of expertise.[21] Being an expert did not reduce excess certainty, and it may even increase it.[21]
What brain studies link to overconfidence. In one study of a reasoning task, feeling confident went with activity in the striatum, a reward-linked area. In another, about adverse life events, people updated risk estimates more after better-than-expected information, and highly optimistic participants showed weaker tracking of worse-than-expected errors in a right frontal region. These are correlations from separate studies, not a tested causal chain.
Diagram · HPC
Brain studies offer a clue to why. One large brain-scan study, of more than three hundred healthy adults doing a social and factual reasoning task, looked at two things. It compared the activity linked to feeling confident with the activity linked to judging your own accuracy well.[15] Feeling more confident went with more activity in the striatum and hippocampus, areas linked to reward and memory. Skill at telling when you were right or wrong is called metacognitive accuracy. More of this skill went with less activity in a region at the front of the brain.[15]
That matters because feeling sure and accurately judging your own performance went with different brain activity. These are correlations, so they cannot show two separate systems at work. But they suggest that feeling sure is not the same as judging your own accuracy well. A review of brain research on judging your own knowledge drew a similar line.[16] Areas at the sides of the front of the brain were linked to judging answers already given. Areas toward the middle were linked to the feeling of knowing beforehand.[16]
A second pattern makes things worse. In a task about adverse life events, people changed their risk estimates more after good news about their risk than after bad news.[17] This is called asymmetric belief updating. In the most optimistic people, a region at the right front of the brain tracked worse-than-expected errors more weakly (point 04 in the diagram above).[17] Separately, one review estimates that about four in five people show some optimism bias, expecting things to turn out better than is likely.[18] That figure is not the share of people who show this updating pattern.
Why would the mind work this way? One answer comes from a mathematical model, not from observation. In this game-theory model, individuals competed for resources, and overconfidence gave them the best results.[11] That held as long as the prize was worth enough relative to the cost of competing.[11] Under that condition it was also stable over generations, meaning more accurate rivals could not displace it.[11]
A related finding concerns people who are weakest at a task. In well-known experiments, people who scored in the bottom quarter thought, on average, that they had done better than about three in five others.[12] In fact they had beaten only about one in eight.[12] The researchers concluded that doing a task well and judging how well you did it often take the same skills.[12] Follow-up studies supported this account.[13] But another analysis, of judgements about intelligence, argued that much of the pattern could arise statistically.[14] It found no support for the idea that people of lower ability judge themselves less accurately.[14] The explanation remains debated.
So far, the findings concern how confidence is updated and judged. Confidence also tends to start too high. Classic research on mental shortcuts found that fast, automatic thinking builds confident, tidy stories from scraps of information.[1][2] Slower, deliberate thinking has no reliable inner alarm to warn it that a bias is at work.[1][2] People also believe they understand complex things far better than they do, a pattern called the illusion of explanatory depth.[10] In these studies, people rated their own understanding on a seven-point scale. Their ratings fell once they tried to explain how something works.[10]
An older study shows the same thing from another angle. Thirty-two judges, including clinical psychologists, read more and more information about one case. Their confidence rose from about a third to just over half, but their accuracy stayed below three in ten.[19] In that study, more information raised confidence without raising accuracy.
Who gets it right? In the calibration research, the professionals who came closest to good calibration worked where feedback is fast and clear. Experienced weather forecasters, for example, learn within hours whether their forecast was right.[8] Delayed or unclear feedback can make it harder to spot errors and learn from them.[37]
03Evidence
The Five Strongest Studies on the Overconfidence Effect
01The claim
The single load-bearing finding
The hero study finds 20 years of forecasts.
Not all evidence carries equal weight. A twenty-year study of experts tells you something different from one lab session with thirty students. Each study below was scored out of a hundred. The six criteria are design, sample, rigour (measurement quality), causality (how well it can show cause and effect), replication and citations. A high score means a sturdier study, not a bigger effect. Together, the five suggest that the gap between confidence and accuracy is not just a quirk of how researchers ask questions. Design counts most and sample size next; replication and citations count least.
Key study result
20 years of forecasts
02How we measured
Grading the calibration studies
Studies scored on design, sample, rigour, causality, replication, citations.
For overconfidence science, ecological validity is decisive: laboratory calibration gaps mean little unless they replicate in real expert populations making real consequential forecasts over years.
Rubric weights
03The spread
Heterogeneity across 5 studies
Methodological quality across the ranked studies.
The gap also appeared in studies of real work. In the twenty-year study, experts made real forecasts about elections, economies and wars, both inside and outside their own fields.[4][2] A recent study of national security officials, who made tens of thousands of assessments of current facts and future events, found the same gap.[26] Earlier lab experiments found it in a purer form. In one of five general-knowledge experiments, answers people called near-certain, at hundred-to-one odds, were right only about three times in four.[7]
Rubric spread
91
→ 70 /100
Highest to lowest rubric score across the ranked studies.
04Beyond the top five
Wider evidence
What the wider evidence adds.
A view called ecological rationality pushes back on treating mental shortcuts as simple errors. Its main proponent argues that simple rules of thumb can beat careful calculation under the right conditions.[3] This does not overturn the overconfidence research; it marks its limits. This article's own reading, untested, is that rules of thumb work best where feedback shows when they fail. That matters because part of the fault may lie in settings that rarely show people when they are wrong. Either way, misplaced certainty can be costly, as the next chapter shows.
5 studies, ranked.
Below: the anchor study in full; then every study's rubric score; then the supporting studies in ranked order.
01Anchor
Expert Political Judgment: How Good Is It? How Can We Know?
This study followed two hundred and eighty-four political and economic experts, such as university professors, government advisers and think-tank analysts, for two decades. It checked each prediction against what actually happened.
Rubric breakdown
The strongest studies, ranked by methodological weight.
Each scored 0 to 100 against a six-criterion rubric, tagged by design and year; the anchor leads.
02
Calibration of Probabilities: The State of the Art to 1980
The review compared people's stated confidence with how often they were right.[8]
85/100
03
The Trouble with Overconfidence
Three kinds of overconfidence (overestimation, overplacement and overprecision) responded differently to task difficulty.[6] Overprecision was the most persistent, appearing across the settings studied.
79/100
04
Neural correlates of metacognitive ability and of feeling confident
In healthy adults doing a social and factual reasoning task, feeling more confident went with more activity in areas linked to reward and memory (the striatum and hippocampus).[15] Judging one's own accuracy well went with less activity in a region at the front of the brain. These are correlations, so they cannot show that separate brain systems produce the two.
74/100
05
Psychological Strategies for Winning a Geopolitical Forecasting Tournament
Training in probabilistic reasoning gave reliable gains in forecasting accuracy in a randomised tournament.[31] Across four tournament years, the gains held on a Brier score, which rates how close forecasts came to what happened.[32]
70/100
04Stakes
The Cost of Misplaced Certainty Across Four Domains
Overconfidence has been linked to real costs in money, health, security and public life. These findings describe the groups studied; they do not predict how any one person's decisions will turn out.
Financial
In a large study of brokerage accounts, men traded nearly half as much again as women.[22] After adjusting for risk, they earned nearly one and a half percentage points less a year.[22] The authors inferred overconfidence from these gender-linked patterns rather than measuring it, so the result fits an overconfidence account but does not test it directly.[22] In another study, the odds of an acquisition were about two-thirds higher for chief executives classified as overconfident.[23] Their average market-adjusted return when they announced a deal was more negative than other chief executives'.[23]
frequent trading, impulsive trades, returns that lag the market
Medical
As the opening noted, in one study of patients who died in intensive care and underwent autopsy, diagnoses made with complete certainty were wrong about two times in five.[20] That figure does not describe medical judgement in general.[20] In a study in Senegal, health workers judged overconfident were about a quarter less likely to manage patients correctly.[29] They also took nearly a fifth fewer diagnostic steps.[29] In an experiment using written cases, the more-confident resident doctors spent about an eighth less time on average, but neither time nor diagnostic accuracy differed significantly between groups.[30]
settling on a diagnosis too early, skipping other possibilities, unchecked confidence
Strategic / Geopolitical
A historical analysis argues that overconfident leaders overrated their own strength and underrated their opponents.[25] Its cases include the First World War, Vietnam and the Cuban Missile Crisis.[25] The study of national security officials described in the opening found the same gap among working professionals, not only in the lab.[26]
intelligence failures, plans that overrun, conflicts that escalate through misread signals
Information / Democratic
In a national survey, three in four Americans overrated their ability to tell real news headlines from false ones.[27] On average they placed themselves more than twenty percentile points higher than their scores warranted.[27] Overconfidence was linked to greater willingness to share false political stories, including in analyses that accounted for people's actual skill at telling real from false.[27] In another survey, the share reporting tobacco use was more than twice as high in the low-health-literacy group as in the functional-literacy group.[28]
sharing without checking, resisting correction, a false sense of expertise
05Protocol
A Four-Step Calibration Protocol for High-Stakes Decisions
These steps help you record predictions, question plans and use past outcomes. They are habits informed by the studies above, not financial or medical advice.
The protocol, as a sequence.
Ongoing → Pre-decision → Estimation → Training
Prediction Tracking
Keep a simple prediction log: write down what you expect, how sure you are as a percentage, and why. Then record what actually happened.
As the mechanism chapter noted, slow or unclear feedback can make errors harder to spot.[37] Confidence judgements can be assessed once their outcomes are known.[8] A log gives you that check on purpose: group your predictions by how sure you were, and see how often each group came true.
Vague notes ("I thought this would go well") defeat the purpose. Put a number on your confidence before you know the outcome. Build up a record over time; a small set of predictions can give a misleading pattern.
Premortem
Before you sign off on a major plan, imagine it has already failed. Write down each likely reason why.
A controlled experiment compared five methods.[35] College students judged a fictional epidemic-response plan.[35] A premortem, imagining failure in advance, reduced their confidence in the plan more than the other four methods.[35] This tested confidence ratings, not whether plans succeeded.[35]
Running the premortem after the decision, when social pressure makes raising concerns feel disloyal rather than useful.
Reference Class Forecasting
When you estimate costs, timelines or odds of success, first find how similar past projects turned out. Use the spread of those outcomes, and choose a margin that fits the risk you can accept.[36]
One paper documents inaccurate forecasts for transport projects.[36] It illustrates reference class forecasting, basing estimates on similar past projects, as a way to make a more cautious estimate.[36] Earlier work argued that planners go wrong mainly by looking only at their own plan (the "inside view").[38] The fix is to look at how similar projects went (the "outside view").[38] In the fourth tournament year, forecasters who reported using comparison classes, another name for reference classes, tended to make more accurate forecasts.[32]
Deciding your project is so unusual that no past projects compare. The research suggests this is usually the inside view in disguise.
Structured Debiasing
Take a structured course in probabilistic reasoning. Look for one that covers base rates (how often something happens in general) and updating estimates as evidence arrives. It should also cover comparison classes, evidence against your view, and combining several people's forecasts.
In one study, a single training session played as a video game reduced measured bias by nearly a third at minimum.[33] The effects were still there two months later.[33] Training aimed at reducing bias is called debiasing. In a later field study, trained graduate students were 19% less likely to choose the worse option in an educational business-case exercise, not a live business decision.[34] That is the corrected figure: a published correction lowered the original estimate of 29%.[34]
Assuming expertise makes training unnecessary. As the mechanism chapter noted, experts tend to be overprecise too.[21] Expect modest gains overall, even though one game-based study found larger ones. In one experiment, a single training session reduced confirmation bias in national risk analysts and a matched student group.[40]
06Verdict
The verdict.
Bottom line
Your brain will keep producing a feeling of certainty. What you can change is whether you test that feeling before you act on it.
The argument
Most people, experts included, feel surer than their record justifies. Slow or unclear feedback can make that gap harder to see and fix.[37] Confidence can be checked against what actually happens: write down predictions with a number for how sure you are, then compare them with the outcomes.[8] In trials, short training and simple habits narrowed the gap, but the gains were modest and depended on the task.[26][39]
Only a record can tell you whether your own sense of certainty holds up better than the national security officials' did.
Judgement can be trained, at least in part. In a forecasting tournament, training in probabilistic reasoning improved forecasting accuracy.[31] Other studies, using games and a field exercise, reduced specific biases through different kinds of training.[33][34] The gains were modest and tied to the tasks studied. The field study tested one business case.[34] It was an educational exercise, not a live business decision, so it does not show gains across decisions in general.[34] The aim is not to feel less sure. It is to turn your confidence into a prediction you can check, using logs, premortems and reference classes.
It also helps to know what these studies do not show. Most of these studies measured groups of people on particular tasks, such as quiz questions, political forecasts, diagnoses and trades. They describe group averages, not any one reader. The brain findings are correlations, and the evolutionary account is a model.
Some researchers call the goal epistemic humility: a clear sense of the limits of what you know. On this view it is a set of habits and outside checks. The useful question is which checks you will put in place.
Confidence can be checked against outcomes.
Calibration is measurable
Confidence can be checked against how often similar judgements prove correct.[8]
Certainty can crowd out checking
Overconfident judgements tend to feel right, which can mean fewer checks: in one study above, health workers judged overconfident took fewer diagnostic steps.
Judgement can improve with training
Structured training of different kinds helped on the tasks studied: forecasting accuracy in one tournament, specific biases in short sessions and one field exercise. The gains were modest and depended on the task; one training made analysts more underconfident on yes-or-no questions. The gains lasted for the periods measured: two months in one study, and across tournament years in another.
Put it to work
Where this science goes next on HPC
07Bibliography
The bibliography.
-
01
Journal
doi: 10.1126/science.185.4157.1124
Judgment under uncertainty: Heuristics and biases
-
02
Book
no doi
Thinking, fast and slow
-
03
Book
no doi
Gut feelings: The intelligence of the unconscious
-
04
Book
no doi
Rubric 91/100
Expert political judgment: How good is it? How can we know?
-
05
Book
no doi
Superforecasting: The art and science of prediction
-
06
Review
doi: 10.1037/0033-295X.115.2.502
Rubric 79/100
The trouble with overconfidence
-
07
Journal
doi: 10.1037/0096-1523.3.4.552
Knowing with certainty: The appropriateness of extreme confidence
-
08
Chapter
doi: 10.1017/cbo9780511809477.023
Rubric 85/100
Calibration of probabilities: The state of the art to 1980
-
09
Journal
doi: 10.1016/0001-6918(81)90005-6
Are we all less risky and more skillful than our fellow drivers? Acta Psychologica, 47(2), 143–148
-
10
Journal
doi: 10.1207/s15516709cog2605_1
The misunderstood limits of folk science: An illusion of explanatory depth
-
11
Journal
doi: 10.1038/nature10384
The evolution of overconfidence
-
12
Journal
doi: 10.1037/0022-3514.77.6.1121
Unskilled and unaware of it: How difficulties in recognizing one's own incompetence lead to inflated self-assessments
-
13
Journal
doi: 10.1016/j.obhdp.2007.05.002
Why the unskilled are unaware: Further explorations of (absent) self-insight among the incompetent
-
14
Journal
doi: 10.1016/j.intell.2020.101449
The Dunning-Kruger effect is (mostly) a statistical artefact
-
15
Journal
doi: 10.1093/scan/nsw093
Rubric 74/100
Neural correlates of metacognitive ability and of feeling confident: A large-scale fMRI study
-
16
Review
doi: 10.1098/rstb.2011.0417
The neural basis of metacognitive ability
-
17
Journal
doi: 10.1038/nn.2949
How unrealistic optimism is maintained in the face of reality
-
18
Review
doi: 10.1016/j.cub.2011.10.030
The optimism bias
-
19
Journal
doi: 10.1037/h0022125
Overconfidence in case-study judgments
-
20
Review
doi: 10.1016/j.amjmed.2008.01.001
Overconfidence as a cause of diagnostic error in medicine
-
21
Review
no doi
Are experts overconfident? An interdisciplinary review
-
22
Journal
doi: 10.1162/003355301556400
Boys will be boys: Gender, overconfidence, and common stock investment
-
23
Journal
doi: 10.1016/j.jfineco.2007.07.002
Who makes acquisitions? CEO overconfidence and the market's reaction
-
24
Meta
doi: 10.1108/RBF-01-2020-0020
Overconfidence and financial decision-making: A meta-analysis
-
25
Book
doi: 10.4159/9780674039162
Overconfidence and war: The havoc and glory of positive illusions
-
26
Journal
doi: 10.1353/tns.00011
The World Is More Uncertain Than You Think: Assessing and Combating Overconfidence Among 2,000 National Security Officials
-
27
Journal
doi: 10.1073/pnas.2019527118
Overconfidence in news judgments is associated with false news susceptibility
-
28
Journal
doi: 10.1007/s10880-022-09895-4
Overconfidence in managing health concerns: The Dunning-Kruger effect and health literacy
-
29
Journal
doi: 10.1016/j.joep.2019.102213
Overconfident health workers provide lower quality healthcare
-
30
Journal
doi: 10.2147/AMEP.S442689
Overconfidence, time-on-task, and medical errors: Is there a relationship? Advances in Medical Education and Practice, 15, 133–140
-
31
Journal
doi: 10.1177/0956797614524255
Rubric 70/100
Psychological strategies for winning a geopolitical forecasting tournament
-
32
Journal
no doi
Developing expert political judgment: The impact of training and practice on judgmental accuracy in geopolitical forecasting tournaments
-
33
Journal
doi: 10.1177/2372732215600886
Debiasing decisions: Improved decision making with a single training intervention
-
34
Journal
doi: 10.1177/0956797619861429
Debiasing training improves decision making in the field
-
35
Chapter
no doi
Evaluating the effectiveness of the PreMortem technique on plan confidence
-
36
Journal
doi: 10.1177/875697280603700302
From Nobel Prize to project management: Getting risks right
-
37
Chapter
no doi
Debiasing. In D. J. Koehler & N. Harvey (Eds.), Blackwell handbook of judgment and decision making (pp. 316–337). Blackwell
-
38
Journal
doi: 10.1287/mnsc.39.1.17
Timid choices and bold forecasts: A cognitive perspective on risk taking
-
39
Journal
doi: 10.1002/acp.4236
The effect of calibration training on the calibration of intelligence analysts' judgments
-
40
Journal
doi: 10.1038/s41598-025-28794-w
Debiasing training reduces confirmation bias in national risk analysts
-
41
Review
doi: 10.3389/fpsyg.2021.802439
The impact of cognitive biases on professionals' decision-making: A review of four occupational areas
No entries match the current filter and search.
The protocol card
The Overconfidence Effect: What the Evidence Shows About Feeling Sure
One sheet. The four moves, in the order the day runs them.
-
01
Ongoing
Prediction Tracking
Keep a simple prediction log: write down what you expect, how sure you are as a percentage, and why. Then record what actually happened.
AvoidVague notes ("I thought this would go well") defeat the purpose. Put a number on your confidence before you know the outcome. Build up a record over time; a small set of predictions can give a misleading pattern.
-
02
Pre-decision
Premortem
Before you sign off on a major plan, imagine it has already failed. Write down each likely reason why.
AvoidRunning the premortem after the decision, when social pressure makes raising concerns feel disloyal rather than useful.
-
03
Estimation
Reference Class Forecasting
When you estimate costs, timelines or odds of success, first find how similar past projects turned out. Use the spread of those outcomes, and choose a margin that fits the risk you can accept.
AvoidDeciding your project is so unusual that no past projects compare. The research suggests this is usually the inside view in disguise.
-
04
Training
Structured Debiasing
Take a structured course in probabilistic reasoning. Look for one that covers base rates (how often something happens in general) and updating estimates as evidence arrives. It should also cover comparison classes, evidence against your view, and combining several people's forecasts.
AvoidAssuming expertise makes training unnecessary. As the mechanism chapter noted, experts tend to be overprecise too. Expect modest gains overall, even though one game-based study found larger ones. In one experiment, a single training session reduced confirmation bias in national risk analysts and a matched student group.
hiperformanceculture.com · 41 sources · 5 April 2026
How this was made
The desk, the instrument, and when this page is next examined.
- Desk
- HPC Editorial
- Instrument
- 6 weighted criteria, 100 points see every score
- Register
- 41 sources, 33 with a resolving DOI read them
- Reading time
- 22 min read
Revision log
- First published.
- Last revised. Sources re-checked against the register above.
- Next review Due by , or earlier if a new controlled trial or systematic review lands.
THE HAND-OFF. Performance Scan: find your limiting factor. Three minutes, 24 questions, six systems. Your answers never leave your device.