Skip to content
Menu
Finabase Learn

Articles · Behavioral finance

Self-attribution bias: crediting skill for wins and luck for losses

What self-attribution bias is, how it feeds overconfidence in finance models, what happened to online traders, and how to keep score of your trades fairly.

On this page
  1. The psychology
  2. Self-attribution in finance models
  3. Evidence: do the slow die first?
  4. Worked example 1: how much is luck?
  5. Worked example 2: lopsided scorekeeping
  6. Keeping honest score
  7. How Finabase helps
  8. Frequently asked questions
  9. Summary
  10. Check your understanding
  11. Exercise

Self-attribution bias is the habit of taking credit for good outcomes and blaming bad ones on bad luck or outside forces. When a stock you picked rises, it was your analysis. When it falls, it was the Fed, a market panic or a surprise nobody could have foreseen. Psychologists also call it the self-serving bias, and it is one of the most consistent findings in social psychology.

For investors the bias has a specific cost: it corrupts learning. Every trade is a small experiment, and an investor who counts the wins as evidence of skill and the losses as noise will conclude, over time, that he or she is better than the record shows. That inflated self-image is the fuel for the overconfidence and excessive trading described in Overconfidence and overtrading: the research on trading too much. This article covers the bias itself, the finance models built on it, what happened to a group of successful investors who moved online, and a worked example of how lopsided scorekeeping turns a coin flipper into a self-declared expert.

The psychology

Experiments since the 1970s show that people attribute success to themselves and failure to circumstances, especially when the outcome matters to their self-image. Dale Miller and Michael Ross reviewed that evidence in 1975. The same year, Ellen Langer and Jane Roth published a study with a memorable title, "Heads I win, tails it's chance." People watching or predicting coin tosses who happened to get early successes rated their own ability higher, expected to do better in the future and remembered more successes than they had actually had, even though the task was pure chance.

Markets are close to an ideal setting for the bias. Outcomes are noisy, so almost any result can be explained as either skill or luck. Feedback is fast and frequent. And there is always a plausible outside culprit for a loss. The bias also works hand in hand with Hindsight bias: why past markets look easier than they were, which rewrites what we believed before the outcome, and with confirmation bias (see Anchoring and confirmation bias: how investors fool themselves), which steers attention to the evidence that supports us.

Self-attribution in finance models

Daniel, Hirshleifer and Subrahmanyam (1998)

Kent Daniel, David Hirshleifer and Avanidhar Subrahmanyam built a model of the market around two biases. Investors are overconfident about the precision of their private information, and their confidence changes through biased self-attribution: when public news later confirms their private view, their confidence rises a lot; when public news contradicts it, confidence falls only a little.

Two predictions follow. Prices overreact to private information, because traders lean on it too heavily, and underreact to public information that contradicts them. And because confidence keeps building after confirmations, an initial move can continue for a while before reversing, a pattern of short-run momentum followed by long-run reversal that matches the evidence discussed in Representativeness and overreaction: why glamour stocks disappoint. The model is one of the main behavioral explanations for why both momentum and long-term reversals appear in stock returns.

Gervais and Odean: learning to be overconfident

Simon Gervais and Terrance Odean asked a simpler question in 2001: if traders learn about their ability from their own results, why do they not end up accurately calibrated? Their model gives traders an unknown skill level and lets them learn from each outcome, but with one distortion: they give themselves too much credit for successes.

The result is a predictable career arc. Overconfidence is highest early, among traders who have had some success but not yet a long record. Successful traders become overconfident; unsuccessful ones do not, or drop out. Over a long career the evidence eventually overwhelms the bias and confidence moves back toward reality, but in the meantime overconfident traders trade too much and take too much risk. A trader's greatest overconfidence, in this account, comes right after a good run.

That fits market-wide evidence. Meir Statman, Steven Thorley and Keith Vorkink found in 2006 that trading volume in US stocks rises after periods of high market returns, as if investors read a rising market as proof of their own skill.

Evidence: do the slow die first?

Brad Barber and Terrance Odean studied 1,607 investors at a large US discount brokerage who switched from trading by phone to trading online during the 1990s. Their paper, published in 2002, found a pattern that is hard to explain without self-attribution:

  • Before going online, these investors had done unusually well: they beat the market by more than 2% a year.
  • After going online, they traded more actively and more speculatively, and their results fell: they lagged the market by more than 3% a year.

The likely story is that a strong run made them more confident in their ability, the confidence drew them to a faster, cheaper way to trade, and the extra trading cost them. The earlier success may have been mostly luck, or the market conditions of the time, but it was read as skill. This paper is distinct from the same authors' broader evidence that heavy traders earn less, which is summarized in Overconfidence and overtrading: the research on trading too much.

Worked example 1: how much is luck?

Suppose a trader has no edge at all: each trade is a fair coin flip, 50% wins and 50% losses of equal size. How often would such a trader still post a record that looks like skill, say winning at least 60% of trades?

Chance that a trader with no edge (50/50 per trade) wins at least 60% of trades 0% 10% 20% 30% 40% 0 100 200 300 400 500 20 trades: 25% 100 trades: 2.8% 200 trades: 0.28% Number of independent trades
Exact binomial probabilities, independent trades, 50% win chance each.
  • With 20 trades, the chance of winning 12 or more by luck alone is about 25%.
  • With 50 trades, winning 30 or more happens about 10% of the time.
  • With 100 trades, about 2.8%; with 200 trades, about 0.28%.

A 60% win rate over a season of 20 trades is weak evidence of skill: a quarter of pure coin flippers achieve it. It takes a record of hundreds of trades before chance becomes an unlikely explanation. Win rate is also only half the picture, since the size of wins and losses matters; see Process over outcome: judging investment decisions under luck.

Worked example 2: lopsided scorekeeping

Now give the same no-edge trader a belief to update. Before trading, she thinks there is a 20% chance she is skilled, where "skilled" means winning 60% of trades and "unskilled" means 50%. She makes 40 trades and wins exactly 20, which is exactly what an unskilled trader would expect.

A fair learner applies Bayes' rule to every outcome: each win multiplies the odds of being skilled by 0.6 ÷ 0.5 = 1.2, each loss by 0.4 ÷ 0.5 = 0.8. A self-attributing learner counts the wins the same way but treats each loss as only half as informative, "partly bad luck," so a loss multiplies the odds by the square root of 0.8, about 0.89.

“Am I skilled?” after 40 trades with exactly 20 wins: two ways of keeping score 0% 20% 40% 60% 80% 100% 0 5 10 15 20 25 30 35 40 Starting belief: 20% Counts wins and losses the same (Bayes) Counts each loss at half weight 51% 10% Trade number (invented sequence; the trader has no edge)
Invented sequence of 20 wins and 20 losses. Skilled = 60% win rate, unskilled = 50%; prior 20%.

Nothing in the trades changed; only the scorekeeping did. That is Gervais and Odean's mechanism in miniature, and it explains why a trader can grow more confident through a period of completely average results.

Keeping honest score

  • Write the reason before the outcome. A dated note of why you bought, what you expect and what would prove you wrong cannot be reinterpreted later. See How to keep an investment decision journal (with a template).
  • Count every trade. Review all closed trades, not the memorable ones, and compute the win rate and the average gain and loss.
  • Grade the decision, not the result. Ask whether the reasoning was sound given what was known, for wins as well as losses. A win for the wrong reason is a warning.
  • Use a benchmark. Compare the portfolio's return with an index fund over the same period. A rising market makes almost everyone look skilled.
  • Be most careful after a good run. The research suggests that is when overconfidence peaks.

How Finabase helps

Finabase keeps the record that self-attribution would rather forget. Paper trading keeps a full transaction history of every simulated trade, wins and losses alike, so a review cannot skip the losers. A symbol's Notes tab stores your reasons with a date before the outcome is known. A portfolio shows your holdings' performance over time so it can be compared with the market rather than judged on its own.

Frequently asked questions

What is self-attribution bias?

It is the tendency to credit your own skill for successes and to blame outside factors or bad luck for failures. In investing it leads people to overrate their ability because their wins count as evidence and their losses do not.

How is self-attribution bias different from overconfidence?

Overconfidence is the belief that you know more or are more skilled than you are. Self-attribution is one of the ways that belief is created and kept alive: by learning from outcomes in a lopsided way.

What did Barber and Odean find about online traders?

In a 2002 study of 1,607 investors who moved from phone to online trading in the 1990s, they found the investors had beaten the market by more than 2% a year before switching, then traded more and lagged the market by more than 3% a year afterward.

How many trades does it take to tell skill from luck?

Many more than most people think. A trader with no edge has about a 25% chance of winning at least 60% of 20 trades and about a 3% chance over 100 trades. Hundreds of trades, judged on both win rate and the size of gains and losses, are needed.

Summary

Self-attribution bias makes wins look like skill and losses look like bad luck. Because markets are noisy, the bias can run for years, slowly converting an ordinary record into a confident self-image. Finance models by Daniel, Hirshleifer and Subrahmanyam and by Gervais and Odean show how it can move prices and produce traders who are most overconfident right after a good run, and Barber and Odean's online investors fit that pattern. The worked examples show the core problem: a small sample can flatter anyone, and lopsided scorekeeping can make an average record look like talent. Writing reasons down in advance and counting every trade the same way are the most direct remedies.

Check your understanding

4 questions on this lesson. Each answer comes with the reason, and nothing is scored.

  1. Which statement is an example of self-attribution bias?

    1. “My winners came from my research; my losers were hit by a market I could not have predicted.”
    2. “I lost money on this trade.”
    3. “The index rose 10% this year.”
    Show the answer

    A. “My winners came from my research; my losers were hit by a market I could not have predicted.” Crediting wins to skill and blaming losses on outside forces is the bias. Applied to both, the same standard would treat either as possibly luck.

  2. A trader with no edge makes 20 trades. Roughly how likely is it that he wins at least 12 of them by chance?

    1. About 1%
    2. About 25%
    3. About 75%
    Show the answer

    B. About 25% With a 50% chance per trade, 12 or more wins out of 20 happens about a quarter of the time. Short records say little about skill.

  3. In Gervais and Odean’s model, when is a trader most overconfident?

    1. At the very start, before any trades
    2. Early in a career, after some success but before a long record has accumulated
    3. After many years of experience
    Show the answer

    B. Early in a career, after some success but before a long record has accumulated Taking too much credit for early wins inflates confidence; only a long record eventually brings it back toward reality.

  4. What did Barber and Odean find about investors who switched to online trading in the 1990s?

    1. They improved their returns after switching
    2. They had beaten the market before switching, then traded more and lagged the market afterward
    3. They stopped trading
    Show the answer

    B. They had beaten the market before switching, then traded more and lagged the market afterward A strong prior run likely raised their confidence, which led them to trade more actively and speculatively, and their returns fell.

Exercise

One task to do in Finabase. Nothing is graded or saved.

Open Paper trading and go to Transaction history (signed out, it shows a sample account). List your last ten closed trades as wins or losses. For each, write one line on why it turned out that way, then mark whether you named your own analysis or an outside cause.

Open Paper trading
What you should find

If your wins mostly cite your analysis and your losses mostly cite the market, news or bad luck, that is self-attribution. With only ten trades, the honest answer for most of them is that the result says little about skill either way.

Sources and further reading

  1. Langer and Roth — Heads I win, tails it’s chance: the illusion of control as a function of the sequence of outcomes in a purely chance task, Journal of Personality and Social Psychology 32(6), 1975
  2. Miller and Ross — Self-serving biases in the attribution of causality: fact or fiction? Psychological Bulletin 82(2), 1975
  3. Daniel, Hirshleifer and Subrahmanyam — Investor psychology and security market under- and overreactions, Journal of Finance 53(6), 1998, pp. 1839–1885
  4. Gervais and Odean — Learning to be overconfident, Review of Financial Studies 14(1), 2001, pp. 1–27
  5. Barber and Odean — Online investors: do the slow die first? Review of Financial Studies 15(2), 2002, pp. 455–487
  6. Statman, Thorley and Vorkink — Investor overconfidence and trading volume, Review of Financial Studies 19(4), 2006