Overtrading: The Behavioural Driver and Its Arithmetic
6 steps · one page
In short
The finding that investors transact more than serves them is among the better-established in the field. The interesting question is not whether it happens but why it is so stable, and the answer is that the behaviour produces its own justification.
Scope, and a boundary that keeps this article from duplicating two others. What frequent transacting costs is computed in Bid, Ask and the Spread, and the documented outcome evidence for short-horizon traders is reported in Trading Styles: Day, Swing and Position. Neither is repeated here. This article covers what drives the behaviour, and one piece of arithmetic that belongs nowhere else: how much apparent evidence of skill activity generates on its own. It supplies no rule about frequency. Figures are small non-canonical illustrations.
The driver, as far as the evidence supports it
The mechanism established in Cognitive Biases in Investing is the load-bearing one, and it is worth restating in this context.
Realised outcomes are salient; open positions are not. An investor who tends to close winners and retain losers therefore accumulates a memory of decisions that worked — and confidence built on that record is an honest inference from a biased sample rather than a character defect.
Greater confidence supports more activity. More activity produces more decisions. And more decisions produce more of the salient, memorable outcomes that sustained the confidence in the first place. That is a closed loop, and no step in it requires anyone to behave unreasonably.
What activity alone produces, before any skill is involved
This is the part that is rarely quantified, and it is the reason the loop is so hard to interrupt from inside.
Take decisions that are genuinely coin-flips — no skill whatever, an even chance each time.
| Number of decisions | Outcome | Probability from chance alone |
|---|---|---|
| 10 | 8 or more correct | 5.47% |
| 20 | 13 or more correct | 13.16% |
| 20 | 14 or more correct | 5.77% |
| 50 | 32 or more correct | 3.25% |
Worked example — a record that looks like skill, produced by nothing. Getting 13 or more of 20 decisions right happens 13.16% of the time by chance. Roughly one person in eight with no ability at all will produce a 65% hit rate over twenty decisions — and that person has a record they can describe accurately, defend in detail, and reasonably believe. Across a population the effect is stark: among 1,000 people each making 10 such decisions, about 54.7 will get 8 or more right, about 10.7 will get 9 or more, and about 1 will get all ten. Those people are not lying and are not fooling themselves through carelessness. They are reading a small sample correctly and drawing the only inference available from it. Figures are illustrative and describe no actual decisions.
Why more activity makes this worse rather than better
The intuition is that a longer record settles the question. It does, eventually, but not at the frequencies individuals operate at — and in the meantime activity produces more of the specific evidence that feels most like skill.
| Decisions taken | Longest run of consecutive successes expected from chance alone |
|---|---|
| 20 | about 4 |
| 50 | about 5 |
| 100 | about 6 |
| 250 | about 7 |
Worked example
The point, stated carefully because it is easy to overstate. Someone making 250 decisions in a year should expect a run of about seven consecutive successes with no skill whatever. A streak is the most persuasive form of self-evidence available — it feels categorically different from a good average — and it is precisely the form that activity manufactures. So the person who transacts most has the most impressive-looking record to reason from, independent of ability. This does not establish that any particular record is luck. What it establishes is that a record of this kind cannot distinguish the two, which is a different and more useful claim. The related finding — that a small minority of short-horizon traders do show persistent skill — is reported in Trading Styles, along with the reason it does not help someone deciding whether to begin. (The expected longest run of successes in n fair even-chance decisions is 3.7, 5.0, 6.0 and 7.3 at n = 20, 50, 100 and 250, computed exactly from the run-length distribution.)
The costs run the other way
While the apparent evidence improves with activity, the arithmetic does not. The spread is charged per transaction, so its annual weight scales directly with frequency — the multiplication set out in Bid, Ask and the Spread. And the delay component of implementation shortfall, computed in Slippage and Execution Quality, is incurred on every decision rather than on every holding.
So the two curves diverge: confidence tracks the number of decisions upward, and net outcomes track the number of decisions downward. They are driven by the same variable in opposite directions, which is what makes the pattern self-sustaining rather than self-correcting.
Frequently asked
8 questions
What drives overtrading?
The mechanism from the biases article is the load-bearing one: realised outcomes are salient and open positions are not, so an investor who closes winners and retains losers accumulates a memory of decisions that worked. Confidence built on that is an honest inference from a biased sample.
Why is the pattern so stable?
Because it is a closed loop. Greater confidence supports more activity, more activity produces more decisions, and more decisions produce more of the salient outcomes that sustained the confidence — with no step requiring anyone to behave unreasonably.
How much apparent skill does chance produce?
A great deal at small samples. Getting 13 or more of 20 decisions right happens 13.16% of the time by chance — roughly one person in eight with no ability produces a 65% hit rate over twenty decisions.
What does that look like across many people?
Among 1,000 people each making 10 coin-flip decisions, about 54.7 get 8 or more right, about 10.7 get 9 or more, and about 1 gets all ten. They are reading a small sample correctly and drawing the only inference available from it.
Doesn't more trading eventually settle whether someone has skill?
Eventually, but not at the frequencies individuals operate at — and meanwhile activity produces more of the evidence that feels most like skill. Someone making 250 decisions a year should expect a run of about seven consecutive successes with no skill whatever.
Does that prove a good record is luck?
No, and the article does not claim it. What it establishes is that a record of this kind cannot distinguish luck from skill — a different and more useful claim.
How do the costs behave?
In the opposite direction. The spread is charged per transaction so its annual weight scales with frequency, and the delay component of implementation shortfall is incurred on every decision rather than on every holding.
Why is this self-sustaining rather than self-correcting?
Because confidence tracks the number of decisions upward while net outcomes track the same number downward. One variable drives both, in opposite directions.
References
- Barber and Odean (2000) — Trading Is Hazardous to Your Wealth: The Common Stock Investment Performance of Individual Investors, Journal of Finance 55(2) —
- Barber and Odean (2001) — Boys Will Be Boys: Gender, Overconfidence, and Common Stock Investment, Quarterly Journal of Economics 116(1) —
- Odean (1998) — Are Investors Reluctant to Realize Their Losses?, Journal of Finance 53(5) —
Educational and informational only — not investment advice, a recommendation, or an offer to buy or sell any security. Investing involves risk, including the possible loss of principal. Worked examples use fictional companies and figures.