Behavioural finance, and why technical analysis works
The answer
Prices are set by people, people make the same errors repeatedly, and some of those errors leave measurable traces in prices. That explains a small number of specific patterns with named mechanisms. It does not license every shape anybody draws on a chart.
Why this costs you money
Here is a cost you can measure today, and most readers are carrying it.
You buy 2 stocks. One rises 20%. One falls 20%. Which do you sell?
Almost everybody sells the winner. Taking a gain feels like being right. Selling the loser converts a temporary disappointment into a permanent verdict. So the winner leaves the portfolio in weeks and the loser stays for years.
Do that for 3 years and 2 things happen. Your average win becomes small, because you cut winners early. Your average loss becomes large, because you hold losers until they are large. Your payoff ratio — average win divided by average loss — falls below 1, and you now need to be right more than half the time to break even.
No improvement to your entries can fix this. It is an exit problem, and it has a name. It is the disposition effect, one of the best documented findings about individual investors anywhere.
The second cost is different. A reader learns that "technical analysis works because of psychology" and uses that sentence to defend any drawing they made. The sentence is doing no work. It has become permission.
How it works
Six mechanisms are documented well enough to build on. For each one, what it is and what it would produce on a chart.
Loss aversion, producing the disposition effect. People feel a loss more strongly than a gain of the same size. So they sell winners early and hold losers. That gives a testable prediction: selling pressure appears near a price where many people bought, and good news is absorbed slowly rather than all at once. Slow absorption of good news is what post-earnings announcement drift describes, documented since the late 1980s. Anchoring. People fix on a reference number and adjust away from it too little. The reference numbers in markets are obvious: your purchase price, the 52-week high, the round number. Anchoring predicts orders cluster at those prices, and price clustering at round numbers is well documented.
Herding. People take the actions of others as information, especially when uncertain. Herding predicts a move continues past the point new information justifies, and then reverses. That is momentum followed by long-horizon reversal.
Recency and availability. Recent and vivid events are weighted too heavily. This predicts overreaction after dramatic news and a slow return afterwards.
Overconfidence. People believe their judgement is better than it is. This predicts heavy trading, high volume and poor net returns. Note what it does not predict: it says nothing about the shape of a price pattern. It explains your behaviour, not the chart.
Limits to arbitrage. This is the mechanism that makes the other 5 matter. In theory a professional removes any mispricing immediately. In practice correcting one requires capital, borrowing the security, and surviving the position getting worse before it gets better. A fund whose clients withdraw during that period must close at the worst moment. So a mispricing can persist even when everybody can see it.
The line
Now the part this article exists for: the difference between a claim with a mechanism and a pattern found in noise.
A claim with a mechanism passes 3 tests.
- You can name the mechanism before you see the result. "People hold losers, so selling appears where many people bought" is a mechanism. "This shape works" is not.
- The pattern is visible at the right edge, before the outcome. If you can only identify it after knowing what happened, you identified an outcome, not a pattern.
- You can say how many alternatives you examined. One shape found after scanning 200 charts is a search result. Article 8 in cluster 15 has the arithmetic.
Apply those 3 and the line becomes sharp.
| Claim | Mechanism? | Verdict |
|---|---|---|
| Selling appears near a price where many bought | Disposition effect | Testable, supported |
| Orders cluster at round numbers | Anchoring, plus order-book habits | Testable, supported |
| Strong recent performers keep going for months | Herding, underreaction | Documented, and it decays |
| Price reverses at 61.8% of a prior move | None known | Decoration |
| A diagonal line through 2 points predicts a turn | None known | Decoration |
| A named shape with 5 parts predicts a target | None for the target | Partly decoration |
That table is the article in 6 rows. The behavioural mechanisms are real. They support far fewer techniques than they are used to defend.
What it tells you, and what it does not
A behavioural explanation is not evidence. It is a reason something could be true. You still have to test whether it is. Every failed strategy in history had a plausible story attached to it.
Documented effects decay after publication. Researchers examined a large number of published return predictors and found measured returns fell substantially after the papers appeared. If an effect is real and public, capital arrives to exploit it. So "there is a study" is not "it still works".
Biases describe you more reliably than they describe the market. Your own overconfidence is measurable in your trade count. The market's is an inference. Start with the measurable one.
The mechanisms compete. Herding predicts continuation. Overreaction predicts reversal. Both are documented, at different horizons. A theory that predicts both directions predicts nothing unless you state the horizon.
The decision rule
Before you trade any pattern, answer 3 questions in order.
- What mechanism would produce this, and would it have produced it before I
looked? No mechanism, no position.
- Is it visible at the right edge? Cover the chart from today onward and
check whether you could name it without the outcome.
- How many shapes did I look at first? More than 20, and the threshold for
believing this one has to be much higher.
And once a quarter: run the exercise below on your own trade history. The biases you can fix are your own.
Try this now
Five minutes, a spreadsheet, and the single most revealing number in your own trading record.
- Export your trade history from your broker's Reports, Tradebook or P&L section. Keep only closed positions, with the entry date, exit date and profit or loss for each.
- Add a column for holding period in days: exit date minus entry date.
- Split the list into winners and losers by whether the profit is positive or negative.
- Compute the average holding period of the winners.
- Compute the average holding period of the losers.
- Divide the second by the first.
What you should see. In most retail accounts the losers are held far longer than the winners. A ratio of 2 is common. A ratio of 4 or more is not unusual.
That ratio is the disposition effect, in your own data, with your own money. You are not reading about a study. You are looking at the finding.
Now the 2-minute second half, which turns the number into a decision. Look at the losers you held longest. For each, write 1 sentence answering: on the day I decided not to sell, what was the reason? Sort the sentences into 2 groups. Sentences about the business — a delayed order, a plant starting up, a margin recovery you expected — are theses, and a thesis is allowed to take time. Sentences about the price — "it will come back", "I am down too much to sell now", "I will exit at my cost" — are the disposition effect wearing a reason. Count the second group. That count tells you what to fix.
If the ratio is below 1.5, check the other direction. Some readers hold winners for years and cut losers immediately, which is the healthier failure. If so, look at whether your winners are small, which would mean you exit at a fixed profit target rather than at a condition.
Three real cases
1. Barber and Odean, individual investor accounts, 1991 to 1996 — the cost of overconfidence, measured Two researchers obtained the trading records of tens of thousands of households at a large discount broker. Households that traded most actively earned annual returns well below the market, and the gap was driven mainly by trading costs rather than bad stock selection. The finding is not that individuals pick bad stocks. It is that they trade too much, because they believe their judgement is better than it is.
2. 3Com and Palm, March 2000 — the mispricing everybody could see 3Com sold a small part of its subsidiary Palm to the public and announced it would distribute the rest to its own shareholders. Simple arithmetic said each 3Com share was worth at least the Palm shares attached to it. For months the market priced 3Com below the value of its Palm holding alone, implying a negative value for the rest of the company. Anybody could see it. Almost nobody could correct it, because borrowing Palm shares to sell was expensive and limited. A mispricing can survive being obvious, which is why limits to arbitrage matters more than any individual bias.
3. SEBI's studies of individual derivatives traders, 2023 and 2024 (India) — behaviour beating analysis, at population scale The Indian regulator published studies of profit and loss among individual traders in the equity derivatives segment. A study released in January 2023 reported that roughly 89% of individual traders lost money. A larger follow-up released in September 2024 covered 3 financial years and reported a loss rate of roughly 93%, across more than 1 crore individual traders, with aggregate losses in the region of Rs 1.8 lakh crore. The sample is not 1 broker's clients. It is close to the whole population. No comparable study exists for any other market.
The question that resolves it
A novice looks at a pattern and asks: does this work?
An expert asks: what would have to be true about the people on the other side for this to keep working, and is anybody paid to remove it?
The second question does 2 things at once. It forces you to name a mechanism. And it tells you how long the effect can survive, because an effect that professionals can remove cheaply will be removed.
What would make this wrong
If prices were set by participants with no systematic errors, and any mispricing were corrected instantly, no persistent price pattern could exist and every technique in this track would be decoration. That is the efficient market position in its strong form, and it is a serious position.
Three honest limits on the case made here.
Documented biases are mostly documented in individuals, and individuals are a minority of trading volume. Most volume in both markets comes from institutions and automated systems. An argument running from individual psychology to index prices has a gap in the middle, filled by limits to arbitrage, which is itself contested in size.
Behavioural finance has a replication problem. Several well-known effects in psychology have failed to replicate at their original strength. Treat any single study as weak evidence, and treat the effects reproduced across many markets and decades — the disposition effect, momentum, price clustering — as the reliable part.
The mechanisms do not deliver the precision technical analysis often claims. "Selling appears near prices where many people bought" is supported. "Resistance is at 1,847.50" is not, and no behavioural finding produces a number to 2 decimal places.
In India
India has something no other market has: population-scale evidence about individual traders, published by the regulator, free to download.
The SEBI studies described above cover crores of accounts rather than 1 broker's clients, and report loss rates near 9 in 10 among individual traders in equity derivatives. Two honest notes. A substantial share of the reported loss is transaction costs rather than market direction, which is the overconfidence finding in another form: high trading frequency guarantees a cost the strategy must overcome first. And the studies do not perfectly separate speculation from hedging. Neither point changes a loss rate that large.
Three Indian features make behaviour more expensive than it is elsewhere.
Derivatives are the entry point. In many countries a new retail participant starts with funds or shares. In India a large share started with weekly index options, the highest-leverage, shortest-horizon instrument available. An error that costs a share investor 10% over a year can cost an options trader the account in a month.
Lot sizes remove precision. You cannot buy 0.6 of a lot. When the correct position size is smaller than 1 lot, the lot size silently replaces the sizing rule. SEBI's October 2024 measures raised minimum contract values, which raises the smallest position an account can take. Trade data is available and almost nobody uses it. Every Indian broker provides a downloadable tradebook and profit and loss statement. The exercise in this article needs nothing else.
In the United States
The United States has the deepest academic literature on investor behaviour, and almost all of it is built on individual broker datasets from the 1990s.
The Barber and Odean work described above is the foundation, built on records from 1 large discount broker over roughly 6 years. Related work by the same authors found that men traded more and earned less net of costs.
The United States also provides a structural protection India lacks. Retirement saving through employer plans defaults many people into diversified funds with automatic monthly contributions. Those savers are protected from their own behaviour by the design of the system, not by their discipline.
American retail trading also has episodes that look behavioural in real time. January 2021 produced a coordinated retail move into a few heavily shorted shares, combining herding with a genuine limit to arbitrage: short sellers had to buy to close, which pushed prices further.
Where they differ, and what that tells you
The best evidence in the world about individual traders is Indian, and Indian readers are the least likely to have read it.
American behavioural evidence is excellent work on small samples from 30 years ago: 1 broker, tens of thousands of accounts, the 1990s. Indian evidence is a regulator publishing outcomes for crores of accounts, updated recently. On sample size and recency the Indian evidence is stronger. On depth of analysis the American literature is stronger. Most readers have met neither.
The second difference is structural, and it decides more than any bias. An American saver is defaulted into diversified funds through an employer plan. An Indian saver has no equivalent default. The American system converts inertia into diversification. The Indian system leaves inertia unallocated, and an active choice into derivatives fills the gap.
In the United States, behaviour is partly managed by structure. In India, it has to be managed by the individual. An Indian reader copying American advice about "controlling emotions" is importing advice written for people whose largest financial decision was made for them automatically. The Indian reader has to build the structure: automatic monthly investment, a written position size rule, and a fixed exit condition decided before entry. Those 3 replace discipline with arrangement, and arrangement is what works.
Carry this
- Name the mechanism, or you found a pattern in noise.
- The disposition effect is the most expensive one, and it is in your exit rule, not your entry rule.
- Losers held longer than winners is the finding. Measure the ratio once a quarter.
- A behavioural explanation is a reason to test something, never a result.
- Documented effects decay after publication. "There is a study" is not "it still works".