Cognitive Biases: Overconfidence, Anchoring and Recency Bias

In a 1981 study, Ola Svenson asked students to rate their own driving. 93% of the American group placed themselves above the median. Half of any group sits below the median, by definition. The students were not lying. They believed it. That is overconfidence, and in markets it has a price tag you can measure.

This module covers four biases that travel together: overconfidence, anchoring, recency and availability. Each one alone costs a little. Stacked, as they usually are near the top of a small cap cycle, they cost a lot. I will show you each with Indian numbers, including my own graded record, because nobody gets a pass on this one.

Overconfidence, the bias that costs the most

Overconfidence shows up in markets in three ways.

• Overestimating skill: gains get credited to skill, losses to bad luck. After a bull run almost everyone believes they can pick stocks.
• Overtrading: if you trust your view more than the market's, you trade more. Every trade costs brokerage, STT, exchange charges, GST, stamp duty and the spread.
• Under diversifying: concentrating in the few names you feel sure about. That raises stock specific risk without raising expected return.

The classic evidence is Brad Barber and Terrance Odean's 2000 paper, "Trading is hazardous to your wealth". They studied 66,465 households at a large discount broker from 1991 to 1996. The average household turned over 75% of its portfolio a year and earned 16.4% net. The most active fifth earned 11.4% net, while the market returned 17.9%. In a follow up paper in 2001 they found men traded 45% more than women, and paid for it in lower net returns.

SEBI's Indian data rhymes almost line for line. In its July 2024 intraday study, the share of loss makers rose with trading frequency, reaching 80% for traders with more than 500 trades a year. Loss makers made more trades on average than profit makers. Female traders had a higher share of profit makers than male traders in all three years studied, and younger traders did worst: 76% of those under 30 lost money in FY23. In F&O, SEBI's August 2026 study found regular traders turned over about 46 times their portfolio in FY26 and lost more per head than new traders.

My own calibration check

I publish a graded scorecard of every call made on The Tanmay Edge, the pre market podcast I ran daily until September 2026. Season 2, 104 graded episodes from 17 April to 15 September 2026, scored 384 out of 558 points. That is a 69% hit rate, with 134 calls marked as outright misses. Every one of them is on the track record page.

I share this because it is the cleanest antidote to overconfidence I know. Write the call down before the event, grade it against the data after, and keep the misses visible. A 69% hit rate is useful. It is also a reminder that roughly one call in three is wrong, which is exactly why position size can never be a function of how sure I feel.

Anchoring, the bias of the first number

Anchoring is leaning too hard on the first number you see, then adjusting too little away from it. In Tversky and Kahneman's 1974 experiment, people spun a wheel that landed on 10 or 65, then estimated the percentage of African countries in the United Nations. Those who saw 10 gave a median answer of 25. Those who saw 65 said 45. A random number moved a factual estimate by 20 points.

Markets hand you anchors all day: the 52 week high, your purchase price, an IPO's issue price, a target in a report. Look at what happened in 2026.

An investor anchored on the 52 week high sees the Nifty 50 "15% off" and the Smallcap 250 "at a high". The P/E column, published by NSE, tells a different story: the large cap index got cheaper on earnings as it fell, while the small cap index climbed back to within 0.5% of its 2024 record at a higher P/E than it had at the record, 34.19 against 32.15. Neither number is a buy or sell signal. The point is that the distance from a high is a fact about the past price, not about value.

The fix: form your own view of value before you look at the chart, the high or a target. Then compare. If you start from the price, you will adjust from it, and not far enough.

Recency bias, the extrapolation trap

Recency bias is treating the last year or two as the new normal. After a great year, investors expect another. After a bad one, they expect worse. You can see it in who opens accounts and when. NSE's new investor registrations track the previous year's market far more closely than they track valuations.

The biggest year for new investors, FY22, came straight after the 71% rally of FY21. FY25 came after a 29% year. Most of those investors met their first real drawdown within a year or two of joining. The pattern is the same in fund flows. The base rate is the better guide: over long periods Indian large cap equity has compounded in low double digits, with regular falls of 15% to 25% on the way. Any two year window, good or bad, is a small sample of that.

Availability heuristic, the memorable vs the probable

Availability means judging how likely something is by how easily an example comes to mind. A friend's options win last month is vivid. The 87.7% of individual F&O traders who lost money in FY26 are a statistic. The vivid story wins.

It works the other way after a crash. A sharp fall is so memorable that for months afterwards many investors overrate the chance of another, sit in cash and miss the recovery. The Nifty Smallcap 250 fell 26.1% from its record into March 2025 and then climbed 34.7% off the low, to within 0.5% of the record, 18 months later. The investors who stayed out were reacting to the most available memory, not the most likely path.

The fix is to ask for the base rate before the story. Of all the people who tried this, how many succeeded? SEBI publishes that number for F&O. For most stories nobody does, which is itself a warning.

Three biases playing out together in Indian smallcap cycles

The biases rarely arrive alone. Small cap cycles are where they stack.

1. Recency sets the expectationAfter a long rally, last year's 40% or 60% starts to feel normal. New investors arrive with the recent past as their only reference.
2. Overconfidence sizes the betA few early winners feel like skill. Position sizes grow, and diversification shrinks.
3. Anchoring holds the loserWhen the cycle turns, the buy price or the old high becomes the target. "I will sell when it gets back."
4. Availability keeps you outAfter the fall, the memory of the loss makes the next good entry feel dangerous.

The defence against all four at once is structural, not heroic. A cap on the size of any one position limits overconfidence. A fixed rebalancing date sells some strength and buys some weakness whether you feel like it or not. A liquidity filter keeps you out of the thinnest stocks, where these biases do the most damage. Each constraint is small. Together they remove most of the behavioural risk.

The screen time tax

Every time you open the app you create a new reference point. Seen at ₹12 lakh in the morning and ₹11.95 lakh at lunch, a ₹5,000 dip feels like a real loss, even though it is noise on a long horizon. More looks means more red, more urges to act and more trades. Benartzi and Thaler's work on myopic loss aversion shows how frequent evaluation makes people take less equity risk than their own goals require.

Checking less often costs nothing and changes behaviour more than any amount of reading. Monthly is plenty for a long term portfolio. Try one honest count: how many times did you open your portfolio last week, and how many of those looks improved a decision?

Common mistakes

• Counting only the wins. If you do not write down your calls before the event, memory will grade them generously.
• Using the 52 week high as a value guide. Price distance from a high says nothing about earnings.
• Extrapolating the last year. Check the long run base rate and the full range of drawdowns before projecting anything.
• Trusting the vivid story. Ask how many people tried the same thing and how many succeeded.

Questions readers ask