The Average Stock Wasn't the Average Investor's Experience

2014 to 2026 was a bull run. The Nifty 50 quadrupled, from roughly 6,300 to roughly 24,000, an annualised 11.17%. Does that mean an investor who bought NSE-listed stocks in 2014 made 4x too?

Four things the data shows

We took every company listed on the NSE on 1 January 2014, 1,230 of them, and tracked each one to today, including the ones that no longer trade. Four conclusions follow.

1. A handful of companies, the top 5–6%, created most of the wealth. Weighted by size at purchase, the top 5% produced 82.9% of everything this cohort made. Weighted equally, the top 6% still made half of it. Only 24 in 100 companies individually beat the portfolio's own average return. The average investor's experience is nothing like the average return.

2. Inside those companies, most of the gain arrived in a small number of trading days. The median top-6% company reached 50% of its entire multi-year gain in about 12 trading days, and 80% in 22 days, out of thousands held. It was not spread evenly across twelve years.

3. Concentration builds with holding period, and it does not protect you in a crash. The median stock fell harder than the index in seven of the eight drawdowns since 2014. This is the mechanism behind #1: compounding turns a market where roughly as many stocks fall as rise into one where the eventual outcomes are heavily lopsided.

4. The downside was real. Roughly a quarter of companies no longer trade on the NSE, 16% lost more than half their value, and 41% (509 of 1,230) failed to beat a 9% fixed deposit over the full 12.6 years.

What this means for strategy

Diversify if you are passive, or have limited conviction in your own stock selection. It will not make you rich, but it keeps you close to the market's own outcome. Concentrate only if you have real conviction in the stock and the timing. That path returned 10x or more for 23% of the companies in this cohort.

Avoiding capital loss matters more than chasing gains. Around 300 companies in this cohort no longer trade on the NSE, and 16% lost over half their value. A loss like that needs a much bigger subsequent gain just to get back to even, so managing it means evaluating each stock before you buy it, the way an underwriter evaluates a loan, not after.

1. The average is carried by outliers

Only 24 in 100 companies individually beat the portfolio's own 19.57% annual return. Rank all 1,230 by rupee wealth created or destroyed, and weighted by size at purchase, the top 5% produced 82.9% of all wealth this cohort created. Weighted equally, the top 6% produced half of it. The bottom 43% net to nothing: their winners and losers cancel out.

Log scale, because on a linear one you'd see a spike and two visible companies. The median company and the Nifty 50 sit almost on top of each other. The typical listed company did about what the index did. The average sits far to the right of the median.

Twenty of the six in a hundred, by name.

Nothing here says these companies were identifiable in January 2014. This is the Indian echo of a result Hendrik Bessembinder published for the United States: the entire net wealth creation of the US market since 1926 traced to the best-performing 4% of companies.

2. The gain arrived in a handful of days

For the 74 companies in the top 6%, we broke each one's full holding-period return into daily moves and found how many of its own trading days it took to reach half, then most, of the total gain. Daily log returns are additive, so this is exact even for a 400x mover: sort each company's own daily log returns descending and find where the running total crosses 50% and 80% of the sum.

Median company: 50% of the entire multi-year gain in about 12 trading days, 80% in 22, out of a median 3,102 days held. Under 1% of the days a winner was held did most of the work. And this is not one company's quirk. The chart below shows the same curve for all 74 companies in the top 6% together: nearly all of them front-load most of their gain into a small cluster of days, then drift for years.

Our hypothesis is that these are earnings days, results and disclosure days, major policy news, and crucial investor calls: the days a company's information actually changes, rather than days its price drifts with sentiment. We have not tested that directly. It is the natural next study, not a claim this one makes. What it does say is narrower and already useful: the winners moved in bursts, not steadily, and being out of the stock on the wrong handful of days would have missed most of the return.

3. Time, not luck

The obvious objection: 2014–2026 is one long bull run, so of course stocks did well. We tested this by re-running the same portfolio inside every Nifty 50 drawdown deeper than 10% since 2014, dated from the index itself.

In seven of the eight, the median company fell further than the index. In the Covid drawdown, 2.3% of companies were positive and over a third lost half their value. Concentration did not protect you on the way down.

The reason is mechanical, and it is Bessembinder's own explanation, not a new one. A single period's return is roughly symmetric: about as many stocks rise on a given day as fall. But returns compound multiplicatively, and a loss is bounded (a company can fall at most 100%) while a gain is not (it can rise by a thousand percent and keep going). Compound a roughly symmetric distribution over enough periods, and the floor at −100% clips the left side while the open-ended right side keeps stretching. The outcome distribution becomes positively skewed purely as an arithmetic consequence of repeated compounding, before any story about which companies were "better" enters at all. That skew needs time to build. Over a six-month window the right tail has barely started stretching, so the concentration vanishes. Over twelve years it has had room to run. Concentration rises monotonically with holding period, from 41% of the gain sitting in the top decile at a one-year horizon to 62% at 12.6 years, which is the direct fingerprint of compounding doing this, not of any one period being lucky.

4. What individual stocks actually did

The portfolio made 19.57% a year. The individual company mostly did not.

Both extremes are true of the same market at the same time. The portfolio-level 19.57% describes neither of them.

A 24.7% chance of no longer trading on the NSE at all, and a 16.2% chance of losing over half your capital, is why downside screening matters more than upside forecasting: a stock that halves needs a 100% gain just to get back to even. The checks that separate the two groups are mostly the ones an underwriter runs before money goes out, not the ones an analyst runs after a stock has already moved:

Diversify or concentrate? The data behind takeaway one

Draw a random portfolio from the cohort, hold it equally weighted, and repeat twenty thousand times at each size.

One random company gives a 35.7% chance of compounding at 15% a year. A hundred random companies give 98.6%. Breadth does not make you rich. It makes you reliable, and it has a ceiling: the odds of beating the portfolio's own average company never clear 43.8%, at any size. Diversification converges on the mean. It cannot pull you above it.

Caveat

This entire study sits inside one secular bull market. The panel spans 12.7 years, and the deepest drawdown in it recovered in 300 days. It has not been tested against a genuine multi-year bear market or a full economic cycle, which in equity markets can run past 90 years peak to peak. None of the numbers above are a claim about the next cycle, only about this one.

This is research, not investment advice. Nothing here is a recommendation to buy or sell any security. Past returns of the companies named above are historical facts and carry no information about their future.

Data, code, and corrections

This study replicates and extends Hendrik Bessembinder's methodology from "Do Stocks Outperform Treasury Bills?" (Journal of Financial Economics, 2018), applied to the Indian market instead of the US one, with a fixed hurdle ladder in place of a matched T-bill series.

The panel is published in full: identity map, corporate-action ledger, cohort, computed results, figures, and the code as run. A finding is only checkable if its input is. It's on GitHub at github.com/dangilaksh23/nav_tools.

No industry breakdown appears in this article. The industry map only exists for companies pulled from current screener snapshots, so it is really a map of survivors: among the companies it does cover, 97.8% are still trading, versus 19.5% among the companies it misses. Any industry table built on it would be a survey of survivors wearing an industry costume. Corrections are the point: an error found in the data is more useful to us than agreement with the conclusions.