INVESTSIGHT CAPITAL is a fintech and capital markets firm in Bengaluru, India. This article is part of our research library — for the official site see investsightcapital.com, or ask us a question.
Investors spend most of their effort on selection and almost none on size. Yet a correct view held in the wrong size produces a poor result, and a merely reasonable view held in a sensible size produces a durable one.
Size is where a portfolio’s actual risk is decided.
What size is really expressing
A position size is a statement about two things at once: how confident you are, and how much you can afford to be wrong.
Most investors size on the first alone. Conviction, though, is the least reliable input available — it is highest exactly when a story is most compelling, which is not the same as when it is most likely to be correct.
The second input is the one that survives contact with reality. Not “how sure am I?” but “if this is wrong, what does the portfolio look like afterwards?”
A position sizing framework is simply a way of forcing that second question to be answered in numbers, before the money moves. The methods below differ in what they measure, but they share one purpose: they replace a feeling with a rule.
Three questions before sizing
What does being wrong cost? Not the probability of loss — the consequence of it. A position that could impair the plan deserves a different size from one that would be an annoyance.
What else moves with it? Three positions that fall together in the same conditions are one position wearing three names. Correlation is where diversification quietly fails.
Would I hold this at a loss? If the honest answer is no, the position is too large. Size that cannot be held through a drawdown will be abandoned at the worst possible moment.
How much of a portfolio should one position be?
There is no universal number, but there is a useful reference point. For nearly three decades SEBI’s mutual fund regulations barred a diversified scheme from holding more than 10 per cent of its net assets in the shares of any one company, with exceptions only for index and sector funds (SEBI, Mutual Funds Regulations, 1996, Seventh Schedule, clause 10). The 2026 rewrite of those regulations moves such limits into prudential norms SEBI sets separately, but the principle stands: a professional manager running other people’s money is not trusted with unlimited concentration, however strong the conviction.
For an individual, the ceiling should usually sit below what a regulator allows a fund, because an individual has fewer positions, less information and no compliance desk. A common working range for a single listed stock is somewhere between 2 and 10 per cent of the portfolio, with the upper end reserved for the most liquid, best-understood holdings. Where a position sits inside that range should come from a method, not from how exciting the idea is.
Fixed-fractional sizing
The simplest named method is fixed-fractional sizing: decide in advance the maximum share of the portfolio you are willing to lose on any one idea, then work backwards to the position size.
For illustration, take a portfolio of ₹50 lakh and a rule that no single idea may cost more than 1 per cent of the portfolio, which is ₹50,000. Suppose a stock is bought at ₹1,000 and the investor decides, at entry, that a fall to ₹800 would mean the thesis is wrong. That is a 20 per cent adverse move. The position that keeps the loss at ₹50,000 is ₹50,000 divided by 0.20, which is ₹2,50,000 — five per cent of the portfolio.
Notice what the method does. It ties size to the distance between the entry and the point of being wrong. A thesis that would be invalidated by a 10 per cent move could, on the same rule, be held at ten per cent of the portfolio. That is why fixed-fractional sizing needs a hard cap alongside it: the arithmetic will happily produce a very large position for a very tight exit, and tight exits are the ones most likely to be hit by noise.
Volatility-scaled sizing
Fixed-fractional sizing asks how far the price must fall before you are wrong. Volatility-scaled sizing asks how far the price tends to move anyway, and sizes so that every position contributes a similar amount of risk.
For illustration, suppose an investor wants each holding to contribute roughly 2 per cent of annualised volatility to the portfolio. A large, stable company whose shares have historically moved about 20 per cent a year would be held at 2 divided by 20, or 10 per cent of the portfolio. A smaller company moving about 40 per cent a year would be held at 2 divided by 40, or 5 per cent. The two positions are different in rupee terms but similar in how much they can move the whole.
The strength of this method is that it stops the most volatile holdings from quietly dominating the portfolio’s risk. Its weakness is that volatility is estimated from the past, and the past is least reliable at exactly the moments that matter. Volatility is lowest at the end of calm periods, which is when this method would size positions largest. It works best with a floor and a cap, and with the volatility figure refreshed rather than set once.
Why the Kelly criterion is usually halved
The Kelly criterion comes from a 1956 paper by John L. Kelly at Bell Labs, which showed that a bettor with a known edge maximises the long-run growth rate of capital by betting a specific fraction of it each time (Kelly, 1956). For a simple even-money bet, the fraction is the probability of winning minus the probability of losing.
For illustration only: if an investor believed a position had a 55 per cent chance of working and would gain or lose the same amount either way, the Kelly fraction would be 0.55 minus 0.45, or 10 per cent of capital. Half Kelly would be 5 per cent.
The reason practitioners halve it is that the inputs are guesses. Edward Thorp, who applied Kelly in both casinos and markets, showed that betting half the Kelly fraction keeps three-quarters of the growth rate while sharply reducing the chance of a large drawdown: the probability of doubling capital before halving it rises from two-thirds at full Kelly to eight-ninths at half Kelly. He also noted that overbetting is far more harmful than underbetting, so a fractional Kelly is prudent whenever the estimate of the edge is uncertain (Thorp, 2006).
In listed equities the estimate is always uncertain. In the example above, if the true probability were 50 per cent rather than 55, the correct Kelly fraction would be zero, and the “half Kelly” position of 5 per cent would already be too large. That is the practical lesson: Kelly is a ceiling to stay well below, not a target to aim at.
Sizing for the SIP investor
Position sizing is not only a stock-picker’s problem. Indian investors put ₹31,961 crore into mutual funds through SIPs in July 2026 alone (AMFI, 2026), and most of that money is sized by habit rather than by design: a monthly amount chosen once, spread across whatever funds were opened over the years.
The same three questions apply. A portfolio of six funds that all hold the same large-cap names is one position wearing six names. A sector or thematic fund is exempt from the single-company cap that applies to diversified schemes, so it can be far more concentrated than its label suggests. And a SIP into a fund the investor would stop at the first 20 per cent drawdown is, by the third question above, too large from the start.
A workable rule is to size fund exposure by what it is actually exposed to — which index, which sectors, which factor — rather than by fund name. Two funds tracking the same benchmark are one allocation. That is the point where asset allocation and position sizing meet.
Sizing as the real risk control
Stop-losses and hedges are reactive; they operate after conditions deteriorate. Size is decided in advance, under no pressure, and it caps the damage before the first adverse move.
Each of the methods above produces a number, and the numbers will disagree. Fixed-fractional sizing rewards tight invalidation points, volatility scaling rewards calm histories, and Kelly rewards confident estimates. Using more than one, and taking the smallest answer, is a reasonable way of letting each method’s blind spot be covered by another.
That is why in our process sizing is not an afterthought to the research. It is the point at which research becomes a portfolio — and the point at which rebalancing later has something defined to return to.
This article is part of our guide, How to Build a Resilient Portfolio.
