Mathematical Position Sizing Strategies

Two traders take the same NIFTY trade on the same morning. Same entry, same stop, same exit. One ends the week flat and takes the next signal without flinching. The other is down a third of the account and has stopped taking signals altogether. Nothing separated them except the number of lots each one carried. Position sizing is the step between having an idea and having a position, and it is the only part of the sequence that is entirely under the trader’s control. The market decides whether the idea works. The trader decides, in advance, what it costs when it does not.

Three numbers do all the work. Lot size is the fixed quantity in one contract — 75 units for NIFTY at the time of writing, revised by the exchange from time to time. Contract value is lot size multiplied by the index level, so a NIFTY lot with the index near 24,500 controls roughly ₹18.4 lakh of underlying. Stop distance is the number of points between the entry and the level that says the idea was wrong. Multiply stop distance by lot size and you have the rupee risk of one lot. Everything else in this article is arithmetic on those three numbers.

Position sizing is not the same thing as margin, and confusing the two is the most common error in the F&O segment. Margin tells you the largest position your broker will allow you to carry. Sizing tells you the largest position your account can survive being wrong about. These are different numbers, they are not close to each other, and the smaller of the two governs. A broker screen that says you can carry four lots is describing collateral, not risk. The margin rules module covers how SPAN and exposure margin are computed; this module covers the number that should override them.

It is worth being clear about what sizing cannot do. It does not make a poor idea profitable. A trade entered at the wrong level, sized perfectly, still loses. Sizing has no influence on whether a trade wins — it only decides how much of the account is standing on that single outcome. That is precisely why it matters. Entry quality varies with skill, market regime and luck. Size is a decision made with a calculator before the order is placed, and it works exactly as intended every single time.

Fixed-fractional sizing means risking a constant fraction of current account equity on every position. Not a constant number of lots, and not a constant rupee amount — a constant percentage of whatever the account is worth on the day the trade is taken. If the account grows, the rupee risk grows with it. If the account is in a drawdown, the rupee risk shrinks automatically, without the trader having to make a discretionary decision at the worst possible emotional moment. That automatic property is the entire reason the method has survived.

The formula converts a percentage into a whole number of lots. Take capital, multiply by the chosen risk fraction to get a rupee budget for this trade. Separately, measure how far the price has to travel against you before the idea is dead, and multiply that distance by the lot size to get the rupee cost of being wrong on one lot. Divide the budget by that cost. The quotient is how many lots the account can carry. It is almost never a whole number, and what you do with the remainder is the subject of a later section.

The risk fraction itself is a policy choice, not a formula output. A smaller fraction means the account survives a longer run of losses but compounds more slowly when the ideas work. A larger fraction does the reverse. What matters far more than the exact figure is that it is fixed before the trade and applied identically to the trade you feel certain about and the trade you feel lukewarm about. Confidence is not an input to this formula, because confidence is not measurable and is systematically highest just before it is wrong.

Suppose NIFTY is near 24,500 and a trader sees a level at 24,350 below which the setup no longer holds. The stop distance is 150 points. With a lot size of 75, one lot costs 150 × 75 = ₹11,250 if the stop is hit. That single number, ₹11,250, is what the risk budget has to be able to absorb. Nothing about the trader’s view, the target, or the strength of the pattern changes it. The only way to reduce it is to place the stop closer, which changes the idea, or to trade a smaller instrument, which changes the market.

Now run several account sizes through the same trade. On an account of ₹5 lakh with a 1% budget, the trader is allowed to lose ₹5,000. Dividing ₹5,000 by ₹11,250 gives 0.44 lots. Lots are integers. Rounded down, 0.44 becomes zero, and the correct action is to not take this trade. Raising the budget to 2% gives ₹10,000, which produces 0.89 lots — still zero after rounding. The formula is not being difficult. It is reporting, accurately, that this particular stop distance is too expensive for this particular account.

The capital at which a 1% budget exactly funds one lot of this trade is ₹11.25 lakh, because ₹11,250 is 1% of it. That figure is a property of a 150-point stop on a 75-unit lot, not a threshold at which someone becomes ready to trade derivatives. Widen the stop to 250 points and the same arithmetic demands ₹18.75 lakh. Tighten it to 80 points and it falls to ₹6 lakh, at the cost of a stop that market noise will now hit routinely. The number moves with the trade, which is exactly why it has to be recomputed for every trade rather than fixed once.

One more check sits after the formula. The lots the risk calculation permits may still exceed the margin the account can fund, especially under the post-2025 regime where a larger minimum contract value means each lot blocks more collateral. Compute both, then take the smaller. A trader whose risk formula allows two lots but whose margin only funds one carries one. A trader whose margin funds four but whose risk formula allows one carries one. The binding constraint is whichever number is lower, always.

This is the collision at the centre of retail F&O in India, and it deserves to be stated plainly rather than worked around. Lots are integers. Since the 2025 revisions raised the minimum contract value the exchanges must maintain, a single index lot now controls a far larger notional than it did before, and lot sizes were reset upward to match. The consequence is arithmetic, not opinion: on a modest account, a disciplined risk budget frequently cannot fund even one lot of a normal-width stop. The formula returns a fraction, the fraction rounds to zero, and there is no position.

What happens next is where accounts are lost. The reflex is to round 0.44 up to one lot, because a fraction of a lot cannot be bought and passing feels like wasted effort. Rounding up is not a small adjustment. On the ₹5 lakh account above, one lot risks ₹11,250 against a ₹5,000 budget — the trade has silently become a 2.25% risk. Do that on four positions at once and the account is carrying 9% of equity on open risk, having never once decided to do so. Every step of that was defensible in isolation, which is precisely what makes it dangerous.

The legitimate responses are narrow, and each one has a cost that has to be accepted openly. A defined-risk structure such as a debit or credit spread lowers the rupee risk of a single unit, because the maximum loss is the strike width minus the net credit rather than the full move — this is covered in the directional spreads and iron condor modules, and it caps the loss at a smaller number, not at a safe one. A closer stop lowers risk per lot but raises the chance of being stopped out by noise before the idea has resolved. Waiting until the account funds the position at the intended fraction is the third option, and it is the only one that changes nothing about the trade itself.

What is not a legitimate response is treating the stop as adjustable to fit the size. Moving the stop from 150 points to 60 points because 60 points makes the lot affordable does not reduce risk. It relocates the exit to a level that has nothing to do with the idea, which converts a considered trade into a coin toss with better paperwork. The stop belongs where the analysis says the idea failed. The size is what adapts. This ordering is not negotiable, and the risk-per-trade module treats it in detail.

A trader who always carries two lots is using fixed-lot sizing. It is simple, it requires no calculation, and it fails in one specific and predictable way. The rupee risk of two lots does not change when the account shrinks, so the percentage of equity being risked climbs with every loss. The method takes progressively larger bets exactly as the evidence that something is wrong accumulates. Fixed-fractional sizing does the opposite, mechanically, without the trader needing to notice or agree.

Take an illustrative account of ₹10 lakh and a run of losing trades — a run, not a disaster, because losing runs are an ordinary feature of any approach that is not being described dishonestly. Under fixed-fractional sizing at 2% of current equity, each loss removes 2% of what is left, so the equity path is a decaying curve that never reaches zero. Under fixed two-lot sizing at ₹22,500 of risk per event, each loss removes the same ₹22,500 from a shrinking base, so the path is a straight line heading somewhere definite.

The table below runs both. After twenty losses the difference in equity is meaningful, but the more important column is the last row. The fixed-fractional trader is about to risk ₹13,352 — still exactly 2% of the account, still a decision the account can absorb twenty more times. The fixed-lot trader is about to risk ₹22,500, which is now 4.1% of what remains, and which will be 6% and then 9% if the run continues. Nobody chose that escalation. It is the arithmetic of dividing a constant by a falling number.

The recovery mathematics compound the problem. A 20% drawdown requires a 25% gain to return to the starting point. A 45% drawdown requires close to 82%. Because the percentage needed to recover rises faster than the percentage lost, every method that lets the risk fraction drift upward during a drawdown is pushing the account toward a hole it cannot climb out of by trading the same way it fell in. Fixed-fractional sizing does not prevent drawdowns. It bounds how deep an uninterrupted run of them can go.

For futures and for long options, the sizing formula has a clean denominator. A long option’s worst case is known the moment it is bought: the premium is paid upfront and it cannot be lost twice. A NIFTY call bought at ₹120 with a lot size of 75 has a maximum loss of ₹9,000 per lot, whatever the index does overnight, whatever happens to volatility, whatever the news is. That figure goes straight into the formula as risk per lot, and the sizing is honest.

A naked short option has no such number. The premium collected is the maximum gain, and the loss has no arithmetic ceiling — it is bounded only by how far the underlying travels. A stop order does not supply the missing ceiling. Between the close and the next open there is no trading, so a gap prints the loss before any order can act on it, and the black swan module works through exactly what that does to the account. Sizing a short option therefore requires the trader to invent the denominator, by assuming an adverse move and sizing to that assumption rather than to a stop.

A practical way to do this is to size the short as though the underlying gaps against it by a defined percentage overnight, and then ask whether the resulting loss is one the account can absorb. If a 6% adverse gap on the index would produce a loss larger than the account is prepared to take, the position is too large — irrespective of how comfortable the margin looks, and irrespective of the fact that the gap has not happened. Margin is set by the exchange to cover ordinary volatility. It is not a risk budget and was never designed to be one.

This is why defined-risk structures are so much easier to size correctly. On a spread, the maximum loss is the strike width minus the net credit, multiplied by the lot size, and that number is fixed at the moment of entry. It goes into the formula exactly like a long option’s premium. The wing costs part of the credit, so the same position collects less. That reduction is the price of having a denominator at all, and it is the cheapest thing on the screen.

The Kelly criterion answers a different question from the fixed-fractional rule. Fixed-fractional asks how much of the account a single trade may cost. Kelly asks what fraction of capital maximises the long-run growth rate of the account, given a known probability of winning and a known ratio of average win to average loss. It is a real result from information theory, and the fraction it produces is the mathematically optimal one — provided the two inputs are correct.

Take the formula purely as arithmetic on assumed inputs. If a hypothetical process wins with probability 0.50 and its average win is twice its average loss, then f = 0.50 − (0.50 ÷ 2) = 0.25, or 25% of capital. That output should immediately make a derivatives trader uncomfortable, and the discomfort is the correct reaction. Kelly is telling the truth about a game whose odds are known exactly. It says nothing useful about a game whose odds are estimated from a few hundred past trades in market conditions that have since changed.

The practical critique has three parts. The win probability is not observable, only estimated, and the estimate is noisiest exactly when the sample is small. The win-to-loss ratio is not stable across volatility regimes, so a figure measured in a quiet market misstates a violent one. And the penalty for overestimating the edge is not symmetric: sizing above the true Kelly fraction reduces growth and increases drawdown at the same time, while sizing below it only reduces growth. The asymmetry argues for erring low, deliberately and by a lot.

The reasonable way to hold Kelly is as a ceiling rather than a target — a number the size must stay well below, useful mainly because it demonstrates that an optimal fraction exists and that exceeding it is destructive even when the edge is real. The fixed-fractional rule is what actually gets used, because it needs no estimate of an edge, cannot be corrupted by a flattering sample, and produces the same discipline on the trade that feels certain as on the one that does not.

Some conditions justify carrying less than the formula permits, and they are all knowable in advance. The first is volatility. When India VIX is elevated, the same setup needs a wider stop to survive ordinary noise, which mechanically reduces the lots the budget funds. A trader who keeps the stop distance fixed while volatility rises is not holding risk constant — they are quietly increasing the chance of being stopped out by movement that means nothing. The implied volatility and IV rank modules cover how to read that environment.

The second is event risk. Around a scheduled RBI policy, a results announcement on a stock in the F&O list, or the single weekly index expiry that the post-2025 regime permits per exchange, the distribution of the next move is wider than usual and known to be wider. Carrying full size into a scheduled event is a decision to accept a larger risk than the rule permits. The honest way to hold a view through an event is to hold it in a smaller size, or in a structure whose maximum loss is defined before the event happens.

The third is correlation across open positions. A 1% risk on four separate trades is only 4% of risk if the four are independent. Four long positions in banking names, or a long index future alongside three long index calls, are one position wearing four names. In a sharp sell-off, correlations across Indian equities converge and everything moves together. Aggregate risk should be measured on what the positions actually do together, not on the count of tickers, and the risk-per-trade module develops this as portfolio heat.

The fourth is the account’s own recent behaviour. A predefined throttle — a rule that halves the risk fraction once equity is a set distance below its high-water mark, and restores it only on a set condition — removes a decision from a moment when judgement is least reliable. It costs something in recovery speed when the drawdown was just variance. It is worth paying, because a trader cannot tell from inside the drawdown whether the cause was variance or a regime that has stopped rewarding the approach.