Selection tells you what to own. Position sizing tells you how much. Two strategies can pick the exact same 30 stocks on the same day and produce very different return series, because one put 20 percent into its top name and the other spread 3.3 percent across all of them.
In a systematic strategy, sizing is a rule rather than a judgement call. It has three parts: a weighting scheme that converts your selected list into target weights, a set of constraints that caps what any single position or group can become, and a rebalance schedule that decides when actual weights are pulled back toward targets.
Weighting schemes
| Scheme | Sizing driver | Suits |
|---|---|---|
| Equal | 1 / N | No view on relative attractiveness within the selection |
| Fixed | Weights you specify | A deliberate, hand-set allocation |
| Market cap | Free-float size | Benchmark-relative strategies, large capital |
| Inverse volatility | 1 / σ | Equalising standalone risk per holding |
| Minimum variance | Covariance optimiser | Lowest forecast portfolio volatility |
| Risk parity | Covariance optimiser | Equal risk contribution per holding |
| Hierarchical risk parity | Clustered covariance | Risk parity without matrix inversion |
| Score proportional | Raw factor score | Letting conviction scale the position |
| Normalised score | Rescaled score | Score weighting with bounded extremes |
| Rank proportional | Rank order | Score weighting that ignores score magnitude |
Sizing in rupees
Target weights become order quantities only after capital and lot constraints are applied. On a ₹10 lakh portfolio a 4 percent target is ₹40,000; in a stock trading at ₹2,850 that is 14 shares, and the rounding leaves you at 3.99 percent rather than 4.00 percent. On smaller accounts this rounding drag is meaningful, and a 50-stock equal-weight portfolio on ₹2 lakh means ₹4,000 per position, which cannot buy a single share of several Nifty 50 constituents.
Failure modes worth testing
Score-proportional sizing is the most common trap. If your factor score is unbounded, one stock with an extreme score can take a third of the book, and the strategy silently becomes a single-name bet. Cap it or use rank-proportional sizing instead.
Volatility-based sizing has an opposite failure. It hands large weights to whatever was recently quiet, and measured volatility is artificially low for illiquid stocks whose prices barely print.
Rebalance frequency compounds all of this. Weekly rebalancing on a 50-stock portfolio can generate several hundred percent annual turnover, and at Indian retail costs, including brokerage, STT, exchange charges and impact cost, that is a real drag that no sizing scheme recovers. Model the costs in the backtest before deciding a sizing rule works.