A concentration limit sits after the weighting scheme and before the orders. The scheme proposes target weights, the constraint clips whatever breaches a bound, and the excess is redistributed across the remaining positions.
Index providers use these routinely. NSE’s Nifty 50 is free-float market-cap weighted, but a single stock’s weight is capped at 10 percent at rebalance, which stops the largest constituent from mechanically taking over the index as it grows. The factor indices apply the same idea to a factor score, so the highest-scoring stock in the Nifty 200 Momentum 30 cannot dominate the basket simply by scoring best.
Types of limit
| Constraint | What it caps | Typical use |
|---|---|---|
| Maximum position weight | Any single holding | 5% to 10% for a diversified book |
| Maximum relative weight | A multiple of a reference field, such as free-float market cap | Keeps small caps from taking large weights |
| Minimum position weight | The floor below which a name is dropped | Avoids 0.3% positions that only add cost |
| Maximum group weight | Total in a sector, industry or country | Caps sector risk without changing scores |
Maximum relative weight is the least obvious and often the most useful. Setting it at three times a stock’s free-float market-cap weight means a stock that would carry 0.4 percent in a cap-weighted book can reach 1.2 percent and no further, which lets a score-driven strategy express conviction while still respecting liquidity.
Illustrative arithmetic
A score-proportional scheme proposes 22 percent in the top name across a 15-stock portfolio. With a 10 percent cap, 12 percentage points are cut and spread over the other 14 holdings, lifting each by roughly 0.86 points. The strategy still holds its highest-conviction name at double the equal-weight share, and a single earnings shock in it now costs the portfolio a tenth of the position, not a fifth.
What limits do and do not fix
Capping single names does not cap correlated risk. Ten positions at 8 percent each, all in private-sector banks, is a concentrated portfolio that passes every single-name test. Group limits are the tool for that, and they need a sector definition you trust.
A binding constraint also costs you signal. If your cap is hit at every rebalance, the weighting scheme is no longer expressing what it computed, and you are running a different strategy from the one you tested. Check how often each constraint binds during the backtest. Occasionally is a safety net; every period is a redesign.
Redistribution has a direction, too. Excess weight from a capped name goes somewhere, usually to lower-scored holdings, which dilutes the signal slightly. That trade is normally worth making, and it should be a decision rather than a side effect.