Concentration Limits

A portfolio can pass every volatility and Value-at-Risk check and still blow up, if the reason it looks fine on average is that one position dominates everything else and that position gaps down. Diversification math assumes positions are genuinely separate bets; concentration limits exist because in practice, a firm's biggest position often becomes correlated with the rest of the book in exactly the moments that matter — a large counterparty defaulting, a sector-wide selloff, a single stock's earnings disaster.

Where the caps get set

Limits are typically layered: a single-name cap (no position above, say, 5% of NAV), a sector cap (no sector above 25%), and a counterparty cap on derivatives exposure. A common single-name formulation ties the cap to how easily the position could be exited:

$$w_{max} = \min\left(w_{policy},\ \frac{ADV \times k}{NAV}\right)$$

In words: the maximum allowed weight is the smaller of a flat policy limit and a liquidity-based limit — some multiple $k$ of average daily trading volume, scaled by the fund's size, so a position never grows larger than the fund could unwind over a reasonable number of days even in normal markets.

Worked example

A $200 million long-short fund has a 5% single-name limit, so no position can exceed $10 million. A biotech stock the fund holds rallies 150% on trial data over six months, growing the position from $4 million to $10 million organically — right at the cap. The average daily volume for the stock is $3 million, and the fund's liquidity policy caps any position at 10% of ADV, or $300,000, times a 10-day unwind horizon, giving a liquidity-based cap of $3 million. Even though the position is within the flat 5% policy limit, it is nearly 3.3x the liquidity-based cap — the risk desk would flag it for trimming even though no rule has technically been breached on the capital side.

What this means in practice

Concentration limits are one of the few risk controls that don't rely on a statistical model of correlation or volatility, which is exactly their value: they cap the worst case directly rather than trusting a model's estimate of how bad the worst case could be. Desks routinely see their best-performing position become their largest, purely from price appreciation, and forced trims at the cap are a standard, unglamorous part of running a risk-managed book.