Portfolio
An equal-weighted portfolio of ten managers is not a balanced portfolio. If two of those managers run at three times the volatility of the others, they will drive the majority of the portfolio's result while occupying a fifth of the capital. Capital weights are an accounting convenience. Risk weights are the actual decision.
Starting from the risk budget
We begin by deciding how much of the portfolio's total risk each strategy should be responsible for, based on our conviction in the source of return, the capacity available, and how the strategy interacts with everything else we hold. Capital allocations are then derived from that budget using each manager's expected volatility, rather than the other way round.
This produces allocations that look strange to anyone reading a capital-weighted schedule. A low-volatility relative-value manager may receive several times the capital of a concentrated equity manager while contributing the same risk. That is the intended result.
Correlation is a moving target
Historical correlation matrices are the weakest input in the process and we treat them as such. Correlations computed in calm markets systematically understate what happens in stress, when leverage unwinds force unrelated strategies into the same direction. We therefore supplement statistical measures with a structural question: if this specific event occurs, which of our managers are on the same side of it?
That question is answered through scenario work rather than regression. We maintain a set of scenarios — a funding squeeze, a sharp rates repricing, a liquidity gap in credit, a crowded-trade unwind — and estimate each manager's exposure to them from position-level data where we have it and from conversation where we do not.
Hidden common factors
The most dangerous exposures are the ones nobody has labelled. Several managers may be independently short volatility, independently long illiquidity, or independently dependent on the same financing counterparty, without any of them describing their strategy that way. We look specifically for shared dependence on leverage availability, on continued market liquidity, and on a small number of prime brokers.
Diversification fails at the moment you need it most, unless it was built on structurally different return drivers rather than on historical correlation.
Rebalancing and drift
Left alone, a portfolio drifts toward whatever has recently performed — which is to say, toward whatever is most expensive and most crowded. We rebalance to the risk budget on a defined schedule, with tolerance bands wide enough to avoid trading for its own sake and narrow enough that drift does not become a decision by default.
Rebalancing constraints are real: lock-ups, notice periods and gates mean we cannot always trade to target. This is one reason liquidity terms are part of the selection decision rather than an afterthought. A portfolio you cannot rebalance is a portfolio whose risk allocation was set years ago by accident.
Position sizing as protection
Finally, sizing carries the weight of everything our diligence might have missed. No single manager is sized such that a total loss — fraud, blow-up, catastrophic style drift — would threaten the portfolio. We would rather give up some return from our highest-conviction position than rely on having been right about it.