Attributing a Drawdown to Positions

A portfolio's max drawdown — the largest peak-to-trough decline — is usually reported as a single number: "the strategy drew down 12% in March." That number says nothing about why. Was the loss concentrated in three positions that all moved together, or spread thinly across the whole book? Was it driven by a single risk factor everyone happened to be exposed to, or genuinely idiosyncratic bad luck across unrelated names? Drawdown attribution answers this by decomposing the loss into contributions from individual positions (or factors, or sectors), turning a single scary headline number into a diagnosis of what actually went wrong.

Why attribution matters more than the headline number

Two drawdowns of the same size can have completely different implications. A 12% drawdown driven by one position going against the strategy due to an idiosyncratic, company-specific event (a surprise regulatory action, say) is a story about diversification and position sizing — should that one position have been smaller. A 12% drawdown where every position lost money simultaneously because they all shared exposure to the same underlying risk factor (say, a rates move) is a story about hidden factor concentration — the strategy looked diversified across names but was never actually diversified across the risk that mattered. Attribution is what distinguishes these two very different lessons, which the headline drawdown number alone cannot.

How the decomposition works

For a drawdown over a period, each position's contribution to the total loss is roughly its position weight times its own return over that period, summed across positions to reconcile back to the portfolio's total loss:

$$\Delta V_{portfolio} \approx \sum_i w_i \, r_i ,$$

where $w_i$ is position $i$'s portfolio weight and $r_i$ is its return over the drawdown window. In plain English: each position's dollar contribution to the drawdown is (how big a bet it was) times (how much it moved), and adding up every position's contribution should reconstruct the total loss — ranking those contributions from most negative to least immediately shows which handful of positions did most of the damage.

Worked example: a 12% drawdown, decomposed

A ten-position portfolio draws down 12% over a month. Attributing contribution by position:

Position	Weight	Return	Contribution
A	15%	−40%	−6.0%
B	10%	−25%	−2.5%
C	10%	−15%	−1.5%
Remaining 7 positions (combined)	65%	avg −3%	−2.0%

Positions A and B alone account for −8.5 of the −12.0 percentage points of drawdown — roughly 71% of the entire loss came from two positions out of ten. That's a meaningfully different finding than "the portfolio had a bad month broadly": it points directly at whether A and B were oversized relative to their risk, or whether they shared a common driver worth investigating, rather than prompting a review of the whole book equally.

What this means in practice

Every material drawdown deserves an attribution breakdown before any conclusions are drawn about position sizing, risk limits, or strategy validity. A drawdown that's concentrated in a small number of positions calls for reviewing those specific bets and the sizing process that allowed them to get that large; a drawdown spread evenly across the book, or traced to a common factor exposure, calls for a different fix entirely — reviewing correlation assumptions or factor hedges rather than any single position.