Abstract. A portfolio is a set of claims. The manager attributes the returns to his own judgement; the risk seat has to establish what the book actually owns. Two disciplines follow, both run nightly by the automated pipeline behind this site on its own strategies: a dissection of intended against unintended exposure across six ordered layers, and a drawdown protocol written before the loss.

Keywords: risk decomposition, factor exposure, intended versus unintended risk, risk contribution, drawdown protocol, de-grossing, liquidity.

1. Intended and unintended exposure

The question is not whether the book is risky, because risk is the mandate. It is which exposures were chosen and which arrived uninvited. A fundamental stock picker is paid for idiosyncratic judgement, and everything else in the book is either deliberate ballast or a stowaway. The six layers below are taken cheapest explanation first, decomposed by risk contribution rather than dollar weight, and run rolling rather than quarterly, because weights drift and passivity does not keep a book neutral to factors.

The six layers. Ordered dissection of the book.
LayerWhat the seat measures
MarketNet and gross beta, the sensitivity to the overall market. The oldest stowaway is a “hedged” book with high-beta longs and low-beta shorts, net long the market while its net dollar exposure reads zero.
StylesSize, value, momentum, quality, low volatility, reversal over the short term. Loadings, the estimated sensitivity to each style, checked against what the manager believes he runs.
Industry and sectorConcentrations, and the view relative to peers. A “diversified” book of one sector’s suppliers is a single bet expressed through many names.
Macro sensitivitiesRates, oil, the dollar, whatever the names transmit. Duration, sensitivity to interest rates, is not only a bond concept.
Crowding and liquidityOwnership overlap with other funds, and the days to liquidate at a reasonable participation rate, a modest share of each day’s trading volume. The position that cannot be exited determines the worst week.
Residual idiosyncratic shareWhat fraction of risk is genuinely idiosyncratic once the layers above are removed. A stock picker whose variance is mostly factor is a closet factor fund charging fees for stock picking.

Two adjustments come from the pipeline’s own workshop. Dollar weights misstate where risk lives, so the blame counter ranks positions by risk contribution, whose pieces sum exactly and whose ranking rarely matches the position sheet. Work on random matrices supplies the second: residuals themselves can conceal a common factor in a concentrated regime, so idiosyncratic is a label to verify rather than a property to assume.

2. The drawdown protocol

The worst time to design a policy is inside the loss. The protocol below fixes the decisions on a calm day so that the worst day only executes them. Mechanical de-grossing, cutting the size of the whole book with longs and shorts alike, is nerve exercised in advance rather than nerve failing. The deepest task of the risk seat is keeping a skilled manager solvent long enough for the skill to compound.

The protocol. Settled before the loss, executed inside it.
StepRule
Diagnose before actingDecompose the loss: factor or idiosyncratic, a few names or the whole book, inside the risk budget or breaching it. A manager losing on his thesis and a manager losing on exposures he did not know he held need opposite remedies.
Run the pre-committed ladderEarly-warning indicator, the manager adjusts; first threshold, escalation; second threshold, mandatory de-risking. Triggers set at promotion are never renegotiated mid-drawdown, and the pipeline applies the same rule to its own strategies.
Cut in the right orderThe de-grossing sequence is fixed in advance; see Table 1.
Separate loss from errorGood process loses money on bad days, which is what risk means. The review asks whether the process was followed, not whether the outcome was pleasant. Repeated breaches are structure rather than weather, and structure escalates.
Talk to the PMEstablish with the portfolio manager whether the thesis is intact or broken by new information, and tell the defence of a position from the defence of a person.
Table 1. De-grossing order in a drawdown.
PriorityInstrument cutRationale
FirstGross exposureLowers the dose while preserving the manager’s relative bets.
SecondUnintended factor exposuresHedge or trim the stowaways; they were never the thesis.
EarlyConcentrated and illiquid positionsThe exit that cannot be made later must be sized correctly while it still can be.
LastIdiosyncratic names of highest convictionThey are the reason the manager holds a seat.

These are methodological field notes from fifteen years on the risk side of that conversation. No dataset or backtest is reported here.

My pipeline runs both disciplines on itself: blame counters by risk contribution and liquidity breakdowns on every backtest, and pre-committed retirement triggers on everything promoted. It rejected the founder’s namesake model at those gates, four versions running. A protocol whose author exempts himself from it is a brochure.