If risk is the chance of permanent loss, we can give it a working shape. For any position, imagine the set of things that could genuinely go wrong, each with a probability and each inflicting some permanent impairment on your capital. Then the risk of the position is captured by the whole distribution of impairment. A single useful summary of it is the expected impairment:
Expected impairment = Σ P(bad scenario i) × (permanent loss fraction in scenario i)
Worked example: expected impairment on a ₹100 position
This is deliberately parallel to the expected-value machinery from M6.03 §4.8, but pointed only at the left tail: the outcomes where capital is destroyed, not merely where the return is a bit lower than hoped. A worked feel for it: suppose a ₹100 position has a 78% chance of no permanent impairment (the thesis holds or you exit whole), a 15% chance the business deteriorates and you permanently lose 40%, and a 7% chance of a severe impairment (a leverage spiral or a fraud) that costs 85%. Expected impairment = 0.78×0 + 0.15×0.40 + 0.07×0.85 = 0 + 0.060 + 0.0595 = 0.1195 ≈ 12%. Notice that the low-probability, high-magnitude tail (7% × 85% ≈ 6.0 points) contributes as much as the more likely, milder one (15% × 40% = 6.0 points): magnitude and probability multiply, so a rare catastrophe and a common bruise can carry equal weight. This is why "it's only a 7% chance" is never, by itself, a reason to ignore a scenario. You must always ask "a 7% chance of what magnitude?"
Advertisement
Why can't you read risk from a track record?
Now the hard part, and the deepest idea in the definition. This entire quantity is a forward-looking estimate that you can never observe directly, and it is not visible in outcomes. This is Howard Marks's great theme (in The Most Important Thing and the memos "Risk" and "Risk Revisited"), and it is worth stating with care because it overturns how almost everyone intuitively judges risk:
- Risk exists only in the future, and the future is a distribution, not a fact. Before the event, many things could happen; afterward, exactly one did. The risk was the range of things that could have happened and the odds of the bad ones, most of which, by definition, did not happen. So the realized outcome shows you one draw from the distribution, never the distribution itself.
- Therefore you cannot read risk off a track record. A manager who earned 18% a year for five years by selling out-of-the-money puts, or by concentrating in one leveraged theme that simply hadn't broken yet, produced returns indistinguishable from a prudent compounder's, right up until the day the risk showed up. "Picking up pennies in front of a steamroller" generates a beautiful, low-volatility track record until the steamroller. Good outcomes routinely come from risky decisions that got lucky; bad outcomes come from sound decisions that didn't. The quality of a decision and the quality of its outcome are different axes (M6.03 §4.8), and risk lives on the decision axis, which the outcome never fully reveals.
- This is exactly why volatility is such a seductive substitute. Volatility has the one property risk lacks: it is visible and measurable. So the industry quietly swaps the thing it can measure for the thing it means, and then acts surprised when a "low-risk" (low-volatility) position, a AAA-rated subprime tranche in 2007 or a quietly-compounding Wirecard, detonates. The map (volatility) is not the territory (permanent-loss probability), and confusing them is not a technicality; it is the recurring cause of blow-ups.
The practical consequence is humbling and liberating at once. You cannot measure the risk of a position; you can only estimate its structure (what could impair it, roughly how likely, roughly how badly) and then, because your estimate is uncertain, build in protection you don't strictly need if you're right. That protection has two names you already know: margin of safety (buy well below conservative value, so the gap absorbs errors) and position sizing (never so much that being wrong is fatal). The rest of this module is those two tools, made precise. Risk management is not risk measurement; it is risk structuring under permanent ignorance of the true odds.
The taxonomy of permanent impairment: the four ways you actually lose
"Permanent loss" is an abstraction until you can name its causes, and there are essentially four. Learn them as a checklist, because every position you ever size is a bet that all four are either absent or paid for by the price. Notice that three of the four have nothing to do with the stock being volatile. They are quiet, structural, and often invisible right up to the impairment.
1. Business deterioration. The most fundamental: the company's economics decay permanently, as the moat erodes, demand secularly declines, technology routes around it, and returns on capital fall below the cost of capital and stay there. The value you bought genuinely shrinks; there is no recovery to wait for because the business that would produce it no longer exists. Kodak (film), Nokia (handsets), regional newspapers, most textile mills, video rental. The diagnostic is everything you learned in Phases 4–5: is the ROIC durable (M2.03, M4.04), is market share stable (M4.05), is the industry's profit pool growing or draining (M4.03)? A value trap, a statistically cheap stock that keeps getting cheaper, is business deterioration the buyer refused to see. Defense: buy quality (durable ROIC > WACC) and monitor the moat's direction, not just its existence.
2. Overpaying for a good business. This one is the most counter-intuitive and the most common way careful investors lose, so dwell on it. You can identify a genuinely wonderful company (wide moat, great management, long runway) and still suffer a permanent loss, because you paid a price that already embedded a decade of flawlessness. The Nifty Fifty of 1972 were, mostly, excellent businesses (Coca-Cola, Xerox, IBM); bought at 50–90× earnings, many took a decade or more just to return to the entry price in real terms. Cisco in March 2000 was a superb, real, profitable business at ~130× earnings; a quarter-century later the stock has still not durably reclaimed that peak, though the company grew the whole time. The business compounded; the investment was impaired at the moment of purchase by the price. This is the direct link to valuation (M3.05): a great business at a demanding multiple is a bet that reality will exceed an already-optimistic implied forecast, and when it merely meets it, you lose. Defense: the margin of safety, the discipline that a wonderful business is not a wonderful stock at any price.
3. Leverage and forced selling. Covered in risk vs volatility and revisited in the anti-ruin overlay: leverage, the company's or yours, converts survivable volatility into permanent loss by forcing a sale at the worst moment. A company with too much debt can be a fine business that hands the equity to creditors in a downturn it would otherwise have survived (the equity is permanently impaired even as the enterprise recovers under new owners). And your leverage (margin, redeemable capital, a near-term cash need) can force you out at the bottom of a temporary decline. This is the impairment that turns the other three from painful into fatal, because it removes your ability to wait. Defense: structural. No forced-selling leverage, ever; scrutinize the company's balance sheet (net debt/EBITDA, ALM for lenders, refinancing walls) as a survival question, not just a returns one.
4. Fraud and accounting deception. The value was never there; the reported earnings and assets were fiction. When it surfaces, the stock does not fall 40%. It goes toward zero, fast, with no recovery, because there is nothing underneath. Enron, WorldCom, Wirecard, Luckin Coffee, Valeant; in India, Satyam, IL&FS, DHFL, Manpasand, Vakrangee. This is the purest permanent loss and the one where volatility is least warning: fraudulent books are often unnaturally smooth right until exposure. Defense: the forensic toolkit (M2.06–M2.07: accruals, cash-vs-earnings divergence, related-party dealings, the Beneish M-Score, auditor and CFO turnover, in India the promoter pledge), plus the governance judgment of Phase 6, plus, decisively, sizing, because no diligence catches every fraud and the only reliable defense against the one you miss is that it was never large enough to ruin you.
How does sizing protect against permanent loss?
Sit with that last clause, because it is the bridge from analysis to sizing. Three of these four defenses are analytical; they live in the work you already know how to do. But none of them is perfect: moats you judged durable erode, prices you thought fair were rich, balance sheets you cleared hid a refinancing wall, frauds you screened for slipped through. The one defense that works even when the analysis fails is that the position was sized so that its total impairment could not end you. Analysis reduces the probability of impairment; sizing caps the magnitude of impairment to your portfolio.
You need both, and the second is the one you are here to learn, because it is the one that keeps you in the game long enough for the analysis to pay.