Everything above converges on one non-negotiable rule that sits on top of all the sizing math and can override it: no single position, and no correlated cluster of positions, may ever be sized so that being wrong about it threatens your survival. This is the anti-ruin overlay, and it is not caution. It is arithmetic, the direct consequence of ergodicity (M6.03), which is worth restating in sizing terms.
Why is ruin different from an ordinary loss?
Wealth compounds multiplicatively, so losses and gains are asymmetric: a 50% loss requires a 100% gain to recover; a 90% loss requires a 900% gain; a 100% loss requires an infinite gain, and you never come back. Ruin is an absorbing state: once you hit zero (or near it, or a level that forces you out of the game), no subsequent brilliance can restore you, because you have no capital left to compound and no seat at the table. The sequence matters, not just the average: a path that touches ruin stays ruined regardless of how favorable the odds were. This is why survival is a first-order goal, not a virtue to balance against returns but the precondition for having any returns at all. Buffett's two rules ("Rule No. 1: never lose money. Rule No. 2: never forget Rule No. 1") are not about avoiding every loss, since of course individual positions will lose; they are about never losing so much that you are removed from the game. Kelly, recall, already encodes this: by maximizing expected log-wealth it will never stake an amount that risks zero. The overlay is what you add because real life has fat tails, correlation, and estimation error that the clean Kelly bet does not.
Advertisement
What are the hard caps on position size?
Concretely, the overlay is a short list of hard rules that bind before any Kelly or conviction number:
- A single-name cap. No position above some fraction at cost (a concentrated-but-sane book might use ~10–12% at cost; more conservative, 5–6%). Even your highest-conviction, smallest-downside idea respects it, because you have been wrong about "certain" things before and will be again.
- A factor cap (the one that actually matters, see hidden correlation). No more than ~20–25% exposed to any single factor, counting the hidden-correlation clusters as one. This is the cap that stops the "thirty names, one bet" catastrophe, and it is the one most portfolios lack.
- No leverage that can force a sale. No margin, no borrowing against the portfolio, nothing that lets a temporary price decline trigger a permanent, forced liquidation. This is the bright line that keeps volatility from ever becoming permanent loss (see risk vs volatility). It is the single most important rule on the list, because leverage is what turns every other mistake from survivable into fatal.
- Cash is a legitimate position. Holding cash when nothing clears your hurdle is not a failure to be invested; it is optionality and the ultimate defense against forced selling (developed in M9.02). The anti-ruin overlay is why cash always has a role.
- Size smaller when the downside floor is unverified or the situation is uncertain (see sizing under uncertainty). If the entire justification for a large size is a capped downside
d, and you cannot prove that floor is real (genuine asset backing, no hidden leverage, no fraud), then the floor is a hope, and you size as if the downside were large, because it might be.
How do you test a position for ruin?
The mental test to run on every position, and on the whole portfolio, is the pre-mortem sized for ruin: if this position, and every position correlated with it, went to its realistic worst case at the same time, would I survive to keep investing? If the answer is anything but a clear yes, the size is too big. Cut it until the answer is yes, before you buy, and let that constraint, not your enthusiasm, set the final number. Survival first; returns are what survival earns you over time.
This is the note Phase 9 opens on and the discipline the whole program has been quietly building toward: you have spent sixty-four weeks learning to be right, and this is where you learn to make sure that being occasionally, inevitably wrong can never take you out of the game.