Position Sizing for Irreversible Decisions

The Kelly criterion says the optimal fraction of your bankroll to bet is edge over odds, and it is right in exactly one world: the world where you get to run the tape an infinite number of times. Full Kelly maximizes long-run growth and, as a free bonus, maximizes the probability that you personally will tap out on the way there. The drawdowns are savage. Half Kelly gives up a quarter of the growth for half the pain, and it is the best trade available in finance: the discount is bought with nothing but restraint.

Now look at the decisions that actually matter — the move, the resignation, the thing said out loud that cannot be unsaid. These are not Kelly bets at all, because there is no rebalancing. You cannot dollar-cost-average into quitting. There is one print, no roll, no next month, and the tape does not run twice. Size them like the instrument they are: a one-lot in an illiquid market where you are also the market.

The real asset class is reversibility

Before sizing any decision, the desk now files it on two axes: how good is my information, and can the position be closed. Almost everything people agonize over is astonishingly reversible — emails, side projects, haircuts, most announcements. The agony is mispriced. The genuinely irreversible list is short and boring: health, the people you keep, the words you refuse to walk back, time. Reversible decisions deserve speed and small ceremony. Irreversible ones deserve the slow average, a second opinion, and a margin of safety measured in years.

Conviction is leverage. Leverage has margin calls, and the margin call for conviction arrives at three in the morning wearing your own voice.

Rules of the desk

One. Never full-Kelly anything you cannot undo. Half-size the irreversible, always, even when — especially when — you are sure.

Two. The unused margin is the position. Money not bet, bridges not burned, exits kept open: these are not cowardice, they are the liquidity that lets the rest of the book compound.

Three. When two irreversible choices both look full-Kelly, the trade is usually to do neither and wait for better data. Flat is a position. Flat is frequently the best position.

Backtest

Run over the full archive of my own worst nights: every disaster traces to a full-size entry into an irreversible instrument, executed on a one-day print (see filing № 003), on conviction borrowed from a story (see filing № 002). The sample is one life long, the result is monotonic, and the strategy — bet like you will be wrong and survive — has never once blown up. The objective was never to win the most. It was to still be at the table, in reasonable health, when the fat pitch finally arrives.