Asymmetric Bet Sizing Ch. 4: Liquidating on a Broken Thesis, Not a Broken Price

Investment Background

way-of-the-turtle chapter 4 specifies the library's mechanical exits in full — the 2N hard stop, the ten-day channel exit, the equity-curve circuit breaker — and it is the correct machinery for a system whose edge is statistical. None of it is restated here.

This chapter is about a different exit trigger, and the difference is not a refinement. It is a different object being monitored.

A price stop monitors the market's opinion of your position. A thesis exit monitors the reasoning that put you there. These come apart constantly, in both directions, and the case that matters — the one no rule-based system can see — is a position that is comfortably profitable while the reason for holding it has quietly ceased to be true.

The Wall Street Translation

Two Triggers, Monitoring Two Different Things

	Price stop	Thesis exit
What is monitored	The market price against your entry	The stated reasoning behind the position
Fires when	Price moves against you by a defined amount	A named condition in the thesis becomes false
Can fire while profitable	No	Yes, and this is its whole reason for existing
Requires	A rule and a price feed	A thesis written down before entry, in falsifiable form
Owned by	way-of-the-turtle ch04	This chapter

Read the third row twice, because it is the entire chapter. A price stop cannot fire on a winning position. A thesis exit is at its most valuable precisely there — the structural condition has resolved, or the institution has capitulated, or the policy has reversed, and the remaining asymmetry that justified the size no longer exists, even though the position is up and nothing on the screen is complaining.

The four cases are worth setting out, because only one of them is handled by conventional risk management.

	Thesis intact	Thesis broken
Position profitable	Hold. The reason and the result agree	Exit, and this is the case nobody takes
Position losing	The genuinely hard case — the thesis may need more time, or you may be wrong and defending	Exit immediately. Both signals agree; no discipline is required

The top-right cell is the contribution of this chapter. Nothing in the library's risk machinery points at it, because every mechanical exit is a function of price and the price is fine.

The bottom-left cell is where discretionary traders actually die, and it deserves naming rather than sympathy: "the thesis just needs more time" is the sentence that precedes most large discretionary losses, and it is unfalsifiable, which is exactly why the falsification condition has to be written before entry rather than argued after.

Why This Requires Something Written in Advance

A thesis exit is only operable if the thesis was stated in a form that can become false, and a thesis stated after the fact never can.

thinking-in-bets-duke chapter 5 owns the decision journal and the reason for it: memory reconstructs reasoning after the outcome is known, so the record has to exist before. thinking-in-bets-duke chapter 2 owns turning beliefs into bets, which is the same discipline applied to belief. Neither is re-taught here.

What this chapter adds is what the journal is used for in this specific method. For Duke the journal improves decision quality by separating decision from outcome. Here it has a second and more operational function: it is the document you read to determine whether the exit condition has fired. Without it, "has my thesis broken" is answered by asking your current self, who is holding the position, has an opinion about it, and is not a neutral party.

The entry that makes an exit possible is short and has to be specific: the structural condition claimed, the mechanism by which it resolves, and the observable events that would establish it is not true. "The currency falls" is not a falsification condition; it is the payoff. "The central bank obtains credible external funding, or the domestic political cost of the defence drops away" is one — it names an event that can be observed and that ends the argument.

A thesis with no such condition cannot be exited on evidence, only on price, which means it has quietly become one of way-of-the-turtle's positions without the benefit of way-of-the-turtle's rules.

Why the Profitable Exit Is the Hardest Action in This Book

Exiting a winner on thesis-invalidation is psychologically harder than taking a loss, and the reasons are documented rather than speculative.

The position is currently agreeing with you. Every screen confirms the decision. Selling requires acting against present evidence on the strength of a document written months earlier by a version of you who did not know how it would go.

thinking-in-bets-duke covers resulting — judging decisions by outcomes — and this is resulting in its most seductive form, because the outcome so far is good. The trader who holds anyway is not being greedy in any recognisable sense; they are being persuaded by evidence. The evidence is simply about the wrong object: it concerns the position's performance, not the thesis's validity.

And there is a second, quieter cost that is specific to this method. The trader who does not exit on invalidation is now holding a large position with no thesis, which means the size — justified by an asymmetry that has been consumed — is now carrying full risk for whatever ordinary opinion has replaced the original reasoning. The position did not stop being large when it stopped being justified.

Where This Method Is Weakest, Stated Honestly

The thesis exit has a failure mode of its own and it would be dishonest to close the chapter without it.

A discretionary trader can rewrite the thesis. The condition fires, and instead of exiting, the reasoning is amended: the mechanism was right but the timing was different, the invalidating event was not really invalidating, the underlying argument still holds in a modified form. Each amendment is individually plausible, and a thesis that can be amended cannot be falsified.

This is the specific advantage way-of-the-turtle holds over everything in this book. A price stop cannot be reasoned with. man-who-solved-the-market-simons chapter 5 states the same principle at the level of a firm: nobody overrides the system, not even the founder, because a system with an override is not a system. The thesis exit is an override-shaped hole by construction, and the only defences available are that the condition was written before entry and that somebody other than you can read it.

Executable Trading Rules

1. Write the falsification condition before entry, or do not size the position as one of the few. A thesis that names no observable event that would end it is not a thesis, and it cannot be exited on evidence. thinking-in-bets-duke chapter 5 has the instrument; this is the use it is put to here.

2. Review the thesis on a schedule, not when the price moves. Price-triggered review guarantees you only re-examine losers, which leaves the profitable-but-invalidated case permanently unexamined. A fixed calendar review sees all four cells.

3. Exit the profitable position when the condition fires, and expect this to feel wrong. It is the hardest action in this book and the one that most distinguishes the method. The screen is reporting on the position; the document is reporting on the thesis, and only one of them justified the size.

4. Do not amend a thesis after an invalidating event, ever. A modified thesis is a new position and should be sized from zero as a new decision, with its own falsification condition. Amendment in place is how a discretionary process quietly loses its only external check.

5. Have someone else able to read the entry. The trader holding the position is not a neutral reader of their own reasoning. This is the closest available substitute for the systematic trader's inability to argue with a stop.

Relevance to a Retirement Portfolio

This chapter contains the single most transferable idea in the book, and it transfers in a form that reduces risk rather than adding it.

Most positions in a retirement portfolio are held for a reason nobody has written down, and the practical consequence is that they can only be reviewed on price — which means they get examined when they fall and never when they rise. A holding that has become five times its intended weight is not reviewed at all, because nothing about it looks like a problem.

Writing down why a holding exists, and what would mean it no longer should, converts an unreviewable position into a reviewable one. For a legacy stock, an inherited concentration, a fund bought for a reason nobody now remembers, the exercise usually terminates the position on its own — because most such holdings turn out to have no statable reason at all.

The warning attached to this is important. A thesis exit is not a licence to trade a diversified index core. A globally diversified index fund is not held on a thesis about anything; it is held because forecasting is unreliable, which is a position that no event invalidates. the-interrupted-plan covers what happens when a retiree liquidates a core holding under stress, and this chapter must never be read as permission for that.

The standard recommendation is unchanged: a core of low-cost, globally diversified index funds, no leverage, a cash buffer covering essential spending, and withdrawal rules containing an adjustment mechanism. The thesis exit applies to the satellite, never to the core — and its honest use for most readers is not exiting at all, but discovering how many holdings never had a thesis to begin with.

Chapter 5 is the chapter this method most needs: what happened when the person who wrote all of this got it badly wrong.