Thinking in Bets Ch. 5: The Decision Journal

Investment Background

Almost every trader keeps some form of log. Almost none of those logs actually support the discipline built in Chapters 1 and 4, because almost all of them record the same five things: symbol, entry price, exit price, size, P&L.

That log answers "what happened." It cannot answer the only question that improves a process: "was the decision that led here reasonable, given what was knowable at the time?"

This chapter's contribution is narrow and concrete: the specific fields a decision journal needs that a trade log does not, and why leaving them out silently defeats the entire pre-mortem/post-mortem discipline from Chapter 4.

The Wall Street Translation

The Five Fields a Trade Log Is Missing

A trade log records outcomes. A decision journal records the decision — written down before the outcome exists, which is the only moment it can be recorded honestly.

Field	What it captures	Why a P&L-only log can't recover it
Thesis	One or two sentences: what has to be true for this to work	Cannot be reconstructed after the fact without hindsight contamination
Confidence number	The Chapter 2 probability — not a word, a percentage	Memory rounds every outcome up to "I knew it" or down to "I never liked it"
Pre-mortem risks	The Chapter 4 "why did this fail" list, written before entry	These are the specific things resulting erases first
What would change my mind	The condition that should trigger an exit or a thesis review	Without this written down, exits become outcome-driven instead of thesis-driven
Position size and why	The sizing logic, tied to the confidence number	Separates "sized correctly for a mediocre edge" from "sized too large for a good one"

Notice P&L is not on this list. It still gets recorded, obviously — but it is the journal's least important column, because it is the one number the market supplies for free and the one number, per Chapter 1, that cannot by itself grade a decision.

A Worked Template

A single entry, filled out at entry and reviewed at exit:

DATE: [entry date]
THESIS: [1-2 sentences — what must be true]
CONFIDENCE: [XX%] — would I take the equivalent even-money bet?
PRE-MORTEM: [top 2-3 reasons this could fail, identified before entry]
INVALIDATION: [the specific condition that proves the thesis wrong]
SIZE & WHY: [position size, tied explicitly to the confidence %]
---
[AT EXIT] OUTCOME: [P&L — recorded last, not first]
DECISION GRADE: [good/bad — judged against THESIS and PRE-MORTEM only]

The order in the template is deliberate. Outcome sits at the bottom, filled in last, specifically so that filling in the top fields cannot be contaminated by already knowing how the trade ended.

Why Most People Who "Journal" Still Fail This

The common failure mode is not skipping the journal — it is filling in the top fields after the outcome is known, reconstructing a thesis to match.

A trader who writes "THESIS: earnings beat expected" only after checking that earnings did in fact beat has not journaled a decision. They have journaled a story. The only way this chapter's tool works is filling in everything above the dashed line before checking the outcome — which is why Chapter 1's one-sentence habit and Chapter 4's pre-mortem exist as separate, earlier steps. This chapter's contribution is simply giving them a permanent home instead of letting them evaporate.

Division of Labor With the Rest of the Library

Book	Owns
Way of the Turtle	The sizing math itself — position sizing formulas and risk-of-ruin calculations
Trading in the Zone ch4	Grading discipline over P&L, as a stated principle
This book, ch5	The specific artifact that makes that principle checkable months later — a written record with the decision-quality fields a plain trade log omits

Executable Trading Rules

1. Add four fields to whatever log you already keep: thesis, confidence %, pre-mortem risks, invalidation condition. Fill them in before the trade, not after. If your platform's log has no room for text, keep a separate note — the P&L log and the decision journal do not need to be the same document.

2. Never write the thesis field after checking P&L. If you catch yourself doing this, discard the entry — a reconstructed thesis is worse than no record at all, because it looks like data while actually being noise.

3. Review the journal quarterly, sorted by decision grade, not by P&L. Read every entry graded "bad decision" regardless of whether it won or lost money. That is where an actual process fix is found — a real trade log sorted by P&L alone will never surface it.

4. Write the invalidation condition as a fact about the world, not a price level. "Guidance cut" is an invalidation condition; "down 8%" is just a stop-loss wearing a disguise. The first tells you the thesis broke; the second only tells you the price moved, which Chapter 1 already established is weak evidence on its own.

Relevance to a Retirement Portfolio

The decision journal's structure transfers directly to the handful of large, infrequent decisions that actually determine retirement outcomes — setting the allocation, choosing the withdrawal rate, deciding whether to adjust either after a bad year.

Most retirement "journaling," if it exists at all, is a spreadsheet of account balances — the retirement equivalent of a P&L-only trade log. It shows what happened. It cannot show whether the plan that produced that balance was reasonable when it was set.

A one-time decision journal entry at retirement is enough, and does not need to be a trading habit: write the thesis (why this withdrawal rate, this allocation, this cash buffer), the confidence level, the pre-mortem risks (Retirement Decumulation Mechanics supplies the standard list — sequence risk, longevity risk, inflation risk), and the specific condition that would justify revisiting the plan (a named drawdown threshold, not "when it feels bad"). That single document is what lets a bad three-year stretch be graded honestly later — as a good decision meeting an unlucky sequence, or an under-planned one — instead of triggering a panicked reallocation graded only by how the balance looks today.

The core recommendation does not change based on any single year's balance. A low-cost, diversified allocation with a written, pre-mortemed withdrawal plan stays the plan; the journal's job is only to let you tell, later, whether staying the course was disciplined or merely lucky.

Chapter 6, the last, states plainly what this discipline cannot do — and where its value actually stops.