Four Years of Quarterly Notes: What the Record Shows

Sixteen quarterly notes, each written on the first working day of a quarter, each answering the same three questions. I reread all of them last month.

The value was not in any individual entry. It was in the pattern across them, which I could not have seen from inside any single quarter.

What the record shows about my predictions

I make directional calls roughly once a quarter, usually without meaning to. Counting them:

• Broadly right about direction: seven
• Broadly wrong: six
• Too vague to score: three

That is a coin flip, which is the correct expected result and was still deflating to calculate. More instructive: my confidence was highest in four entries, and three of those four were wrong.

The confident ones are systematically worse. I have seen this claimed in the literature and it is different to observe it in your own handwriting.

What the record shows about my errors

Reading sixteen entries in sequence, the mistakes fall into three repeating types.

Timing structural views. I am reasonably good at identifying what will eventually happen and consistently wrong about when. Every date I have attached to a structural prediction has been early by a year or more.

Confusing salience with information. After any dramatic event, the following quarter’s note over-weights it. The 2022 failures produced three consecutive entries about counterparty risk, which was reasonable, and one about abandoning a position that had nothing to do with any counterparty, which was not.

Rationalising during drawdowns. The month-three pattern shows up repeatedly. Notes written during declines contain more responsible-sounding vocabulary and worse reasoning than notes written in flat periods.

What the record shows about my process

The mechanical decisions outperformed the discretionary ones, consistently, across all sixteen quarters.

Scheduled purchases beat timed ones. Scheduled rebalances beat opportunistic ones. The positions I left alone did better than the ones I managed.

The single exception was one quarter in 2022 where I added outside schedule during a sharp decline, which worked. I have thought about whether to count that as evidence for discretion. I do not think it is. One favourable outcome from one deviation, in sixteen quarters, is what you would expect from chance.

What has not changed in four years

I still feel the same thing during declines. The notes from 2022 and the notes from 2025 describe an identical internal experience in almost identical language.

I had assumed experience would reduce this. It has not. What it has done is separate the feeling from the action, which turns out to be the only achievable goal.

The format, for anyone who wants to copy it

Three questions, same day each quarter, fifteen minutes.

1. Has any invalidation condition occurred? Evidence, not impression.
2. Is each position within its band? Numbers.
3. What do I believe now that I did not believe last quarter, and why?

The third question is the one that produces the useful record. It forces you to state a view in a way that can later be checked, which is uncomfortable by design.

The recommendation

Start now, whatever the market is doing. The entry you write today will be worth very little to you today and a great deal to you in three years, and there is no way to acquire it retroactively.

Mine cost fifteen minutes a quarter and have been worth more than every piece of analysis I have ever read.