Skip to content
Menu
Finabase Learn

Investor psychology Lesson 15 of 15

Articles · Behavioral finance

How to keep an investment decision journal (with a template)

A step-by-step way to record investment decisions and forecasts, score your calibration with the Brier score, run a premortem, and what the evidence says about each.

On this page
  1. What goes in an entry
  2. Scoring a year of entries: calibration
  3. Three additions worth the time
  4. What the evidence says
  5. Common ways a journal fails
  6. How it differs from a trading journal
  7. How Finabase helps
  8. Frequently asked questions
  9. Summary
  10. Check your understanding
  11. Exercise

A decision journal is a dated record of what you decided, why, what you expected to happen and how sure you were, written before you know the result. Later you add what actually happened and compare.

Most of the biases in this section share a weakness: they work by editing memory. Hindsight moves your old forecast toward the outcome. Outcome bias regrades the decision by the result. A written record is the one thing they cannot edit. This guide gives a template that takes about two minutes per decision, shows how to score a year of entries, and is plain about the evidence: several of the ingredients have been tested, and the journal as a whole has not.

What goes in an entry

One journal entry: what to write before, and what to add after Before: written on the day Date and price: Mar 3, $84.20 Decision: buy 40 shares (4% of account) Why, in two sentences Forecast: revenue growth above 10% next four quarters. Confidence 70%. What would prove me wrong Review date: Mar 3 next year Mood and what prompted this After: written at the review Did the forecast come true? Yes / no Was the reasoning right, or was I right for another reason? What did I not know that I could have found out? Would I make the same decision again with the same information? Score: process and outcome, separately The left side is never edited after the day it is written. That is the whole point of the record.
A decision-journal entry. The left side is written before the outcome and never edited; the right side is added at the review.

Before, on the day (about two minutes)

  1. Date, price and what you did. Include the size as a share of the portfolio. "Bought 40 shares at $84.20, 4% of the account."
  2. Why, in two sentences. If it takes more, you may not have a reason yet.
  3. A forecast with a number and a date. Not "I think it will do well" but "Revenue growth will stay above 10% over the next four quarters: 70%." It must be something a stranger could check on the review date.
  4. What would prove you wrong. The observation that would make you sell or change your mind. Written now, it becomes an if-then plan: "If growth falls below 5% for two quarters, I sell."
  5. A review date. Put it in a calendar or an alert.
  6. How you feel and what prompted this. One line. "Read three articles this week; a little afraid of missing out." Over a year this line is surprisingly informative.

Deciding not to act is a decision too. An entry that says "Considered selling during the drop; did not, because nothing in the thesis changed" is as valuable as a buy.

After, at the review

  1. Did the forecast come true? Yes or no, by the definition you wrote.
  2. Was the reasoning right? A forecast can come true for a reason you never considered.
  3. What did you not know that you could have found out? This separates bad luck from a gap in research.
  4. Would you make the same decision again with the same information?
  5. Two grades: process and outcome. Kept separate, as in process over outcome.

One rule holds the whole thing together: the "before" side is never edited.

Scoring a year of entries: calibration

Once you have a few dozen forecasts with probabilities, you can test something most investors never learn about themselves: whether their confidence means anything.

You are well calibrated if, of all the things you called 70% likely, about 70% happened. Research on calibration going back to the 1970s finds that people are generally overconfident, most of all on hard questions, which is what investment forecasts are. Overconfidence and overtrading covers the evidence.

A worked example

An illustrative investor made 40 checkable forecasts in a year and grouped them by the confidence written down at the time.

Confidence stated Forecasts Came true Hit rate
60% 12 7 58%
70% 12 7 58%
80% 10 6 60%
90% 6 4 67%
All 40 24 60%
A calibration check from a journal of 40 forecasts 40% 50% 60% 70% 80% 90% 100% 50% 60% 70% 80% 90% 100% Perfect calibration 7 of 12 right 7 of 12 right 6 of 10 right 4 of 6 right Confidence written down before the outcome Share of those forecasts that came true Illustrative journal: stated confidence averaged 73%, 24 of 40 came true (60%). Brier score 0.26.
Calibration of the illustrative journal: stated confidence against the share of forecasts that came true. Points below the diagonal are overconfident.

The picture is the common one. This investor is right about 60% of the time whatever confidence they state. At "60% sure" they are calibrated. At "90% sure" they are right two times in three. Average stated confidence was 72.5% and the hit rate was 60%.

The practical reading is specific: this investor's feeling of near-certainty carries no extra information, so decisions that depend on it, such as a larger position because "I'm sure about this one", are not justified by the record.

The Brier score

The Brier score, introduced by the meteorologist Glenn Brier in 1950 to grade weather forecasts, turns the table into one number.

For a forecast stated at 70% that came true, the score is (0.7 − 1)² = 0.09. For one stated at 90% that failed, it is (0.9 − 0)² = 0.81. Confident misses are punished heavily.

The illustrative journal scores 0.262. Had the investor simply said "60%" every time, the score would have been 0.240, and saying "50%" every time scores 0.250. So this investor's stated probabilities were slightly worse than a coin's. That is not a verdict on their stock picking: 60% correct may be good. It is a verdict on their confidence, which is what position sizes are built on.

Forty forecasts is a small sample, and a few results either way would move these numbers. The score becomes meaningful over a hundred or more.

Three additions worth the time

A premortem. Before committing, imagine it is a year later and the investment has failed badly. Write the three most likely reasons. The technique was popularized by Gary Klein in a 2007 Harvard Business Review article, which cites a 1989 study by Deborah Mitchell, Edward Russo and Nancy Pennington for the claim that imagining an event as already having happened increases the ability to identify reasons for it by 30%. The study's own summary is more modest: what helped people generate reasons was treating the outcome as certain rather than merely possible. The exercise costs five minutes and its output goes straight into "what would prove me wrong".

A short checklist. The same five or six questions before every purchase: Do I know how this company makes money? What is the position size? What would make me sell? What is the case against? Have I checked this against what I already own?

If-then rules. "If the price falls 20% and my thesis is unchanged, I do nothing. If the thesis is broken, I sell that day." These are what psychologists call implementation intentions.

What the evidence says

It is worth being exact about what has and has not been shown.

Ingredient Evidence Caution
Keeping score with probabilities In a two-year geopolitical forecasting tournament, Barbara Mellers and colleagues found that brief training in probabilistic reasoning, plus scoring, improved accuracy. Geopolitical questions, not investments.
Training about biases Carey Morewedge and colleagues found one training game reduced several decision biases by about 32% immediately and 24% two months later. Measured on test questions, not on real decisions.
If-then plans A meta-analysis of 94 tests by Peter Gollwitzer and Paschal Sheeran found a medium-to-large effect (0.65 standard deviations) on following through on goals. Later reviews correcting for publication bias report smaller effects.
Checklists A surgical checklist in eight hospitals was followed by a fall in deaths from 1.5% to 0.8%. A later study of 101 Ontario hospitals that adopted checklists found no significant reduction.
An investment journal as a whole None found. We searched and did not find a study that tests whether keeping an investment or trading journal improves returns. Anyone who says a journal "is proven to" improve results is going beyond the evidence.

The checklist pair is the most instructive. The first result was large and famous. The second suggests that a checklist helps when people use it as a thinking tool and does little when it is a form to be completed. A journal is surely the same. Its value is in the comparison at review time, and an unreviewed journal is a diary.

Common ways a journal fails

  • Vague forecasts. "Should do well long term" cannot be wrong, so it cannot teach.
  • No probability. Without a number there is nothing to calibrate.
  • Writing after the fact. An entry written a week later is already contaminated by what happened in that week.
  • Recording only trades. The decisions not to act are where most long-term returns come from, and they vanish from memory fastest.
  • Never reviewing. Set the review date when you write the entry.
  • Editing the left side. Add a note below if you must. Never change the original.

How it differs from a trading journal

A trading journal, covered in trading plan and journal, records each trade's setup, entry, stop and result in multiples of risk, and its main use is statistical: win rate, average win and loss, maximum drawdown over hundreds of trades. A decision journal suits investors who make a few decisions a year. Its unit is the forecast, and its main use is checking beliefs and confidence. Active traders can keep both in one file.

How Finabase helps

Each symbol page has a Notes tab where a signed-in user can keep a note on that symbol, which syncs with the Stock Master app. It is a natural home for the "before" entry, next to the chart and the financials the reasons came from. An alert on a price or percentage move can stand in for the "what would prove me wrong" line, so the review happens when the condition is met and not when you happen to remember. A portfolio records each trade with its date and price and shows realized and unrealized gain, which supplies the "outcome" half of each entry without relying on memory.

Frequently asked questions

What is a decision journal?

It is a written record of a decision made at the time: what you decided, why, what you expected with a probability, and what would change your mind. You review it later against what happened.

What should an investment journal entry include?

The date and price, the action and position size, two sentences of reasoning, one checkable forecast with a confidence level and a date, the condition that would prove you wrong, a review date, and a line on your state of mind.

What is a good Brier score?

Lower is better. Zero is perfect and 0.25 is what you get by always saying 50%. For yes-or-no investment forecasts, anything reliably below about 0.2 over a hundred or more forecasts shows real calibration. Compare your score with the score from always stating your overall hit rate.

Does keeping an investment journal improve returns?

There is no published study showing that it does. What a journal reliably provides is an accurate record of what you believed, which is the precondition for learning anything from your own results.

Summary

A decision journal records the decision, the reasons, a forecast with a probability and a date, and what would prove it wrong, all before the outcome is known. Reviewed later, it gives two things memory cannot: an unedited account of what you thought, and enough forecasts to measure calibration with the Brier score. Scoring forecasts, if-then plans and checklists each have some experimental support, with caveats, and the journal as a whole has not been tested against investment returns. Its honest promise is smaller and still worth having: you will know what you actually believed, and how much your confidence is worth.

Check your understanding

5 questions on this lesson. Each answer comes with the reason, and nothing is scored.

  1. Which forecast can be checked on its review date?

    1. “This company should do well over the long term.”
    2. “Revenue growth stays above 10% for the next four quarters: 70%.”
    3. “I have a good feeling about this one.”
    Show the answer

    B. “Revenue growth stays above 10% for the next four quarters: 70%.” A checkable forecast states an observable result, a date and a probability. Without all three there is nothing to compare with what happened and nothing to calibrate.

  2. Over a year an investor calls 20 things “90% likely” and 12 of them happen. What does this show?

    1. Good calibration
    2. Overconfidence: events called 90% likely happened 60% of the time
    3. Underconfidence
    Show the answer

    B. Overconfidence: events called 90% likely happened 60% of the time Calibration compares stated confidence with the hit rate. A 60% hit rate at stated 90% means the feeling of near-certainty carries less information than the investor thinks.

  3. A forecast stated at 90% fails to happen. What is its Brier score, and why does that matter?

    1. 0.01, so confident misses hardly count
    2. 0.81, so confident misses are punished far more than cautious ones
    3. 0.50, the same as any miss
    Show the answer

    B. 0.81, so confident misses are punished far more than cautious ones (0.9 − 0)² = 0.81. A miss stated at 60% scores 0.36. The squared error makes overconfidence expensive, which is the behavior the score is meant to discourage.

  4. Why must the “before” side of an entry never be edited?

    1. Because editing takes too long
    2. Because its whole value is being a record that hindsight cannot revise
    3. Because brokers require it
    Show the answer

    B. Because its whole value is being a record that hindsight cannot revise Memory moves old forecasts toward the outcome. The entry is useful only as evidence of what was believed on the day.

  5. What does the published evidence say about investment journals and returns?

    1. Journals are proven to raise returns
    2. Some ingredients, such as scoring forecasts and if-then plans, have experimental support, but no study tests a journal against investment returns
    3. Journals have been shown to lower returns
    Show the answer

    B. Some ingredients, such as scoring forecasts and if-then plans, have experimental support, but no study tests a journal against investment returns The honest claim is narrower: a journal gives an accurate record to learn from. The claim that it improves results has not been tested.

Exercise

One task to do in Finabase. Nothing is graded or saved.

Open the Overview page of a company you own or follow (AAPL is linked). Write one journal entry on paper or in the symbol’s Notes tab: today’s date and price, one checkable forecast for a date three months from now (for example “the price will be above today’s close”), your confidence from 50% to 100%, and one observation that would prove you wrong. Then set a reminder for the review date.

Open the AAPL page
What you should find

A usable entry has a result a stranger could verify, a date and a number. If your confidence was above 80% for a three-month price forecast, note that: over three months a large stock is up only somewhat more often than it is down. After ten such entries you can work out your first hit rate and Brier score.

Sources and further reading

  1. Brier — Verification of forecasts expressed in terms of probability, Monthly Weather Review 78(1), 1950
  2. Lichtenstein, Fischhoff and Phillips — Calibration of probabilities: the state of the art to 1980, in Kahneman, Slovic and Tversky (eds.), Judgment under Uncertainty, 1982
  3. Mellers et al. — Psychological strategies for winning a geopolitical forecasting tournament, Psychological Science 25(5), 2014
  4. Morewedge, Yoon, Scopelliti, Symborski, Korris and Kassam — Debiasing decisions: improved decision making with a single training intervention, Policy Insights from the Behavioral and Brain Sciences 2(1), 2015
  5. Gollwitzer and Sheeran — Implementation intentions and goal achievement: a meta-analysis of effects and processes, Advances in Experimental Social Psychology 38, 2006
  6. Haynes et al. — A surgical safety checklist to reduce morbidity and mortality in a global population, New England Journal of Medicine 360(5), 2009
  7. Urbach et al. — Introduction of surgical safety checklists in Ontario, Canada, New England Journal of Medicine 370(11), 2014
  8. Mitchell, Russo and Pennington — Back to the future: temporal perspective in the explanation of events, Journal of Behavioral Decision Making 2(1), 1989
  9. Klein — Performing a project premortem, Harvard Business Review, September 2007