In one sentence: In part two, Jake Taylor explains how to track rejected ideas and score your predictions with a Brier score so you can separate luck from skill, plus how environment, time of day and a growth mindset improve your process. (Continues 294.)
Key ideas
- Repeatable process. Like a controlled experiment, a fixed process lets you change one thing at a time and learn what helps. [01:30–03:00]
- Track your rejections. Taylor logs why he passed on a company and how it did afterwards. Most investors only track what they bought. [03:00–05:00]
- Test the too-hard pile. Buffett's "too hard" basket is fine, but you don't know its cost unless you check what you put in it. If everything there rose, dig deeper. [05:00–06:30]
- Loud misses distort memory. A missed 100x stands out, while the rejected names that went to zero fade. Review the whole reference class, not the anecdote. Danielle names Amazon and Lululemon as her own misses. [06:00–09:00]
- Test each filter. Taylor is wary of leverage. He says he can't yet say whether that filter helps or hurts his results, but he is collecting the data. [09:00–10:30]
- Environment is personal. Buffett and Munger memorabilia, paper books to mark up, family photos, and couches. He keeps electronics out of reach, since even their presence drains attention. [12:00–20:00]
- Timing. He does deep work early, when "vigilance" is highest, and creative work later when his mind is looser. Danielle finds mornings best for everything. The two disagree on the Pomodoro method: she dislikes it; he suggests longer blocks. [19:30–27:30]
- Compare yourself with last month's you. Small gains compound and setbacks are normal. [28:00–31:00]
- Brier score. Taylor predicts the five drivers of return (revenue, margins, earnings, multiple, dividends) with probabilities, then scores each. One year gives one return but five scored predictions. A rising stock with five wrong predictions was probably luck, and a falling stock with five right ones was bad luck. Paper predictions on stocks you don't own work, though real money adds seriousness. [31:00–37:00]
How it maps to RuleOne
- The screen could keep a log of rejected tickers with the reason and later price. That is not built, but the data (price history, filters) is already there.
- /holdings/ shows what you own. The rejection log is the missing half.
Buffett, Munger and Graham links
- The "too hard" pile is Buffett's well-known desk basket.
- The Brier score is from the meteorologist Glenn Brier (1950), later used in Philip Tetlock's forecasting research (Superforecasting, 2015).
- Munger's "sit-down money" and inversion pair with the idea of judging process rather than outcome. Check the source before quoting.
Words to know
- Brier score: a measure of how well your stated probabilities match what happened.
- Reference class: the group of similar past cases you compare against.
- Mistake of omission: passing on a good idea.
Try this
Open /stocks/, pick 5 companies you passed on recently, and write the reason for each. Set a calendar reminder in 12 months to look at how they did.
Check yourself
- Why track what you rejected?
Answer
To learn whether your filters (and too-hard pile) cost you good ideas. - How does a Brier score separate luck from skill?
Answer
It scores your probability forecasts of the drivers, giving several data points per year instead of one return. - What should you do if a stock rose despite wrong predictions?
Answer
Treat it as luck and don't raise your confidence.
Short quotes
"Compare yourself with you, a month ago." (Taylor, ~29:30, auto-transcribed paraphrase)