Subscribe
Share
Share
Embed
Two men who won the Nobel Prize for pricing options helped found a fund that lost 4.6 billion dollars in four months. Ed Thorp ran the formula that names the right bet size and spent 19 years betting half of it, because you can count your odds at a blackjack table and only estimate them in a market. Eliud Kipchoge, an Okinawan dinner table and the race to the South Pole all run the same rule at different speeds.
Show Notes
Two men who won the Nobel Prize for pricing options helped found a fund that lost 4.6 billion dollars in four months. Ed Thorp ran the formula that names the right bet size and spent 19 years betting half of it, because you can count your odds at a blackjack table and only estimate them in a market. Eliud Kipchoge, an Okinawan dinner table and the race to the South Pole all run the same rule at different speeds.
Featuring: Eliud Kipchoge, Ed Thorp, John Kelly, Roald Amundsen, Robert Falcon Scott, Peter L. Bernstein
Brought to you by:
Trends.vc: Discover new markets and opportunities.
HeadsUp: Know when your competitors move. Before your customers do.
Multi: Turn any content into a YouTube video.
Chapters:
(00:00) Bernstein on survival
(00:05) Long-Term Capital Management's collapse
(00:20) Ed Thorp's Kelly formula fund
(01:12) Thorp bets half of the formula's number
(01:35) Kipchoge trains easy most weeks
(01:58) Amundsen paces 15 miles a day to the pole
(02:09) Scott's race ends in a blizzard 11 miles short
Where to find Dru:
I write an essay daily. Get them free
Follow me on X
Watch my video essays on YouTube
Follow Dru's Notes on Apple Podcasts or Spotify
Subscribe to Dru's Notes for a new episode every day.
Creators and Guests
PB
Guest
What is Dru's Notes?
One mental model each day in a few minutes. Habits, business, building products and thinking clearly. From Dru Riley, founder of Trends.vc.
"Survival is the only road to riches."
Peter L. Bernstein
Two men who won the Nobel Prize for pricing options helped found a fund that bet 25 dollars for each dollar it held.
Long-Term Capital Management lost 4.6 billion dollars in four months, almost everything it owned.
Ed Thorp ran a fund on a formula that told him exactly how much to bet. For 19 years he compounded at about 19% a year.
Thorp's formula came from John Kelly. Give it a bet you want to make and it returns one number: the share of your money that bet should get.
Say you find a bet you win 60 times out of 100.
Win and you double what you put in. Lose and it's gone.
Start with 100 dollars and bet a fifth of what you hold each round.
Win the first and you hold 120. Lose the next and you hold 96.
Bet two fifths instead. Win the first and you hold 140. Lose the next and you hold 84.
Run all 100 rounds with the same 60 wins.
A fifth each round turns your 100 dollars into about 750.
Two fifths turns it into about 78.
Thorp never bet the number the formula gave him. He bet half of it, so where the formula said a fifth he put down a tenth.
That earned about three quarters of what the full number would have earned. His fund's value rose and fell half as far.
The formula only works if you know your chances exactly.
He could count them at a blackjack table. In the market he could only estimate.
Eliud Kipchoge runs up to 220 kilometres a week and runs hard on only about a fifth of those kilometres.
He won the Olympic marathon twice on training he never ran flat out.
Diners at an Okinawan table stop eating when they are 80% full.
The rule is called hara hachi bu.
Ask which of your weeks you could repeat all year.
Roald Amundsen held his party to 15 nautical miles a day on the way to the South Pole, well short of what the dogs and the men could have covered. His men rested up to 16 hours a day.
Robert Falcon Scott raced him for that pole the same Antarctic summer and marched on whatever the weather gave him. Long days under a clear sky. Tent-bound days when the storms came.
Amundsen stood at the pole on 14 December 1911 and walked back into his hut on 25 January, all his men alive and heavier than when he left.
Scott reached the same spot 34 days behind him and found the Norwegian flag already flying there. His party of five turned for home.
Edgar Evans died in February. Lawrence Oates said he was just going outside and might be some time, then walked into a March blizzard.
Scott and the last two lay down 11 miles from a depot stocked with food and fuel.
A blizzard held them there until they died.
The attention you have today is all the capital you have to bet with.
Give each habit a smaller share than your best week could carry.
Run it at the smallest share you'll still keep on a bad day.
Too big a share ends the habit.
Which habit can you protect by asking less of it?
CLOSED 2026-08-04 (round 6). The post ships with no personal story. Dru: "i removed my personal story since it felt forced in." That is the third consecutive round this one slot was rejected: round 3 on the learning-habit beat, round 4 on the segue and the low stakes, round 6 on the launch loop. Per process-feedback ยง5c a fourth agent-written variant is the wrong move, because the defect sits in the slot rather than in any candidate's wording. The slate below stays as a record of what was tried, not as a queue.
Two consequences, both accepted rather than worked around. audit-composition returns MISSINGI, overridden on his words above. And the absence run the personal beat used to break is now broken by a reader-facing line instead (see ## Reader Beat Options).
Slate generated 2026-08-04 after the round-4 note "you have a learning habit? you jump right into it without segue or anything and learning habit feels low stakes after talking about death. we need to list personal story premise options out." All five are sourced to context.md with dates. No decoys.
1. The launch that ended the marketing habit. (context.md 2026-07-08) Build to a launch, win product of the day, stop marketing because it felt handled, watch customers keep arriving, then watch them stop. Right about the product, wrong about the size of the bet on it. (SHIPPED 2026-08-04, CUT the same day. Dru's note above. It cleared every gate and still read as forced.)
2. The cold shower got better, the business didn't. (context.md 2026-06-04) Pulled the cold-shower floor back to 34 seconds because it was already clearing its bar, then named the pattern: pushing habits that were already winning while marketing, the stuck one, sat below the line. Exact big-idea match. The post already links the published post of that name. Cost: needs a plain gloss for the rate vocabulary.
3. Twelve reps against thirty-four. (context.md 2026-06-04) Sorted each habit by age. The 12 oldest all cleared the bar. Marketing sat at 12 reps against a target of 34. The habit at the top of the list was the smallest position held. Tightest arithmetic match. Cost: rate vocabulary plus Fibonacci numbers to introduce mid-post.
4. The floor that was too high. (context.md April 2026) Ran an experiment floor of 55 a day until it collapsed every third week, then dropped it to 34. The lower number produced more total output because it survived hard Mondays and illness. Exact match for the closing law and for Amundsen. Cost: it is about one habit's size rather than priority across habits, so it serves the second half.
5. Banking the hard habit while the easy one sat at zero. (context.md 2026-07-28) Banked further ahead on an already-deep hard habit while an easier one sat due at zero, then missed learning that night after starting late on low energy. Sizing by perceived difficulty instead of priority. Smallest change from the round-4 draft, since it keeps the learning habit.
Note on the chosen premise. context.md describes the launch loop as a pattern across years and names no particular launch, so the body renders the loop in its own beats without naming one. If a single named launch would land harder, the launch and its outcome have to come from Dru.
New section, 2026-08-04 (round 6). Cutting the personal story took the post's only I/you beat out of a 261-word stretch, over the 200-word hard block in scripts/listenability.mjs. One line placed between Kipchoge and Amundsen splits it into 84 and 151 words. It sits BETWEEN two stories rather than inside either one, and it has to be a question, a conditional or an imperative: a declarative about the reader's life is the reader-fact violation, since nobody measured it.
1. "Ask which of your own weeks you could repeat all year." (Imperative, CHOSEN. In body, between the Okinawa pair and the polar block.) Plants repeatability, which is the mechanism under all three restraint examples and which the post never states outright. States no principle, which matters because the polar stories are still queued and a law here is LESSONEARLY.
2. "Count the days last week you ran flat out." (Imperative. REJECTED by the gate, 2026-08-04.) Shipped first and audit-signal-density returned REPEAT, restates 16: reflect on flat-out running days. That is the bare-reader-address-beat class the gate names by rule: strip the "you" and nothing survives that the Kipchoge line did not already carry. The gate's prescribed fix is a substantive beat, never deletion, because deleting it brings the 200-word absence block back.
3. "How many of last week's days did you run at your limit?" (Question.) Same payload as 2 in interrogative form, so it inherits the same REPEAT problem.
4. "Ask what your best week would cost you if you ran it every week." (Imperative + conditional.) Carries the cost rather than the count. Risk: it is one step from stating the post's law outright, and the law belongs after the stories.