The idea behind Napkinbook

Whiteboarding the way you think

Three years of wanting one surface that works at the speed I think and computes without a context switch. What it is, why it is not a spreadsheet, and what I ran on it during the World Cup.

3 min read
  • napkinbook
  • origin

Napkinbook is a computational whiteboard I built for myself: somewhere to diagram my ideas, do the analysis, and run the back-of-the-napkin math it is named after. My background is prediction markets, and every trade I have ever liked started on a napkin. A base rate, two or three things I believed about this case, and a number at the end. That is how I think, loose and fast. The trouble is that napkins die: the math goes stale the moment you write it, and a week later you are holding a number with no memory of how you got there.

+BASE RATE18%EVIDENCE+9COMBINE→PYOURS27%MARKET34%EDGE−7move a driver,your edge moves

Napkinbook is the napkin that stays alive. Excalidraw and tldraw are great at diagramming; I needed the diagram to do something. A real forecast asks you to hold more variables than a head can carry, and the board holds them for me.

Why not a spreadsheet?

A spreadsheet holds the same numbers and hides the reasoning: the argument lives in formulas nobody reads. A notebook hides it in code. On a canvas the argument is the picture. When I disagree with my own forecast a month later, I can see exactly which wire to argue with.

The World Cup test

This summer I ran my World Cup positions on boards. A knockout match, a favorite priced richer than I believed. A base rate node on historical knockout data, three pushes wired in, and the combine holds it all to a real probability, sitting next to the market’s price.

++base rate55%two extrarest days+4 ptsstarting CBsuspended−6 ptstournamentform+3 ptsfavorite advances56%base 55%market price0.62market says 62¢board says 56%the gap is the tradeBut like...is it though?

The gap between those two nodes is the trade. When a number moved (an injury update, a lineup leak) I edited one node and the forecast re-derived itself. The data lives on the board too: drop a CSV on the canvas, query it with SQL in the browser, wire the result into a base rate.

Not just markets

The same boxes do a month’s budget. Your take-home flows out to everything you spend, and a pool node keeps the running balance, so “what’s left” updates itself the moment any number changes:

take home4,200rent1,500groceries600transport400utilities300phone100fun300whats left over1,000running balance

or lock in an arbitrage:

++book A: yes0.54book B: no0.42Σcost of both0.96pays 1.00either way.4¢ locked in.

and when there is no math at all, it is still a good napkin:

meSamPriyaJakeNinaTheoMiaDevonOllie, 1yr 3exes. it's fine.ended at the weddingthe crush everyone knowssituationship (alleged)roommatesstill owes me $40sistersallegedly
Ollie is, allegedly, Theo's

Present it from your phone

You author a walkthrough on the board, then run and annotate it live from your phone or iPad, front and center on stage instead of hiding behind a laptop. That is the feature I did not know I was building.

Share it live

+past launches hit35%paid + creators+20 ptsoff-season−8 ptshits Q3 goal47%base 35%PRIYASAMJAKE
one link, no account

People can join a room, and they can be booted out of one. Boards are end-to-end encrypted: the key lives in the link itself, so the server never sees your work. And it costs nothing.

Who it’s for

Prediction markets people, obviously. But really, anyone who has to turn data into a decision in front of other people.

napkinbookpredictionmarketsteachers& tutorsfinancialadvisorsdata & bizanalystspolicyanalystsanyonepresenting

What’s coming next

  1. Deeper live collaboration, and AI on the board
  2. An enterprise version on your own infrastructure
  3. A hosted version, so your boards follow you
  4. A semantic layer over your data
  5. More compute: a hosted database for when your data outgrows the browser
  6. Open source

The method behind the tool: base rates come first, a canvas reads as a forecast, and calibration is a practice.

Start a board.