bet-evaluatelisted

# Size the bet Framing tells you what the decision is. Sizing tells you how much certainty to demand before you make it. The answer is rarely "as much as possible" — that is just a slower way of deciding, and delay has its own cost. ## First, size the response - **Trivial** — cheap to undo, small blast radius. Give the recommendation and the one thing that would change it. No table, no ceremony. - **Moderate** — classify reversibility, name the stakes, surface alternatives, give a bet size. - **Consequential** — everything below, including backcasting and the strongest counterargument, written to the bet file. ## Confidence, with its basis State how confident you are and immediately what that rests on. A confidence with no basis is decoration. Never invent a percentage. "70%" implies you could be calibrated at 70%, and for most decisions nothing supports that. Prefer plain words with the evidence attached: "low — this rests entirely on one customer conversation" says more than any number. Use a number only when something real backs it: a base rate, a measurement, a track record over enough similar cases to mean something. **Take the outside view.** Before reasoning from the specifics of this case, ask what usually happens in cases like it. Most migrations overrun. Most senior hires take longer to find than planned. Most features are used less than predicted. The inside view — the detailed story about why *this* time is different — is exactly the story everyone te