Decision tool

Your "90% sure" should come true nine times out of ten. Usually it doesn't.

Being well-calibrated means your confidence matches reality: the things you're 90% sure of happen about 90% of the time. Almost everyone is overconfident — our 90% ranges hold the truth far less than 90% of the time, because a narrow, precise-sounding range feels smarter than an honest one. The good news is that calibration is measurable, and it's trainable. This is a game to measure yours, then a dial to see why honest beats narrow.

01 — The game

Can you draw a range you're 90% sure of?

Eight numbers nobody carries in their head. For each one, enter a low and a high bound so that you're 90% sure the true value sits between them — wide enough to be safe, tight enough to still mean something. Reveal each answer and watch how many of your 90% ranges actually hold.

Question 1 of 8 Ranges hit: 0 of 0
 

Set a low and high bound you're 90% sure contains the answer. Aim to be wrong only about 1 time in 10.

km
to
km
0 / 8

02 — The width dial

Narrow feels smart. Honest is wide enough to be right.

Here's the same estimate made forty times over — each dot is one occasion you gave a 90% range. Drag the dial to change how wide you make that range, and watch two things pull against each other: how often you're right, and how much the range actually tells anyone. The goal isn't the narrowest range. It's the honest one.

narrow

03 — How it's different

Calibration isn't accuracy, and it isn't confidence.

Three things get muddled together whenever someone says "I'm sure." Pull them apart and the overconfidence bias — that steady pull toward narrow, precise-sounding ranges — stops fooling you.

Lens What it measures The question it answers Where it fails you
Accuracy Whether a single answer turned out right. "Was I correct this time?" One hit or miss says almost nothing, and it ignores how sure you were when you said it.
Raw confidence How sure you feel. "How certain am I?" Feelings aren't evidence. Confidence runs ahead of correctness — that's the overconfidence bias.
Calibration Whether your confidence matches reality across many judgements. "When I say 90%, is it true 90% of the time?" You can't feel it from the inside. It has to be measured — and then it can be trained.

Not the same as the Forecast Cone

The Forecast Cone is about uncertainty widening over time — a range that fans out the further ahead you project. The Calibration Game is about a different question: is the confidence you're stating right now honest at all? A cone can be beautifully drawn and still be dishonestly narrow at every point along it.

Reach for the cone when you're projecting forward and need to show uncertainty growing. Reach for calibration when you want to know whether any stated range — today's or next year's — can be trusted. Compare with the Forecast Cone →

Overconfident

The narrow range

Two percent either side of a number you can't actually pin down. It looks rigorous and sounds authoritative — and it's wrong far more than one time in ten.

Right ~55% of the timeFeels precise
Calibrated

The honest range

Wider, and it admits what you don't know. Less impressive in the room — but right about nine times in ten, so people can actually plan on it.

Right ~90% of the timeActually useful
04 — Where it bites

False precision hides risk. Honest ranges surface it.

A single confident number looks decisive and quietly buries everything you don't know. The same judgement, stated as a calibrated range, tells people the truth about how sure you really are. Four places it changes the decision.

Delivery

Project time & cost

Instead of"It'll take six weeks."
Say"Eight to fourteen weeks, 90% confident — and here's what would widen it."
Finance

Revenue forecast

Instead of"We'll do $4.2M next quarter."
Say"$3.1M to $5.0M, and I'd bet nine to one on that range holding."
Engineering

The design review

Instead of"This will definitely scale."
Say"I'm ~70% sure it scales to 10x — here's the test that would make me sure."
Risk

Risk assessment

Instead of"That can't happen."
Say"Low, but not zero — call it 1 in 20 a year, and here's the blast radius if it does."
05 — Go deeper

Calibration is a habit you can build.

The idea is simple; living it is the work — catching yourself mid-sentence about to give a number you can't defend, and widening it on purpose. Teams that estimate in honest ranges plan better, argue less about false precision, and get blindsided far less often. That's the work we do with the leaders and teams we coach: turning gut confidence into judgement you can actually bet on.

Talk to us about coaching
TAKE ONE

Your 90% should be right 90% of the time

That's the whole test. If the things you're 90% sure of come true far less often, your confidence is inflated — and now you can see by how much.

TAKE TWO

Widen the range until it's honest

Overconfidence is the default, so the fix has a direction: almost always, outward. Stretch your bounds until being wrong would genuinely surprise you.

TAKE THREE

Confidence is not accuracy

How sure you feel and how often you're right are different numbers. Calibration is the discipline of closing the gap between them — and it improves with feedback.