LEVEL vs CALIBRATE: check the claim, or check your own judgment

LEVEL checks a claim. CALIBRATE checks the person doing the checking. LEVEL, from The Evidence-Aware Life, is a five-step check of how far one claim can carry one decision. CALIBRATE, from The Calibrated Mind, is a nine-skill practice for finding out, over many judgments, whether your confidence matches what actually happens. Both books are by Len P. van der Hof and both are available now in English and Dutch. You do not need either book to use the frameworks; this page spells out both.

They are often confused because they sit so close together. Both are about evidence and confidence, and both even use the word “Lean”. But they answer different questions, and mixing them up leads to a common mistake: treating one good call as proof of good judgment.

What is the difference, in one table?

	LEVEL	CALIBRATE
The question	Can this claim carry this decision?	Does my confidence match what happens?
What you check	The claim and its evidence	Your own judgment, as an instrument
Unit of work	One claim, one decision	Many judgments over weeks and months
Needs outcomes?	No. It is done before you act	Yes, and many of them
What you end up with	A five-line record ending in a bounded next step	A prediction journal and a regular review of it
Steps	5	9 skills
Book	The Evidence-Aware Life	The Calibrated Mind

The short version: LEVEL asks whether you should trust the claim. CALIBRATE asks how far you should trust yourself.

What is LEVEL?

LEVEL is a five-step check for deciding how far to rely on one claim before you act on it.

1. Locate the claim: write it as one sentence someone could check.
2. Estimate the stakes: what it costs if you act and are wrong, if you ignore it and it was right, and if you wait.
3. Value the evidence: name what kind of support you hold, then check its quality, independence and relevance.
4. Expose the gap: name what is missing between that evidence and what this decision needs.
5. Lean accordingly: act only as far as evidence and stakes allow, with an owner, a limit and a review date.

LEVEL never asks you to wait for the outcome. It is finished when the record is written. The LEVEL framework page has the one-line version of each step.

What is CALIBRATE?

Calibration is the match between how sure you say you are and how often you turn out right. If you are well calibrated, then of all the things you call 70% likely, about 70% happen. CALIBRATE is the practice for getting there and staying there. Its nine skills:

1. Calibrate confidence: treat certainty as a measurement, not a feeling. Say how sure you are as a number you can check later.
2. Assess assumptions: write down the hidden premises your judgment rests on, before they surprise you.
3. Lean on probabilities: put numbers or ranges on words like “likely”, so one word stops hiding three different bets.
4. Investigate alternatives: test your view against the strongest competing explanation you can build.
5. Build updating rules: decide in advance how much a given result should move you, while you can still be honest about it.
6. Record and review: keep a prediction journal and compare it with what actually happened.
7. Act with appropriate uncertainty: you do not need certainty to act. Set a threshold based on what each kind of mistake would cost.
8. Teach and extend: explain your reasoning to someone else. It is the fastest way to find its holes.
9. Evolve your model: review regularly, notice where you drift, and correct it.

Notice the word collision. In LEVEL, “Lean” is the last step: act accordingly. In CALIBRATE, “Lean on probabilities” means putting numbers on your judgments. Same word, two jobs. The CALIBRATE framework page lists the skills with their one-line summaries.

One decision, two records: a worked example

Imagine your kitchen is being renovated. The builder says it will be finished in six weeks. You are thinking of hosting a family dinner for twelve on the Saturday after. This example is hypothetical.

The LEVEL record checks the builder’s claim before you decide about the dinner.

• Locate: “finished” means usable: water, power, worktop fitted, appliances connected. Not “cabinets on the wall”.
• Estimate: if you rely on the date and it slips, you cancel on twelve people at short notice. If you do not rely on it and it holds, you miss hosting. Moving the dinner early costs little; moving it late costs more.
• Value: the builder’s promise is an assertion from someone who wants to keep the job. Two online reviews saying “on time” are anecdotes. Your own last two renovations are the closest thing you have to a track record.
• Expose: the missing piece is whether the parts are ordered and have confirmed delivery dates.
• Lean: pencil in the dinner, but send no invitations until the end of week four. Ask the builder for the delivery confirmations this week. Owner: you. Decision point: end of week four.

That record is complete. It does not tell you whether your own sense of “six weeks sounds right” can be trusted. That is a different question.

The CALIBRATE record is one entry in a prediction journal. The fields below follow The Calibrated Mind’s format: six written now, two when the outcome is known.

Date:          Monday 28 September
Prediction:    The kitchen is usable (water, power, worktop,
               appliances) by Friday 6 November
Type:          binary (it happens or it does not)
Confidence:    60%
Rationale:     The schedule assumes the worktop arrives in week 3;
               my last two projects both ran late
Status:        open
-- filled in after 6 November --
Outcome:       1 if it happened, 0 if it did not
Process note:  one line on what you learned

Here is the uncomfortable part. Whatever happens on 6 November, this one entry tells you almost nothing about your judgment. A well-calibrated 60% call should miss about four times in ten, so a miss does not prove your 60% was wrong, and a hit does not prove it was right. Only a run of resolved forecasts shows whether your 60% behaves like 60%. That is why The Calibrated Mind advises logging many small, fast-resolving predictions (“this meeting will run over”) rather than a few grand ones.

How do the two fit together?

LEVEL feeds CALIBRATE. The evidence you weighed in Value and the gap you named in Expose are exactly what goes into the Rationale line. When you review your journal later, you can see which kinds of evidence you tend to over-trust.

CALIBRATE tells you how much to trust your LEVEL checks. If your journal shows that your 80% calls come true only about half the time, your confidence is running ahead of your evidence somewhere. Your Value step is a good first place to look, and until you find the leak, your Lean steps should be more cautious.

Stakes change the action, not the odds. Keep the same 60% forecast and change the dinner from twelve guests to two. The forecast is identical; the right action is not. CALIBRATE’s seventh skill makes this explicit: the threshold for acting depends on what each kind of mistake costs. LEVEL’s Estimate step says the same thing from the other side.

Which one should you use?

Your situation	Start with
A claim is in front of you and you must decide soon	LEVEL
You want to know whether your gut feeling can be trusted	CALIBRATE, starting with Record and review
A promise comes with a date that will resolve	Both: LEVEL now, a journal entry for later
Your team keeps saying “likely” and meaning different things	CALIBRATE, Lean on probabilities
You are about to make a big call you cannot undo	LEVEL, then CALIBRATE’s Act with appropriate uncertainty

What are the common mix-ups?

“LEVEL is the quick one, CALIBRATE is the slow one.” Not quite. A LEVEL check on a large decision can take weeks of digging. CALIBRATE’s threshold skill can shape a decision today. The difference is what you check, not the clock.

“I called it right, so I am calibrated.” One outcome cannot confirm or refute a probability. Calibration only shows up across many resolved forecasts.

“This matters a lot, so it is less likely to go wrong.” Stakes raise the bar for acting. They do not move the probability. The same holds in reverse: a deadline does not make thin evidence stronger.

“Using the words means I am doing it.” Both are authorial frameworks. Knowing the letters proves nothing; the records do the work. What calibration means makes the same point about confidence as a report rather than a feeling.

Try this today (15 minutes)

1. Pick one decision this week that depends on someone else’s promise with a date: a delivery, a hire’s start date, a repair.
2. Write the five LEVEL lines (5 minutes).
3. Write one journal entry for the same promise: date, prediction with its resolution date, type, confidence, one-line rationale, status (5 minutes).
4. Put the resolution date in your calendar. When it arrives, fill in the outcome and one line on what you learned (5 minutes).

Then keep going with small, quick forecasts, and wait until a few dozen have resolved before you draw conclusions about yourself. For bigger calls there is the five-line decision journal, which you write before you act for exactly this reason.

Cite this:LEVEL vs CALIBRATE: check the claim, or check your own judgment.Len P. van der Hof. https://lenvanderhof.com/en/blog/level-vs-calibrate/ · Published 26 September 2026.