Judgment instrument · the anchoring paradigm · Tversky & Kahneman 1974

I hand you a random number. Your estimate will look like it.

The first number you see drags your estimate after it — even when the number is arbitrary, even when you know it. You get 4 randomly drawn anchors, estimate on a slider, and the page shows the pull in your own figures: anchor, estimate and truth on one ruler.

Local
Random in URL
4 questions

Anchors are drawn at random and written into the page address — reloading keeps the same anchors. Answers never leave your browser. Cost: about 2 minutes.

Start — 2 minutes

Play: 4 rounds

Each round first shows the anchor ticket, you say whether the truth is more or less, then you give your figure on the slider.

Loading rounds…

Without JavaScript: the conclusion still stands — in the lab, estimates slide toward the anchor received, even an arbitrary one (Tversky & Kahneman 1974). Round truths: Danube 2,850 km; Everest 8,849 m; adult skeleton 206 bones; first crewed Moon landing 1969. Method and limits below, in text.

Why it happens

The mind never estimates from zero. It starts from the available anchor and adjusts insufficiently: you move away from the figure, but stop too early. That is why the anchor matters even when drawn at random in front of you.

Proof it is not ignorance: replay with flipped anchors and your estimates move with the anchor. Same mind, same knowledge — different starting figures, different landing figures.

Where to watch out

"Reduced from" prices, opening offers in negotiations, first figures thrown around in the news. The defending question: am I starting from facts, or from the first figure I heard?

Limit next to the claim: the page measures 4 general-knowledge estimates, not your intelligence nor your professional expertise. The index is an illustrative estimator, not a psychometric measure.

Truths on the table

The scoring keys, frozen 3 September 2026. Nothing hidden, everything checkable.

Question	Truth	Source
Length of the Danube	2,850 km	Britannica
Height of Everest	8,849 m	Britannica
Bones in the adult skeleton	206	Britannica
First crewed Moon landing	1969 (Apollo 11)	Britannica

Margins: the Danube's length varies with the source chosen (2,850–2,888 km); the "close" verdict allows ±10%. Bone counts vary individually; 206 is the adult standard.

Method, sources, limits

How it plays: each round randomly gets the high or low anchor (the ?a= address parameter, one bit per round). You answer comparatively, then estimate on the slider. The same URL anchors deal the same tickets on reload.

The index: per round, pull = (estimate − truth) / (anchor − truth), clamped to −50%…+150%; the index is the mean of 4, in percent. 0% ≈ neutral, 100% ≈ dragged all the way to the anchor. The "dragged toward" verdict means the estimate sits on the anchor's side of the truth.

The paradigm: Tversky & Kahneman (1974), Judgment under uncertainty: Heuristics and biases, Science (doi); the classic index in Jacowitz & Kahneman (1995). The lab is mechanism context; your figures are your proof.

Limits: 1) 4 rounds do not measure how smart you are. 2) The bank is general knowledge, not your trade. 3) Remembering the bank on replay inflates the score — which is why the flipped replay is part of the demonstration. 4) Anchors are stratified high/low by design; otherwise the pull would not show.

Reuse: CC BY 4.0. Cite "Marius Comper — The Anchor (freeze 3 Sept 2026)".