Anchoring

Definition

Anchoring is insufficient adjustment away from an initial reference number — estimates gravitate toward whatever value was considered first, even when that value is arbitrary (Tversky-Kahneman 1974). Markets are dense with candidate anchors — entry prices, round numbers, 52-week highs, IPO prices, analyst targets — and the documented result is that some of them carry real predictive structure precisely BECAUSE the crowd anchors: George-Hwang (2004) showed nearness to the 52-week high predicted returns as well as conventional momentum.

How it works / structure

• The mechanism: adjustment from a starting value stops too early; the anchor need not be relevant (experimentally, random numbers anchor estimates) — in markets, salient prices become “fair” by exposure.
• The documented market anchors: the 52-week high (George-Hwang — traders under-react to news near the high because the anchor says “already expensive”; the drift continues through it); analyst estimate strings (sent-analyst-revisions — under-adjustment institutionalized); entry-price anchoring (the disposition effect’s reference point — bias-disposition-effect); round-number clustering in orders and strikes (documented in order-book data; part of why pivot/level lore self-fulfills weakly — indicator-pivot-points).
• The two-sided platform read: as a BIAS, anchoring is an error to engineer out of estimates (blind re-estimation before seeing prior values; falsifiers stated in market terms, not entry-relative terms); as a PHENOMENON, crowd anchors are measurable structure (distance-from-52-week- high as a documented factor input — strategy-momentum variant).
• Engine-executable forms: anchor-distance features (52w-high proximity, round-number distance), and anchor-hygiene rules (thesis targets derived from valuation/vol frames, never from “it was at X last month”).

When it applies

Estimate hygiene everywhere (price targets, “cheap vs its high” reasoning — the high is an anchor, not a valuation); exploiting crowd anchors as features (the George-Hwang factor family); exit design (entry-relative exits inherit the entry anchor — R-multiples make it explicit and disciplined rather than hidden).

Risk profile & failure modes

• “Down 50% so it’s cheap”: the prior price is not a valuation — the classic anchor error; falling-knife cohorts are full of stocks down 50% that went down another 50%.
• Target stickiness: theses re-anchored to stale targets after fundamentals change; the platform requires target re-derivation on thesis-relevant news, not adjustment from the old number.
• Feature-vs-error confusion: using anchor structure as a feature while committing the error in one’s own estimates — the two uses need separate machinery.
• Round-number crowding: stops clustered at obvious levels are documented liquidity targets (mgmt-stop-loss placement discipline).

Evidence & limits

Anchoring is foundational heuristics-and-biases psychology (1974), heavily replicated; George-Hwang (2004) is the peer-reviewed market expression with successor literature. Round-number effects are documented in microstructure data at modest effect sizes. The platform separates the psychology (engineer it out) from the factor (replay it in).

Falsifiable-thesis examples

Illustrations only, not signals:

• “The near-52-week-high decile (quality-filtered) will outperform the far decile over the next 6 months (George-Hwang thesis)” — falsified by the cohort pair.
• “Re-derived price targets on thesis-relevant news will differ from adjusted-from-prior targets by more than 10% in a third of cases (anchor-drag measurement)” — falsified by the paired estimates.

Cross-references

• The institutional form: sent-analyst-revisions (under-adjustment strings)
• The reference-point sibling: bias-disposition-effect (entry anchoring)
• The exploitable structure: strategy-momentum (52w-high variant), indicator-pivot-points (self-fulfilling levels)
• The hygiene frame: qualitative-analysis (falsifiers in market terms)