New highs / new lows

Definition

The new highs/new lows indicator counts issues making fresh 52-week highs versus fresh 52-week lows across an exchange or index each day. It is a breadth gauge with a distinct personality: unlike advance/decline (which counts every small up-day equally), NH/NL only registers stocks at yearly EXTREMES — so it measures how broadly trends are strong enough to break out, and how broadly damage is severe enough to break down. Its two best-known uses are practitioner canon (labeled as such): expanding new lows while an index holds up as internal-deterioration warning, and the near-total DISAPPEARANCE of new lows as a bear-market-end condition (Zweig).

How it works / structure

• Computation forms: raw daily counts; net (NH − NL); cumulative net (running sum, trend-followable); percentage of issues (normalizes across eras — raw counts drift with listing counts, an important hygiene point); 10-day averages to de-noise.
• The asymmetric readings (practitioner, labeled): bull-market tops historically show new highs CONTRACTING for months while indices grind up on narrowing leadership (1999-2000, 2007, 2021 are the standard exhibits); bear-market bottoms show new lows exploding to washout extremes (>40% of issues) then contracting sharply even as indices retest — Zweig’s documented condition: retests on FEWER new lows precede durable turns.
• Why extremes carry information: a 52-week high requires a completed base and absorbed overhead supply (strategy-breakout logic aggregated); a 52-week low means every holder of the past year is underwater — the counts aggregate meaningful per-stock states, not daily noise.
• Cross-checks (engine-relevant): NH/NL divergence
  • indicator-breadth-advance-decline divergence + widening indicator-credit-spreads is the compound distribution warning; NH/NL washout + opt-term-structure inversion is the compound capitulation read.

When it applies

Regime and health dashboards (the narrowing-leadership check on index strength); bear-market-bottom process identification (the retest-on-fewer-lows pattern); momentum-universe health (strategy-momentum breadth of trend participation); exchange-level application needs issue-mix hygiene (NYSE counts include funds and preferreds — documented distortion, filter or use common-stock-only versions).

Risk profile & failure modes

• Raw-count era drift: listing counts and index composition change — thresholds in raw counts decay; use percentages (quant-data-hygiene).
• Divergence duration: leadership can narrow for QUARTERS before consequences (1999’s divergence ran a year) — the signal is condition, not trigger, and shorting on it has documented failure history.
• Issue-mix contamination: rate moves can flood NYSE new lows via bond funds/preferreds without equity information (the classic distortion) — uncurated counts mislead in rate regimes (2022’s counts needed this filter).
• Evidence tier honesty: formal academic support is thin relative to A/D-line literature — the KB carries NH/NL as practitioner canon with strong episodic exhibits, labeled accordingly.

Evidence & limits

Zweig (1986) documents the new-lows-contraction condition; the divergence exhibits (2000, 2007, 2021) are public record. Formal out-of-sample studies are limited; parameter conventions (10-day smoothing, percentage thresholds) are practitioner defaults, not optimized constants.

Falsifiable-thesis examples

Illustrations only, not signals:

• “Index highs with new highs below half their prior-peak count precede flat-or-negative 6-month returns (narrowing-leadership thesis)” — falsified by the conditional return distribution.
• “Bear-market retests on 50%+ fewer new lows than the initial low mark durable bottoms within a quarter (Zweig condition check)” — falsified by a failed retest cohort.

Cross-references

• The breadth siblings: indicator-breadth-advance-decline, indicator-mcclellan-oscillator
• The per-stock logic it aggregates: strategy-breakout, strategy-momentum
• The compound-signal partners: indicator-credit-spreads, regime-volatility