A market anomaly is an empirical regularity in asset prices that is inconsistent with the prevailing asset-pricing theory — most often a predictable pattern in returns that the efficient-market hypothesis (EMH) says should not survive once it is known. This section is the corpus's catalogue of the specific, named, academically-documented edges (post-earnings drift, momentum, value, low-volatility, accruals, calendar effects, and others), each treated as its own expert node. The core tension that runs through the entire section is interpretive and never fully resolved: every anomaly is simultaneously consistent with three different stories — (1) a genuine mispricing investors leave on the table because of behavioral biases plus limits to arbitrage; (2) a rational risk premium the test's benchmark model simply failed to capture (Fama's reading — it was never alpha, just an unmeasured beta); or (3) a data-snooping artifact that was never real. The same backtest is compatible with all three. That ambiguity is the subject, not a defect to be smoothed over.

What belongs here (and what doesn't)

This section covers cross-sectional and time-series return predictabilities documented in the academic literature — the things a quant calls "factors" or "anomalies" and a discretionary trader calls "edges." It deliberately overlaps with, but does not duplicate, the quant/factor-investing branch: there, factors are treated as portfolio construction inputs (how to build a momentum or value sleeve); here, each is treated as an anomaly — its discovery, mechanism, measured effect size, and decay. Pure indicator mechanics (RSI, moving averages) live in the technical-analysis branch and are not anomalies — most have weak or no standalone documented edge, and that distinction matters: do not let a charting tool borrow the academic credibility of a replicated factor.

A working taxonomy

The classic four-way split (Schwert 2003; standard finance texts) organizes the children:

• Calendar / seasonal — return depends on the date: turn-of-the-month, January effect, "Sell in May"/Halloween, weekend effect. Oldest class, most exposed to data-mining. → 014-calendar-and-seasonal-anomalies
• Cross-sectional return predictors (the "factors") — return depends on a firm characteristic: momentum (002), value & size (006), quality / profitability (007), accruals (005), the 52-week-high effect (012), low-volatility (003) and its cousin betting-against-beta (004), and the lottery / low-price effect (008).
• Event / corporate-action — abnormal returns around a discrete event: post-earnings-announcement drift (001), IPO underperformance (009), the index-inclusion effect (011).
• Microstructure / arbitrage-bound — apparent violations of the law of one price: closed-end-fund discounts (010) and the overnight-vs-intraday return split (013).

The capstone node 015-why-anomalies-persist-or-decay is not an anomaly — it is the meta-framework every other node inherits: the rules for deciding whether a documented edge is still live.

When it matters vs. when it doesn't

These nodes matter when the question is "is there an evidence-based reason to expect this kind of position to have a tailwind?" — they convert folklore ("buy strength," "cheap stocks win") into claims with measured effect sizes, sample periods, and named authors. They matter less, and can actively mislead, in three situations. First, most of these are diversified-portfolio, long/short effects documented across dozens-to-hundreds of names; they do not transfer 1:1 to a single concentrated discretionary trade (the most common misuse of the momentum and value literatures). Second, several live disproportionately in microcaps and illiquid names where the paper edge is eaten by transaction costs — Hou, Xue & Zhang (2020) found 96% of anomalies in their trading-frictions category failed to clear t > 1.96 once microcaps were mitigated via NYSE breakpoints and value-weighting. Third, a published number is a stale prior, not a live edge (see below).

Adoption, debate & evidence (the honest landscape)

The field is genuinely contested, and the contest is the expertise:

• Decay after publication is real. McLean & Pontiff (2016) found predictor returns roughly 26% lower out-of-sample and ~58% lower post-publication across ~97 predictors — about a third of the in-sample return attributable to publication/crowding itself. Schwert (2003) documented that the size, value, weekend, and dividend-yield effects weakened or disappeared after their founding papers.
• The "factor zoo" / replication crisis. Harvey, Liu & Zhu (2016) catalogued 300+ proposed factors and argued the usual t > 2 hurdle is far too lenient given the multiple testing (they propose t > ~3). Hou, Xue & Zhang (2020) replicated 452 anomalies and found ~65% could not clear even t > 1.96 once microcaps were handled with NYSE breakpoints and value-weighting (~82% at a multiple-testing hurdle); economic magnitudes of the survivors were much smaller than first reported.
• The counter-evidence. Replication is itself disputed: Chen & Zimmermann (open-source asset pricing) and Jensen, Kelly & Pedersen (2023, "Is There a Replication Crisis in Finance?") find high reproduction rates when methodology is applied consistently, arguing the "crisis" is overstated and partly a methodology choice. So the honest position is: a meaningful core of anomalies (momentum, value, quality, profitability, PEAD) replicates robustly; a long tail does not.
• A few are not seriously disputed. The existence of momentum (Jegadeesh-Titman) and PEAD (Ball-Brown lineage) is broadly accepted across both behavioral and rational camps; only their cause is debated.

How the section is used

Read any individual anomaly node through the 015 filter: identify whether it's risk-based or behavioral, how long ago it was published, whether it survived out-of-sample, how exposed it was to multiple testing, and whether it only works in tiny/illiquid names. Treat every reported effect size as decayed-until-proven and apply a publication haircut by default.

Sources

• Schwert, G. W. (2003), "Anomalies and Market Efficiency," Handbook of the Economics of Finance — taxonomy; size/value/weekend/dividend effects weakened after publication. https://www.nber.org/papers/w9277
• McLean, R. D. & Pontiff, J. (2016), "Does Academic Research Destroy Stock Return Predictability?" Journal of Finance 71(1): 5–32 — ~26% out-of-sample / ~58% post-publication decline. https://onlinelibrary.wiley.com/doi/abs/10.1111/jofi.12365
• Harvey, C. R., Liu, Y. & Zhu, H. (2016), "…and the Cross-Section of Expected Returns," Review of Financial Studies 29(1): 5–68 — 300+ factors; t > ~3 hurdle. https://www.nber.org/papers/w20592
• Hou, K., Xue, C. & Zhang, L. (2020), "Replicating Anomalies," Review of Financial Studies 33(5): 2019–2133 — 452 anomalies, ~65% fail t > 1.96 (~82% at multiple-test hurdle); microcap/value-weighting sensitivity. https://www.nber.org/system/files/working_papers/w23394/w23394.pdf
• Jensen, T. I., Kelly, B. & Pedersen, L. H. (2023), "Is There a Replication Crisis in Finance?" Journal of Finance — high reproduction rates; counter to HXZ. https://onlinelibrary.wiley.com/doi/full/10.1111/jofi.13249
• Sibling nodes (this section) — esp. 015-why-anomalies-persist-or-decay (the inherited discount framework), 002-momentum-anomaly, 001-post-earnings-announcement-drift-pead, 006-value-and-size-premia.