APEX factor · Value

Value factor — explained

Value bets that stocks trading cheaply relative to their fundamentals (book value, earnings, cash flow) outperform expensive ones over multi-year horizons. The mechanism is investor over-extrapolation: people pay a premium for recent winners and discount recent losers further than the fundamentals warrant.

Where this comes from

Academic anchor

Fama-French 1992 — The Cross-Section of Expected Stock Returns
Established that book-to-market ratio explains cross-sectional return variation independently of beta and size. Top-decile-cheap stocks beat top-decile-expensive by ~5% annually over 1963-1990. Lakonishok-Shleifer-Vishny 1994 sharpened the result by combining multiple value multiples (P/E, P/B, P/CF) and showed the premium comes specifically from glamour-stock disappointment, not value-stock surprise.
Plain English

What it actually measures

Imagine you're shown two companies: one earned $1 per share last year and trades at $10; another earned $1 and trades at $30. The market is telling you the second company will grow earnings faster than the first. Sometimes that's true — sometimes the cheap one is dying for a reason. But statistically, over decades and across countries, the market is too optimistic about the expensive cohort and too pessimistic about the cheap cohort. The difference between expectations and reality is the value premium. It plays out over years, not weeks.

No calibration constants

Math sketch

bm = log( book_value / market_cap )            // book-to-market
ep = log( trailing_12m_eps / price )            // earnings yield
cfp = log( trailing_12m_cf / market_cap )       // cash-flow yield
value = z_score( w₁·bm + w₂·ep + w₃·cfp )

Composite of three value multiples z-scored across the universe. The log transformation keeps cross-sectional comparisons meaningful when ratios span orders of magnitude (a stock at P/B 0.5 is just as far from the median as one at P/B 4 in log-space). Composite weights w₁, w₂, w₃ are not disclosed — public weighting would let competitors trade against our exact tilt. The academic blend itself is in LSV 1994.

Pipeline

How DeepVane implements it

Book value, EPS, and cash-flow figures come from SEC EDGAR XBRL filings — quarterly cadence, refreshed within hours of 10-Q / 10-K. Market cap is end-of-day. The composite refreshes daily at 06:00 UTC. We z-score within each sector cohort as well as across the full universe — sector-neutral value protects against the factor accidentally picking up an energy / financials tilt during sector rotations.

One coherent posterior

How it composes with APEX

Value pairs structurally with Quality (Novy-Marx 2013 explicitly argues quality + value works better than either alone) and with Momentum (Asness-Moskowitz-Pedersen 2013's value-and-momentum-everywhere). When all three fire bullish on the same ticker, the QUALITY COMPOUNDER pattern triggers. When value is high but quality is low and accruals are weak, the VALUE TRAP pattern fires bearish — Lakonishok-Shleifer-Vishny 1994's exact warning. Regime amplifier favours value in late-cycle and risk-off.

Honest limitations

When it fails

Value traps: a stock trading cheaply because its business model is permanently broken. We mitigate via the cross-product with Quality and Accruals — VALUE TRAP fires when value is high but quality is low. Second failure: long droughts. Value underperformed for ~12 years (2008-2020) before reverting. The reason isn't a broken factor; it's that growth stocks during ZIRP got priced for perfection. The 2022-2023 reversal showed value rebounded sharply once rates normalised. Position-sizing via the conformal prediction interval helps — we never go all-in on value, we size to the model's confidence.

Read next

Related factors

QualityMomentum

See Value score on a real ticker

Every ticker page shows the per-factor decomposition. The Value score is one of twelve composing the 0–100 APEX composite.

Try NVDA →Full methodology35 invariants live