The evidence

A survivorship-adjusted backtest of our own fair value model over 34 years, 1992 to 2026. Hypothetical and in-sample, with every caveat shown openly.

34years tested1992 to 2026

+5.2 ppedge per yearvs broad market, survivorship-adjusted

96,091stock-yearsin the test

2,172delisted stocksincluded (against survivorship bias)

Key findings

Over the full 34 years the strategy returned about 13.7%/yr (13.2% after a data-backed survivorship deduction) versus about 8%/yr for the broad market — a +5.2 pp/yr gap after the deduction, with the typical diversified ~15-stock portfolio beating the market in the large majority of simulations.
Dividends reinvested, real prices, point-in-time fair values (no look-ahead), losers cut by the trend/quality rules before they can rot.
Honest note: a buy-and-hold of the OVERVALUED, low-quality group (the opposite of the strategy) ended near a similar gross level over the full window. The strategy's measured edge is versus the broad market — not a promise that every undervalued pick beats every expensive stock.
This is a hypothetical in-sample backtest of our own model; a large part of the long-run return comes from broad, equal-weight diversification, and past results never guarantee the future.

Why this holds up

📅 34 years, not a lucky windowTested 1992→2026 — the longest period where our point-in-time fair values and clean prices both exist. Not a hand-picked stretch.

💸 Dividends reinvestedEvery return uses adjusted closes — dividends and splits are fully reinvested, on both our strategy and the broad market.

🔁 Always reinvestedWhen we sell, the proceeds roll straight into the next undervalued, high-quality pick. A real rolling portfolio, not buy-and-forget.

⏳ No look-aheadEach decision uses only data that was actually available that day. The future never leaks in.

🧹 Clean data onlyHigh-evidence stocks, ≥12 valuation models, and dense, gap-free price histories. Junk is excluded.

⚰️ Survivorship removedWe pulled 32,000 delisted stocks and MEASURED how much failed companies drag returns — then subtracted it. Most backtests skip this; we didn't.

🌍 Realistic benchmarkCompared to the realistic broad market (~8%/yr) — a broad index built from real share prices going all the way back to 1992, total return. A real price path with its true ups and downs, not a reconstruction.

Risk, honestly

yearly NAV, median of 100 diversified portfolios vs the realistic broad market.

	Strategy	Broad market
Volatility	23.6%	19.8%
Max drawdown	40.5%	41.6%
Worst year	-40.4%	-35.9%

When we hold, when we rotate

On track Below fair value, thesis intactPrice is under our Fair Value and the business is sound. The position stays until the gap closes.

Running Above fair value, up-trend intactReached fair value but still trending above its 300-day average. Nothing is flagged at fair value — the winner keeps running.

Review Up-trend brokeAbove fair value AND the price crosses below its 300-day average — the up-trend has turned. The position is flagged for review and rotated into the next undervalued pick.

Review Quality slipped below the medianQuality Score dropped below the median of the universe — the business is weakening. The position is flagged before it becomes a value trap.

Review Held 5 yearsAfter ~5 years a position is rotated into a fresh, more-undervalued opportunity — keeps the portfolio working on the best current discounts.

Survivorship bias, measured

Measured from a representative sample of delisted US stocks (held-to-last, capturing the collapse). The quality filter shrinks the drag because high-quality companies rarely fail.

Academic sources

Fama & French (1992, 1993) — value & size premia (survivorship-free CRSP).
Novy-Marx (2013) — gross profitability premium.
Piotroski (2000) — F-Score / quality of value stocks.
Asness, Frazzini & Pedersen — quality minus junk.
Bessembinder (2018) — most stocks underperform T-bills; a few winners drive all wealth.
Shumway (1997) — the delisting bias in stock returns.

Educational research, not financial advice — see the references above. This is a hypothetical, in-sample backtest of our own Fair Value model: point-in-time (no look-ahead), dividends reinvested, and survivorship-adjusted using real delisted stocks, benchmarked against the realistic broad market (~8%/yr) built from real share prices back to 1992 (total return). Because the strategy lets winners run, a large part of the return concentrates in a few large winners (the Bessembinder effect), so it carries high variance and is path-dependent — it reflects one 34-year history with value-typical down years, not a guaranteed forward return. Past performance does not indicate future results.