I buy shit companies cheap, with a reasonable chance they become less shitty. That's the pitch. This page is the audit of the pitch.
Fair warning about what you'll find here. There is no study proving that Tangible Bargains works. There's a large body of research supporting the ideas it's built from, a smaller body of research attacking those same ideas, and a live portfolio that will settle the argument the only way arguments like this get settled. I've put all three on this page, because a research process that only cites the studies agreeing with it isn't research, it's marketing with citations.
If you read one sentence on this page, read this one: the evidence supports the architecture of this strategy much more strongly than it supports the calibration. Buying cheap has a century of data behind it. Screening cheap companies for survival has real support. But no paper on earth says 0.50x tangible book is a magic number, or that my 15% averaging-down rule is optimal, or that an F-Score of 4 is the correct cutoff. Those are house rules. They exist to impose discipline, and the honest way to test them is the way this site tests them: publish the rules before the results, then let the results accumulate in public.
Now the evidence. Both directions.
Part one: the case for
Cheap has beaten expensive for as long as anyone has measured it
The foundation under everything else here is the most studied result in empirical finance. Fama and French (1992) sorted US stocks on book-to-market equity from 1963 to 1990 and found the cheapest decile earned an average 1.83% per month against 0.30% for the most expensive decile. Not a rounding error. A chasm. Their conclusion was that size and book-to-market, not beta, did the real work of explaining average returns.
Two years later, Lakonishok, Shleifer and Vishny (1994) asked the more interesting question: why? Sorting on book-to-market over 1968 to 1990, value stocks returned about 19.8% a year over the five years after formation against 9.3% for glamour stocks, and the gap survived size adjustment. Their answer to "why" matters more to me than the gap itself: investors extrapolate. They pay up for companies with beautiful recent histories and dump companies with ugly ones, past the point either deserves. Value strategies profit from that mistake, and in their tests the value stocks were not measurably riskier. They actually held up better in bad markets.
That's not just an American artifact. Fama and French (2012) found value premiums across Europe, Japan and Asia-Pacific from 1989 to 2011, largest among small stocks nearly everywhere. (Honest footnote: in that particular sample the North American value premium was positive but not statistically significant. The international evidence is the stronger leg, which suits a portfolio screening 33 countries just fine.) Asness, Moskowitz and Pedersen (2013) found value premia in every one of eight markets and asset classes they tested. And if you want the practitioner's compilation, Tweedy, Browne's What Has Worked in Investing collects over fifty studies of cheap-asset and cheap-earnings strategies, roughly half from outside the US. Yes, that Tweedy Browne: the firm in whose office Walter Schloss sub-leased his famous desk space. More on him on the Schloss page.
The behavioral logic is the part I actually rely on. De Bondt and Thaler (1985) found that the market's biggest three-year losers went on to beat its biggest winners by about 25% over the following three years, and the losers were less volatile. Pessimism overshoots. I'm in the business of buying the overshoot, with rules for telling overshoot apart from accurate diagnosis.
One more supporting beam. My matrix prices companies on normalized earnings, a multi-year average, rather than last year's print. Campbell and Shiller (1988) showed why that's not just a stylistic preference: a thirty-year average of real earnings against price explained over half the variance of ten-year returns, while single-year figures explained almost nothing at long horizons. Averages carry information single years don't.
The deeper the discount, the weirder (and better) the evidence gets
Tangible Bargains lives at the extreme end of cheap: discounts to tangible assets. The academic record down here is thinner but startling.
The classic is Oppenheimer (1986), who tested Ben Graham's net-net rule (buy below two-thirds of net current asset value) from 1970 to 1983. The net-net portfolios averaged 2.45% per month, roughly 29% a year on an arithmetic basis, against about 11.5% for the NYSE-AMEX index. Ten thousand dollars became roughly $255,000 in thirteen years. Xiao and Arnold (2008) ran the same idea on the London Stock Exchange over 1981 to 2005: raw first-year returns above 31%, and an 11.3% annual premium that survived even after adjusting for the small-size effect. Montier (2008) took it global across developed markets from 1985 to 2007 and got net-net returns over 35% a year against 17% for the comparable universe.
And it's not purely ancient history. Mohanty and Oxman (2026), published this year in the Review of Financial Economics, tested US net-nets from 1969 to 2019: a risk-adjusted alpha of about 1.09% per month, near 13.9% a year, after controlling for the standard factor zoo. The same paper found the edge weakened over 2004 to 2019. I cite it anyway, and I cite that second finding on purpose. A source that reports the decay alongside the alpha is worth ten that don't.
Before you get excited: every one of those net-net studies has serious implementation problems, and I've listed them in part three. The clean takeaway is narrower. Extreme discounts to conservative balance-sheet value have carried real information in every decade and every market where someone bothered to look. What they haven't carried is a guarantee you could trade them at the printed prices.
Cheap plus alive beats cheap alone
This is the core design decision of the whole methodology, and it has the most direct academic support of anything on this page.
Piotroski (2000) started from the same place I do: the cheapest quintile of the market, where most academics stopped looking. Then he asked whether nine boring accounting signals could separate the cheap stocks that recover from the cheap stocks that die. They could. Within high book-to-market stocks from 1976 to 1996, the financially strong names earned a 13.4% average market-adjusted return against 5.9% for the cheap universe as a whole, and a long-short version returned about 23% a year. The detail that matters most to me: fewer than 44% of cheap stocks earned positive market-adjusted returns at all. Deep value is a population with a fat left tail, and the score's real job was cutting that tail off.
That is exactly, and only, the job I give it. The F-Score here is a survival veto, not a quality ranker. A full page covers the score, the four changes I made to the textbook formula, and the case against it: the Piotroski F-Score page.
The follow-up work strengthens the logic. Piotroski and So (2012) found the value premium concentrates almost entirely where price and fundamentals disagree: value spreads were roughly 22.6% a year where the market's pessimism contradicted the accounting evidence, and roughly zero where the pessimism was justified. Cheapness pays when the fundamentals say the market is wrong. Walkshäusl (2020) tested the score across 20 developed and 15 emerging markets from 2000 to 2018, out of sample in both time and geography, and found it worked broadly: high-minus-low F-Score spreads around 0.79% per month in developed markets and 0.95% in emerging ones, surviving the standard factor controls. For a portfolio screening obscure names in 33 countries, an out-of-sample international replication is worth more than another US backtest.
The mirror image comes from the distress literature. Campbell, Hilscher and Szilagyi (2008) found that the stocks most likely to fail didn't compensate anyone for the risk: the most distressed names delivered anomalously low returns, with the riskiest slice losing double digits a year while looking statistically "cheap" the whole way down. Cheap and dying is not a bargain. It's a value trap with a ratio attached. That finding is the reason leverage, liquidity and the F-Score all get checked before the cheapness matrix ever sees a candidate.
Getting paid while you wait is measurable
The methodology promotes and demotes names based on shareholder yield, dividends plus buybacks, with dilution tracked separately. Boudoukh, Michaely, Richardson and Roberts (2007) is the reason it's built that way: measures of total payout predicted returns far better than dividend yield alone, and net payout (subtract the share issuance) was better still, explaining 26% of return variance at the annual horizon where dividend yield alone explained about 6%. A company paying a 5% dividend while quietly issuing 5% more shares hasn't paid its owners anything. The research agrees.
The governance discount is real, especially where I fish
A dollar of book value is not worth a dollar to you if the person controlling the company can route it somewhere else. This has three of the strongest papers on the page behind it. La Porta, Lopez-de-Silanes, Shleifer and Vishny (2002) found companies in countries with stronger minority-shareholder protection carry systematically higher valuations. Claessens, Djankov, Fan and Lang (2002) found, across 1,301 East Asian companies, that value falls when a controller's voting power exceeds his actual cash-flow ownership. Djankov, La Porta, Lopez-de-Silanes and Shleifer (2008) built an anti-self-dealing index across 72 countries and found it predicts stock-market outcomes. For a global microcap strategy full of controlled companies and state-adjacent names, this literature is why the capital-allocation red-flag check is a separate gate from cheapness, not a footnote to it.
Why a hundred names instead of ten good ideas
Bessembinder (2018) measured every US stock from 1926 to 2016 and found that 57.4% of them underperformed one-month Treasury bills over their lifetimes. The single most common lifetime outcome for an individual stock was a total loss. All of the market's roughly $35 trillion in net wealth creation came from about 4% of companies. His own stated implication: this is why concentrated portfolios usually lose to the index.
Read that back against a strategy that deliberately buys disliked companies, and the conclusion writes itself. Individual outcomes here are wildly skewed, some of these companies deserve their obituaries, and I don't know in advance which ones. Wide diversification isn't a hedge on the strategy. It is the strategy. Schloss held over 100 names for the same reason.
Why the portfolio moves slowly
One more, because it justifies a structural choice most people never notice. Novy-Marx and Velikov (2016) studied what trading costs do to published anomalies and found the survivors are the low-turnover ones, with one specific trick standing out: making the bar for buying higher than the bar for continuing to hold, which cut turnover and costs by roughly 40% in their tests. That is structurally what this portfolio does. Entering requires clearing the full gauntlet; holding only requires the thesis staying intact; adding requires the 15%-cheaper-on-the-ratio rule. In illiquid microcaps, the cheapest thing you can do is nothing.
Part two: has anyone run this in public? (The AAII question)
Academic backtests are one thing. I wanted to know whether anyone has tracked strategies like this in public, in something close to real time, for decades. The American Association of Individual Investors has, twice over, and the results deserve an honest reading in both directions.
The screen. AAII has tracked a Piotroski High F-Score screen (F-Score 8 or 9 inside the cheapest 20% of the market by price-to-book, which is a close cousin of what I do) in its stable of stock screens since 1998. As of April 2024 it showed a 14.0% annualized price gain since 1998 against 6.1% for the S&P 500, and at its peak, as of late 2013, it was AAII's best performer of all tracked screens at 31.7% annualized. In 2008 it was the only one of 56 screens that finished positive, up 32.6% while the S&P lost 40%.
Now the caveats, which are large. AAII's own methodology page says screen returns ignore commissions, spreads, dividends and slippage, assume frictionless monthly rebalancing, and are "unachievable even in a best-case scenario." The Piotroski screen averaged about four holdings and spent twenty separate months holding nothing at all, including most of late 2008, meaning part of its famous bear-market win was the win of sitting empty. Its max drawdown was 84.8%, the deepest of any AAII screen. And do the arithmetic on the two endpoint figures: going from 31.7% annualized (1998 to 2013) to 14.0% annualized (1998 to 2024) implies the screen roughly lost money for a decade after 2013. That decade happens to be the value winter covered in part three, but decay is decay and it belongs on this page.
The real-money version. More interesting to me than any screen: AAII's Model Shadow Stock Portfolio, an actual-dollars microcap deep-value portfolio running since 1993. Rules a Graham disciple would recognize: microcap size bands, price-to-book near or below 0.9, positive earnings, no financials, no ADRs, quarterly maintenance. As of March 31, 2026 it showed a 13.6% compound annual return since inception against 10.4% for the Vanguard 500 index fund, with real money paying real spreads. The ride was violent (down 50.8% in 2008, calendar-year swings from +73% to -50%), which is what a microcap value basket honestly looks like. That's the closest thing that exists to a decades-long public, real-money test of small, cheap and ugly, and it beat the market while being far bumpier than the market.
The referees. Because I'd rather you hear it from critics than from me: Schadler and Cotten (2008) audited AAII's screens and found that while 91% beat the S&P on paper, only about a third did so significantly once transaction costs entered. North and Stevens (2015) reran the audit through 2011 and found the Piotroski screen ranked first of 56 on alpha, but with realistic costs on a $50,000 account most screens' edges shrank drastically. CXO Advisory's tally through 2018 found the Piotroski screen had the highest net annualized return of all sixty screens (25.4% under their cost assumptions), and also found screen outperformance tends to deteriorate over time. And Alpha Architect independently rebuilt thirteen AAII value strategies over 1963 to 2013 in a liquid universe with delisting data: the F-Score strategy came first at 16.74% a year. Their conclusion was that the simple value models did as well as the sophisticated ones, which any Schloss reader could have told them, but it's nice when the data agrees.
So: does AAII show deep value working? Yes, across 25 to 30 years, in both hypothetical and real-money form. Does it show anything like the printed screen numbers reaching an actual account? No, and AAII says so itself. Both halves are true. Plan on the second one.
Part three: the case against
This section isn't decorative, and it isn't here so I can knock down strawmen. Some of these criticisms have changed how this portfolio operates. All of them could turn out to matter more than I think.
Book value measures less than it used to
The strongest structural criticism of any book-value strategy: accounting hasn't kept up with the economy. R&D, software, brands and organization-building are mostly expensed, not capitalized, so the book value of intangible-heavy companies understates their real capital. Lev and Srivastava (2022) argue this is a core reason book-to-market value investing broke down, measuring its average return over 2007 to 2018 at 0.65% a year, statistically zero, and showing that cheap companies also stayed trapped in cheapness about a third longer after 2007. Eisfeldt, Kim and Papanikolaou (2022) built a value factor with intangible capital added back and found it beat the traditional one, including in the years traditional value failed. Research Affiliates' work estimates intangibles have gone from about 30% of the average firm's tangible book value in the 1960s to roughly 100% now. (For fairness even to the critique: Dimensional's counter-research finds intangible estimates too noisy to reliably improve value strategies. The fix is contested too.)
I think this criticism is correct, and it's the reason this strategy is shaped the way it is rather than a reason to abandon it. Tangible book value is a terrible universal yardstick and a perfectly good specialized one. You will not find me valuing a software company on P/TBV. The screen deliberately fishes where tangible assets still mean something: manufacturers, shippers, distributors, holding companies, businesses whose balance sheets contain things you could theoretically sell. Notice also where the intangibles critique bites hardest: it's an argument about book value missing unrecorded capital. A strategy that buys recorded, tangible capital at a discount, in asset-heavy industries, is standing in the corner of the market where the critique has the least to say. That's by design, not luck.
Value spent thirteen years getting beaten up, and I have to assume it can happen again
From December 2006 to June 2020 the standard academic value factor suffered a drawdown of 54.8%, its worst ever, per Arnott, Harvey, Kalesnik and Linnainmaa. The small-cap version of the pain was grotesque: Blitz and Hanauer measured small value at -13% cumulative from January 2017 through August 2020 while small growth gained 71%. Thirteen years is longer than most investors' careers, let alone their patience.
The rebuttal camp has real evidence too. The Arnott paper decomposed the loss and found more than all of it came from value getting relatively cheaper, not from the underlying businesses failing; Asness at the 2020 trough measured value spreads at their 100th percentile ever. What happened next fit that reading: a violent comeback in 2021 and 2022 (MSCI world value beat world growth by 21 points in 2022, the widest gap since 2000), another growth-led stretch in the US during the 2023 to 2024 AI run, and then a result that matters a lot to a strategy like this one: international value had its best year in over 25 years in 2025, with MSCI EAFE Value up 33.8% through October. As of the start of 2026, AQR's assessment is that value spreads have normalized from historically extreme to merely healthy. Meanwhile, one for the skeptics' side of the ledger: by 2025, US small value's twenty-year record against the S&P 500 had gone negative for the first time.
What I take from all of it: the value premium is not dead, it is episodic, its worst episodes last years, and nothing entitles me to skip the next one. The portfolio is built for that (diversification, payout while waiting, no leverage), but built-for-it and immune are different words.
The F-Score has real critics, and some of them land
The F-Score page engages the strongest critique in detail, so here's the strategy-level summary of the ones that matter most, in descending order of how directly they hit my use of it.
Yuval Taylor's replication at Portfolio123 rebuilt Piotroski's exact test inside the value universe for 1999 to 2020 and found the ranking inverted: low-F-Score value stocks returned about 9.4% a year while high scorers returned about 4.3%, and the long-short version lost money for two decades. That's not a critique of some other use of the score. It's a within-value test, my use case, and it's the single best reason not to treat the score as gospel. (Note what it doesn't show: the high scorers still made money long-only. What failed out of sample was the score's power to rank, which is part of why I use it only as a floor.) Anderson, Chowdhury and Uddin (2024) found the score is partly a macro thermometer: in contractions, economy-wide conditions matter about five times more to a company's F-Score than in expansions, so a fixed cutoff quietly tightens and loosens with the cycle. Hyde (2018) found in Australia that the score's paper returns mostly vanished under factor adjustment except in small caps, driven by the hard-to-short side. Hou, Xue and Zhang found the score fails replication entirely under microcap-suppressing protocols, and Novy-Marx and Velikov found its market-wide net-of-costs edge indistinguishable from zero. And the general force pushing on every published signal: McLean and Pontiff (2016) measured anomaly returns dropping 26% out of sample and 58% after publication. The F-Score was published in 2000. Assume the 2000-era numbers are the ceiling, not the base case.
Why do I still use it? Because every one of those results attacks the score as a return generator, a ranker, or a market-wide factor, and I ask it to do none of those jobs. It's a veto against the specific left-tail failure documented by Piotroski himself and by the distress literature. The four modifications described on the F-Score page exist because of these critiques, not in spite of them. But if the day comes when the evidence says the veto itself adds nothing, this page is where you'll read it.
Published edges shrink, and backtests flatter everyone
Beyond McLean and Pontiff: Hou, Xue and Zhang reran 452 published anomalies under conservative protocols and 65% failed outright, 82% under stricter thresholds, with the failures concentrated exactly where the flashy numbers came from, which is microcaps: 60.7% of listed companies, 3.2% of market cap. Linnainmaa and Roberts (2018) took accounting anomalies back into pre-publication decades and watched their power fall 50 to 70%. The uncomfortable implication for me: the part of the market I fish in is precisely where paper alpha inflates easiest. The equally true flip side: it's also the one part of the market where a small individual account has a structural edge over any institution, because the alpha that exists there is too small in dollars for professionals to harvest. Both things are true. This site is a live experiment in which one dominates.
Paper portfolios trade for free. I don't.
Every backtest in part one assumed fills at printed prices. In actual microcaps, spreads run to percents, size moves prices, and some of the cheapest names on any screen cannot be bought in meaningful quantity at all. The net-net literature has a specific version of this problem: Alpha Architect's review of the classic studies notes Montier's 35% figure is an arithmetic mean (compound results would be materially lower), that Oppenheimer's sample had a median market cap around $4 million, and that the smallest names in these studies were, in their words, virtually untradeable. Add delisting bias, where databases historically recorded dying companies' final returns too kindly, and the honest conclusion is that every historical deep-value number you've read on this page overstates what an investor actually captured. It's why this site publishes actual fills, actual spreads and actual broker statements instead of a model portfolio. Trust the broker, not the blogger.
Sometimes cheap is just the right price
The last criticism is the oldest one. A discount to tangible book is an opinion about mispricing, and the market's counter-opinion is sometimes correct: the assets are obsolete, the receivables are fiction, the controller will never let a minority shareholder see a dollar, the business burns the book value before anyone arrives to unlock it. The distress literature quantified how expensive that mistake is. I run TBV-erosion limits, the survival veto, and the governance gate specifically against it, and I still expect to own value traps. The strategy doesn't promise to avoid them. It promises to keep them small, diversified and honestly logged when they happen.
Part four: what none of this proves
Time to say the quiet part in bold.
No academic study has tested this exact system. Not the P/TBV-by-normalized-P/E matrix, not the 0.50x and 0.65x band edges, not the 15% add rule, not the roughly 10% TBV-erosion limit, not the F-Score-4 cutoff, not the sell range read off each stock's own ten-year history, not the sizing tiers, not the caps. Those numbers are house policy: internally consistent, built from the evidence above, and unvalidated as a package. Anyone who tells you their bespoke rule set is "proven by research" is selling something. The research proves the raw materials. The construction is mine.
Which is exactly why the site works the way it does. The rules are published in full, they get frozen when the first live trade posts, changes get dated and logged from then on, and the portfolio page reconciles against actual broker statements. The evidence on this page is why I believe the bet is worth making. The live record is the only thing that will ever show whether I built the machine right. If this page ever starts reading like a victory lap, or the shelf of evidence against stops growing, stop trusting this site.
The Library
Everything cited above plus the primary sources behind the rest of the site, shelved so you can browse. Free full-text links where they exist; paywalled journal pages otherwise. If a link rots, the citation has enough detail to find it again.
Shelf one: the practitioners
Schloss, Walter J. (1994). "Factors needed to make money in the stock market." The one-page memo, sixteen rules. Full text, and this site's rule-by-rule implementation mapping.
Buffett, Warren E. (1984). "The Superinvestors of Graham-and-Doddsville." Hermes, Columbia Business School. The founding document for "this corner of the market works": nine Graham-school records, including Schloss's 28-year run. Full text at Columbia.
Buffett, Warren E. (2007). Berkshire Hathaway 2006 shareholder letter, the Schloss tribute: 47 partnership years, "one of the good guys of Wall Street." PDF.
Tweedy, Browne Company (1992, rev. 2009). "What Has Worked in Investing." Fifty-plus studies of cheap-asset and cheap-earnings investing worldwide. PDF.
Graham, Benjamin, and David Dodd (1934). "Security Analysis." The source. No link needed; every library on earth has it.
The modern lineage, still writing: Dave Waters' Oddball Stocks and Alluvial Capital letters, and David Orr's X feed and Idea Brunch interview, whose "actually analyze your own results" push is part of why my broker statements are on this site. No affiliation; none of these people know I exist.
Shelf two: the evidence for
The value premium
Fama, E. & French, K. (1992). "The Cross-Section of Expected Stock Returns." Journal of Finance. Cheapest book-to-market decile 1.83%/month vs 0.30% for the priciest, 1963–1990. Journal · free PDF
Lakonishok, J., Shleifer, A. & Vishny, R. (1994). "Contrarian Investment, Extrapolation, and Risk." Journal of Finance. Value beat glamour ~10.5 points/yr, 1968–1990, explained by extrapolation, not risk. Journal · NBER version
Fama, E. & French, K. (2012). "Size, Value, and Momentum in International Stock Returns." Journal of Financial Economics. Value premiums across Europe, Japan, Asia-Pacific; strongest in small stocks. Free PDF
Asness, C., Moskowitz, T. & Pedersen, L. (2013). "Value and Momentum Everywhere." Journal of Finance. Value premia in all eight markets and asset classes tested. AQR page
De Bondt, W. & Thaler, R. (1985). "Does the Stock Market Overreact?" Journal of Finance. Three-year losers beat winners by ~25% over the next three years. Journal
Campbell, J. & Shiller, R. (1988). "Stock Prices, Earnings, and Expected Dividends." Journal of Finance. Long-average earnings explain over half of ten-year return variance; the case for normalization. Free PDF
Deep asset value and net-nets
Oppenheimer, H. (1986). "Ben Graham's Net Current Asset Values: A Performance Update." Financial Analysts Journal. ~29%/yr arithmetic vs ~11.5% market, 1970–1983. Journal page
Xiao, Y. & Arnold, G. (2008). "Testing Benjamin Graham's Net Current Asset Value Strategy in London." Journal of Investing. 31% raw first-year returns; 11.3%/yr size-adjusted premium, 1981–2005. SSRN
Montier, J. (2008). "Graham's Net-Nets: Outdated or Outstanding?" SocGen Mind Matters; reprinted in his book Value Investing (Wiley, 2009). Global net-nets 35%/yr arithmetic vs 17%, 1985–2007. Alpha Architect's critical review (read this alongside it)
Mohanty, S. & Oxman, J. (2026). "Does Ben Graham's net current asset value investing continue to generate excess returns?" Review of Financial Economics. 13.9%/yr alpha 1969–2019, weakening after 2004. Journal
Survival screening
Piotroski, J. (2000). "Value Investing: The Use of Historical Financial Statement Information to Separate Winners from Losers." Journal of Accounting Research. The F-Score paper. SSRN · free PDF
Piotroski, J. & So, E. (2012). "Identifying Expectation Errors in Value/Glamour Strategies." Review of Financial Studies. The value premium lives where price disagrees with fundamentals. SSRN
Walkshäusl, C. (2020). "Piotroski's FSCORE: international evidence." Journal of Asset Management. Works across 20 developed + 15 emerging markets, 2000–2018. Journal · free PDF
Campbell, J., Hilscher, J. & Szilagyi, J. (2008). "In Search of Distress Risk." Journal of Finance. The most distressed stocks earn anomalously low returns. NBER version
Payout, profitability, governance
Boudoukh, J., Michaely, R., Richardson, M. & Roberts, M. (2007). "On the Importance of Measuring Payout Yield." Journal of Finance. Net payout beats dividend yield as a return signal. Free PDF
Novy-Marx, R. (2013). "The Other Side of Value: The Gross Profitability Premium." Journal of Financial Economics. Profitability complements value; the basis for the quality-exception lanes. Free PDF
La Porta, R., Lopez-de-Silanes, F., Shleifer, A. & Vishny, R. (2002). "Investor Protection and Corporate Valuation." Journal of Finance. Weak minority protection, lower valuations, 27 countries. Journal
Claessens, S., Djankov, S., Fan, J. & Lang, L. (2002). "Disentangling the Incentive and Entrenchment Effects of Large Shareholdings." Journal of Finance. Control above ownership destroys value; 1,301 East Asian firms. Journal
Djankov, S., La Porta, R., Lopez-de-Silanes, F. & Shleifer, A. (2008). "The Law and Economics of Self-Dealing." Journal of Financial Economics. Anti-self-dealing law predicts market outcomes across 72 countries. Free PDF
Skew, diversification, implementation
Bessembinder, H. (2018). "Do Stocks Outperform Treasury Bills?" Journal of Financial Economics. 57.4% of stocks underperform T-bills lifetime; ~4% of firms created all net wealth. SSRN
Novy-Marx, R. & Velikov, M. (2016). "A Taxonomy of Anomalies and Their Trading Costs." Review of Financial Studies. Low turnover and buy/hold spreads are what survive costs. NBER version
Shelf three: the case against
House rule: if this shelf ever stops growing, stop trusting this site.
Lev, B. & Srivastava, A. (2022). "Explaining the Recent Failure of Value Investing." Critical Finance Review. Expensed intangibles broke book-to-market; value averaged 0.65%/yr, 2007–2018. SSRN
Eisfeldt, A., Kim, E. & Papanikolaou, D. (2022). "Intangible Value." Critical Finance Review. Intangible-adjusted value beats traditional value. NBER version
Leadbetter, B., Li, F. & Linnainmaa, J. (2020). "Book Value Is an Incomplete Measure of Firm Size." Research Affiliates. Intangibles grew from ~30% to ~100% of tangible book since the 1960s. Article
Rizova, S. & Saito, N. (2020). "Internally Developed Intangibles and Expected Stock Returns." Dimensional. The counter-counterpoint: intangible adjustments are too noisy to trust. SSRN
Arnott, R., Harvey, C., Kalesnik, V. & Linnainmaa, J. (2021). "Reports of Value's Death May Be Greatly Exaggerated." Financial Analysts Journal. The −54.8% value drawdown, decomposed: nearly all revaluation. Free PDF
Blitz, D. & Hanauer, M. (2021). "Resurrecting the Value Premium." Journal of Portfolio Management. Small value −13% vs small growth +71%, Jan 2017–Aug 2020; and how composite value survives. SSRN
Israel, R., Laursen, K. & Richardson, S. (2021). "Is (Systematic) Value Investing Dead?" Journal of Portfolio Management. The defense, conceded weaknesses included. SSRN
Taylor, Y. (2021). "Why Piotroski's F-Score No Longer Works." Portfolio123. The within-value replication where the score's ranking inverts, 1999–2020. Article
Hyde, C. (2018). "The Piotroski F-score: evidence from Australia." Accounting & Finance. Raw spreads significant; factor-adjusted alphas mostly not. SSRN version
Anderson, K., Chowdhury, A. & Uddin, M. (2024). "Piotroski's Fscore under varying economic conditions." Review of Quantitative Finance and Accounting. The score is partly a macro thermometer; least reliable in contractions. Free PDF
Gray, W. (2015). "Simple Methods to Improve the Piotroski F-Score." Alpha Architect / AAII Journal. The equity-issuance signal flaw (fixed in this site's implementation). Article
McLean, R.D. & Pontiff, J. (2016). "Does Academic Research Destroy Stock Return Predictability?" Journal of Finance. Anomalies fade 26% out of sample, 58% post-publication. SSRN
Hou, K., Xue, C. & Zhang, L. (2020). "Replicating Anomalies." Review of Financial Studies. 65–82% of 452 anomalies fail conservative replication; the failures live in microcaps. Free PDF
Linnainmaa, J. & Roberts, M. (2018). "The History of the Cross-Section of Stock Returns." Review of Financial Studies. Accounting anomalies lose 50–70% of their power out of sample. Free PDF
Shumway, T. (1997). "The Delisting Bias in CRSP Data." Journal of Finance. Why backtests are kind to dying companies. Free PDF
Merani, G. / Alpha Architect (2019). Critical reviews of the net-net classics: Oppenheimer and Montier. Arithmetic means, micro market caps, untradeable tails.
Shelf four: the real-world trackers (AAII)
AAII, "Separating Winners From Losers: Low Price-to-Book Stocks." The Piotroski screen: 14.0%/yr since 1998 vs 6.1% S&P, price-only, as of Apr 2024. Article
AAII, screen methodology and caveats: "unachievable even in a best-case scenario." Article
AAII, "2008 Stock Screen Roundup: Piotroski Strategy Defeats the Bear" (+32.6% vs −40.3%, with an asterisk: it sat empty for months). Article
AAII, "Adjusting for the Real World": ~4 average holdings, 20 zero-stock months. Article
AAII, Model Shadow Stock Portfolio: real-money microcap value since 1993, 13.6%/yr vs 10.4% Vanguard 500 as of Mar 2026. Portfolio page · rules · April 2026 update
Gray, W., Vogel, J. & Xu, Y. (2014). "Does Complexity Imply Value? AAII Value Strategies from 1963 to 2013." F-Score best of thirteen at 16.74%/yr. Article · SSRN
Schadler, F. & Cotten, B. (2008). "Are the AAII stock screens a useful tool for investors?" Financial Services Review. 91% beat on paper; ~32% after costs. Journal
North, D. & Stevens, J. (2015). "Investment performance of AAII stock screens over diverse markets." Financial Services Review. Piotroski ranked first on alpha; costs are the killer. Journal
CXO Advisory, "AAII Stock Screens" audit: gross 12.1% average falls to 9.4% net; Piotroski highest net CAGR at 25.4%; edges deteriorate over time. Article
Shelf five: other ways to make money (that aren't mine)
Deep value is one documented edge among several. If you read this whole site and realize you're wired for a different one, that's the site working. Start here.
Jegadeesh, N. & Titman, S. (1993). "Returns to Buying Winners and Selling Losers." Journal of Finance. The founding momentum paper. Free PDF
Hurst, B., Ooi, Y.H. & Pedersen, L. (AQR). "A Century of Evidence on Trend-Following Investing." 1880 to present. AQR page
Faber, M. "A Quantitative Approach to Tactical Asset Allocation." The practical retail on-ramp for trend. SSRN
Alquist, R., Israel, R. & Moskowitz, T. "Fact, Fiction, and the Size Effect." The skeptical take on small-cap investing, included on purpose. AQR page
Asness, C., Frazzini, A., Israel, R. & Moskowitz, T. "Fact, Fiction, and Value Investing." The same honest treatment for my own side. PDF
Nothing on this page is investment advice, and a citation is not an endorsement of everything its author believes. Sources were last verified August 2026. If you find a dead link or, better, a study I should be forced to reckon with, send it. The against shelf has room.