The idea in one minute
Cheap stocks are usually cheap for a reason. Most companies trading at a low price compared with their book value are weak businesses, and in Piotroski's data fewer than half of them beat the market over the following year.
But they are not all the same. Some are quietly recovering: profits are back, cash is coming in, debt is falling, margins are improving. Piotroski's insight was that a short checklist of simple facts from the annual report is enough to tell the two groups apart, and that the market is slow to notice the recovery, especially in smaller companies that few analysts follow.
The F-Score is that checklist. Each of nine tests scores one point if passed, so every company gets a score from 0 to 9. The strategy is to look only among cheap stocks, and only at those scoring 8 or 9.
Where it comes from
Joseph D. Piotroski wrote the paper as an accounting professor at the University of Chicago: "Value Investing: The Use of Historical Financial Statement Information to Separate Winners from Losers", published in the Journal of Accounting Research in 2000. He is now at Stanford.
He was clear that the nine signals are not a clever, optimised formula. In Chicago Booth's magazine he described them as an intuitive set of performance measures, chosen because they are easy to understand and easy to check. That simplicity is part of why the score is still so widely used twenty-five years later.
The nine tests
All nine come straight from the annual accounts. "This year" means the latest financial year and "last year" the one before. A test that is merely unchanged scores zero: improvement has to be real.
| Test | In plain English | One point if | |
|---|---|---|---|
| 1 | Return on assets | The company made a profit | Net income before one-off items is positive |
| 2 | Operating cash flow | The business brought in real cash | Cash flow from operations is positive |
| 3 | Change in return on assets | Profitability is improving | Return on assets is higher than last year |
| 4 | Accruals | Profit is backed by cash, not accounting | Operating cash flow is larger than net income |
| 5 | Change in leverage | The debt burden is shrinking | Long-term debt relative to assets fell |
| 6 | Change in liquidity | Short-term finances are stronger | The current ratio rose |
| 7 | No new shares | It didn't need to raise money from shareholders | No new common shares were issued |
| 8 | Change in gross margin | Better pricing power or lower costs | Gross margin is higher than last year |
| 9 | Change in asset turnover | Assets are used more productively | Sales relative to assets rose |
The groups matter. Tests 1 to 4 ask whether the company is profitable and whether that profit is real. Tests 5 to 7 ask whether its finances are getting safer. Tests 8 and 9 ask whether the business itself is running better.
The detail most guides miss. The F-Score measures change, not level. It asks whether debt is falling, not whether debt is low. A company with plenty of debt can still score 9 if it is paying it down. That is deliberate: it is a recovery signal.
How the original strategy worked
- Cheapness first. Each year, take only the cheapest fifth of the market by price compared with book value.
- Then score. Compute the F-Score from the latest annual report.
- Keep the 8s and 9s, and avoid the 0s and 1s.
- Wait for the report. Portfolios were formed in the fifth month after the financial year ended, so the annual report was certainly public.
- Hold for a year, then repeat. Two-year results were also reported.
Nothing else was added: no requirement for a long record of stable earnings, low debt or high profitability beyond the score itself. The point was to look inside the unloved, often troubled, cheap end of the market.
What the evidence shows
The original study covered US companies from 1976 to 1996. Measured against the market over the following year:
Cheap stocks as a group beat the market by 5.9 points, but fewer than half of them did so. High scorers beat it by 13.4 points; low scorers lagged by 9.6. Low scorers were also several times more likely to be delisted for poor performance. A simpler split, score of 5 or more against less than 5, worked in 18 of the 21 years.
Does it still work, and in Europe?
The most useful test is one done entirely after the paper was published. Christian Walkshäusl (2020) looked at 20 developed and 15 emerging markets from 2000 to 2018. High scorers beat low scorers by about 10 percentage points a year worldwide, and Europe was the strongest region, with the effect largest among small and mid-sized companies and still present among large ones. A good part of the gain came from avoiding the low scorers.
A separate European study (Tikkanen and Äijö, 2018) found that adding an F-Score screen improved every European value strategy tested, whether value was measured by book value, earnings, dividends or operating profit.
The honest caveats
- Practitioner back-tests are mixed. One recent ten-year student-fund test using the cheapest fifth by book value earned slightly more than the S&P 500, but with a much deeper worst fall; without the cheapness filter the extra return disappeared.
- Piotroski himself flagged possible data-snooping, said binary scoring throws information away, and left open whether the extra return is mispricing or hidden risk.
Where it works, and where it doesn't
It works best in small and mid-sized companies, especially those with no analyst coverage and little trading, which is where good news takes longest to be noticed. In Piotroski's original US data the largest companies showed no significant effect; internationally, after 2000, it still worked in large companies, but less strongly.
It works poorly, or not at all, for banks and insurers, whose accounts have no meaningful gross margin, current ratio or asset turnover.
Eight common mistakes
- Scoring every stock. The documented strategy is "cheap first, then score". The score helps elsewhere too, but that is a different strategy.
- Lowering the bar to 7 or 6. The paper's winners score 8 or 9. Many screeners call anything above 6 "good".
- Getting the low group wrong. Some guides compare 8–9 with 0–3. The paper's low group is 0–1.
- Letting negative book value through. A company with negative equity has a negative price-to-book ratio. It isn't cheap; it has a hole in its balance sheet.
- Including banks and insurers. Several tests don't make sense for them.
- Adding "quality history" filters. Requiring five years of stable profits removes exactly the recovering companies where the strategy is strongest.
- Re-scoring every quarter and trading often. The evidence is for annual data and roughly a year's holding.
- Counting shares instead of share issues. A company that issues shares and buys back as many looks unchanged on share count, but fails Piotroski's test 7.
How ValuJack applies it
The ACE card follows Piotroski as closely as the data allows, and says openly where it doesn't.
Score of 8 or 9 only. The paper's high-score group, not the "6 and above" many screens use.
Banks and insurers excluded. The tests don't fit their accounts. Piotroski didn't need to say this; practical implementations do.
The cheapest third, not the cheapest fifth. Strictly following the cheapest fifth by price-to-book leaves too few companies once the other checks are applied. The post-2000 international evidence, which covers all companies, supports the slightly wider cut. Negative book values are excluded.
Our data provider's version of the score. It uses the nine Piotroski tests on the latest twelve months, with operating margin in place of gross margin and the share count in place of share issues. Close and recognisable, but not identical.
Positive earnings and a non-shrinking five-year trend. Our fair value model needs them. The cost is honest: some genuine recoveries with a bad five-year record are left out.
A ceiling on debt. The F-Score rewards falling debt, not low debt. We add a simple safety net against very high debt.
Price has turned up from its low. A cheap stock that is still falling is hard to hold. We want the price clearly above its one-year low before the card is dealt.
A note for European investors. Since 2019, lease accounting (IFRS 16) moves part of rent payments out of operating cash flow. That makes tests 2 and 4 a little easier to pass for lease-heavy European companies such as retailers and restaurant chains than they were in the US data of the 1990s. It's worth a second look when a high score rests on those two tests.
The same final steps for every card
Whatever the strategy, a company only becomes a card after three more steps. Our own model estimates a fair value from the company's reported earnings, and only companies priced below it (or, for the two momentum and growth cards, not far above it) can be dealt. The Edge Check, described below, then reads the latest reports, filings and news for what the numbers can't show. Finally the company must be tradable at major European online brokers. The full process is on the methodology page, and every card ever dealt stays on the card history.
The Edge Check: what the screen can't see
Every original strategy leaves something to judgement that no screen can measure. The Edge Check is ValuJack's answer: before a card is dealt, an AI research agent reads the company's latest reports, official filings and news, and asks the one question that matters most for this strategy.
The ACE question: Is the improvement real and still going?
A high F-Score says last year's accounts improved. It can't say whether the improvement came from the business or from one-offs, or whether it has continued since.
Counts in favour
- The latest interim report continues the improvement across margins, cash flow, debt and return on assets
Counts against
- Improvement mostly from one-offs: asset sales, released provisions or cash squeezed out of working capital
- The improvement reversed in the latest interim report
- A breach of loan terms, or a waiver from the lenders
- The share price is still falling
Checked for every card
- Insider trades: are the company's own managers buying or selling shares with their own money? Awards, option exercises and routine trades are ignored.
- Short sellers: are professional investors betting against the company, and is that bet growing?
- Report wording: has anything changed since last year in the risks, targets, dividend policy, accounting rules or the auditor's report?
- Customers, suppliers and peers: what have they said about demand since the company's own latest report?
- Management's record: have they met their own targets, and have the finance chief or the auditor left unexpectedly?
- The latest report: a clear beat or miss, and has guidance been raised or cut?
What stops a card outright
A profit warning, a takeover bid, a trading suspension, profits not backed by cash, doubt about the company's future or a qualified audit opinion, restated accounts, a large new share issue, a regulator investigating the accounts, or an unanswered fraud allegation from short sellers.
Argue the opposite
Before any verdict, the agent writes down the three strongest reasons this card could fail within the strategy's time frame, and looks for evidence on each. A reason that is confirmed counts against the company, or stops it.
How it changes the cards
Every trading day the best-ranked companies in each tier are checked. A stop removes a company; the other signals move companies up or down the order, and the best-placed company that passes is dealt. A company due to report within the next week usually waits until after its report. Open a card's details to see its Edge Check: the question, a short answer, up to three signals marked ✓ or △, what would break the case, and the date it was checked.
Every point needs a source and a date: the company's own reports, official registers of insider trades and short positions, and exchange news. Forums, social media and analysts' price targets never count, and a source that can't be reached counts as no signal, not as good news. The Edge Check never changes the fair value or the star rating, and never says buy or sell. It is research done by an AI agent under fixed rules, live since 28 October 2026; it can miss things or be wrong.
Questions people ask
What is a good Piotroski F-Score?
In the original research the winners scored 8 or 9 out of 9, and the losers 0 or 1. Most companies sit in the middle, between 3 and 7.
Does the F-Score work in Europe?
Yes. A study of 2000 to 2018, entirely after the paper was published, found the gap between high and low scorers was largest in Europe, especially among small and mid-sized companies (Walkshäusl, 2020).
Can I use the F-Score on growth stocks?
You can calculate it for any company, but the strategy that was tested applies it only to the cheapest stocks by price-to-book. Used on expensive growth stocks it becomes a different, untested strategy.
Why does the F-Score ignore how much debt a company has?
It measures direction rather than level: it rewards debt that is falling, because that signals recovery. That is why ValuJack adds a separate ceiling on debt.
Sources
- Piotroski, J. (2000). Value Investing: The Use of Historical Financial Statement Information to Separate Winners from Losers. Journal of Accounting Research 38.
- Chicago Booth Review: Separating winners from losers
- Walkshäusl, C. (2020). Piotroski's FSCORE: international evidence. Journal of Asset Management 21.
- Tikkanen, J. & Äijö, J. (2018). Does the F-score improve the performance of different value investment strategies in Europe? Journal of Asset Management 19.
- QuantConnect research: Piotroski F-Score investing
- ProfitProphets student fund pitch (April 2026)
- Investopedia: Piotroski Score
KING
QUEEN
JACK
JOKER
TWO