Lesson 2 of 7
A rank is not a threshold
The cheapest fifth of the market and P/B under 1.5 sound alike. A rank always returns a fixed share; a threshold returns whatever qualifies today.
By Javier Sanz
Founder of ValueScreener. Built Ninety Nine, a securities services provider, and sold it to a tier-one global bank; ran operations and internationalization at Alpaca, the brokerage API.
Who writes thisEditorial processLinkedIn (opens in a new tab)X (opens in a new tab)
In one minute
- A rank sorts the whole market and keeps a fixed share of it, such as the cheapest fifth.
- A threshold keeps whatever clears one fixed number, so its count depends on how the market is priced today.
- The two are described in similar words but return very different lists in a cheap year and a dear one.
"The cheapest fifth of the market" and "P/B at or below 1.5" sound like the same instruction. A rank always returns a fifth of whatever universe you sort. A threshold returns however many companies happen to clear that number today, which can be almost none or several hundred depending on how the market is priced.
Where a P/B ceiling of 1.5 sat on 2026-09-07
A P/B ceiling of 1.5: between the 25th percentile (1.14x) and the median (2.17x).As a ceiling it admits between a quarter and a half of the cohort.
- 5th percentile: 0.33x
- 10th percentile: 0.60x
- 25th percentile: 1.14x
- median: 2.17x
- 75th percentile: 5.02x
- 90th percentile: 10.87x
- 95th percentile: 19.62x
A rank keeps a share of the market
A rank sorts every company in a universe by one measure and keeps a fixed fraction: the cheapest fifth, the top decile, the best half. Greenblatt's Magic Formula ranks companies on earnings yield and on return on capital, then buys the names with the best combined rank across both. He published no minimum figure for either measure, because a ranked portfolio does not need one: it always holds the same share of the universe, however the market is priced that year. Piotroski's sample was a rank too: he applied his nine tests only inside the cheapest fifth of the market by book-to-market, a rank he built before running a single one of the nine tests.
A threshold keeps a level
A threshold is a fixed number: P/B at or below 1.5, P/E at or below 15. It returns every company that clears the line today, and nothing tells you in advance how many that will be. As the figure above shows, a P/B ceiling of 1.5 sat between the 25th percentile and the median of the cohort built on 2026-09-07, admitting between a quarter and half of it. That share is a fact about that day's prices, not a fact the threshold itself carries.
The same line, a different year
Move every price in the market down and the same P/B ceiling of 1.5 admits more companies, because more of them now sit under it. Move every price up and the ceiling admits fewer, in the extreme almost none.
A rank
- Always returns the same share, whatever the market did.
- Fills its quota even when every company in it is expensive by any absolute standard.
A threshold
- Keeps an absolute price standard that does not move with the market.
- Its count moves with the market: it can admit almost none, or several hundred.
A rank tells you what is cheapest. A threshold tells you what is cheap.
What a threshold cannot express
A condition in this product's own screener is one field, one operator, one fixed value, checked against a stored row. It cannot rank the live universe on a ratio you choose, such as book-to-market: there is no combined ordering across the whole universe, and no instruction that means "the best fifth, whatever that turns out to be." Every discipline here that was published as a rank, including Greenblatt's, is run as a fixed threshold instead, and the screen says so rather than pretending the two are the same test.
Check yourself
Which of these always returns the same share of the market, whatever year it is?
- A fixed P/B ceiling, such as 1.5
- A rank, such as the cheapest fifth by book-to-market
- Neither does
Show the answer
Answer: B.
A rank sorts the whole market and keeps a fixed fraction of it, so its count does not depend on how the market happens to be priced.A P/B ceiling of 1.5 sat between the 25th percentile and the median of the cohort built on 2026-09-07. If the whole market gets cheaper, what happens to the number of companies it admits?
- It falls
- It rises
- It stays exactly the same
Show the answer
Answer: B.
A cheaper market pushes more companies under any fixed price ceiling, so the same threshold admits more of them.Greenblatt published his Magic Formula as a rank across two measures. What minimum figure did he give for either one?
- An earnings yield of 8 percent
- A return on capital of 15 percent
- None. A ranked portfolio does not need a minimum
Show the answer
Answer: C.
A rank always returns the same share of the universe, so there is nothing for a minimum figure to do.Sources
- The screener's own account of what a condition can and cannot express: one field, one operator, one value against a stored row, never a rank across the live universe (
frontend/src/components/screener/strategies.ts, doc comment). - Greenblatt, The Little Book That Beats the Market, 2005: earnings yield and return on capital, ranked and combined.
- Piotroski, Journal of Accounting Research, 2000: the F-Score, applied inside a high book-to-market cohort.
- The P/B distribution published on the open research pages of ValueScreener, built 2026-09-07.
What this leaves out
- The distribution above is dated 2026-09-07, built from prices last stored on 2026-08-26. A later reader will see different breakpoints and a different cohort size.
- This product's screens run fixed thresholds only. None of them reproduces a published rank exactly, and each says so where it substitutes one.
- The cheap-year and dear-year comparison uses round invented counts, not two real dates of the distribution.
- Nothing here is advice about any security.