Home / Indicator library / Financial strength scores
Financial strength scores
Hundreds of line items. Which ones matter?
Four published academic scorecards, computed from the official quarterly filings and shown side by side: whether the business is getting stronger, how far it sits from distress, whether the earnings look managed, and a composite read across profitability, safety, growth and ownership. Each score is reported with the inputs it was built from, and withheld where the filings do not support it.
- Three financial statements open and no idea which line to start at.
- Wanting to drop weak balance sheets from a universe without hand computing ratios for every name.
- Strong headline earnings, with no cheap way to check whether the accounting flatters them.
See it
Four scorecards across five tickers, computed only from filings that were public as of the date shown. Rendered on the server with no key, no signup and no quota.
| Ticker | F-score | Z-score | M-score | Composite health | Basis quarter |
|---|---|---|---|---|---|
| 2330 TSMC | 6/6 | inputs not filed | -2.128 (no flag) | 81.2 A | 2025Q1 |
| 2454 MediaTek | 4/6 | inputs not filed | -2.348 (no flag) | 60.3 C | 2025Q1 |
| 2317 Hon Hai (Foxconn) | 2/6 | inputs not filed | -2.306 (no flag) | 36.6 D | 2025Q1 |
| 2308 Delta Electronics | 3/6 | inputs not filed | -2.442 (no flag) | 58.4 C | 2025Q1 |
| 2412 Chunghwa Telecom | 5/6 | inputs not filed | -2.467 (no flag) | 61.6 C | 2025Q1 |
| 2882 Cathay Financial | 1/2 | not applicable | -2.481 (no flag) | 34.9 D | 2025Q1 |
How it is built
Four published models, applied to Taiwan filings with the adaptations stated rather than hidden.
The F-score, with an honest denominator
Nine yes-or-no tests on profitability, balance sheet and operating efficiency, one point each. Where a filing does not carry the input a test needs, that test is reported as unavailable and the denominator shrinks — six out of six is stated as six out of six, not silently rounded up to a nine-point score.
The Z-score, and the two ways it can be absent
A distress model with separate formulas for manufacturers and non-manufacturers. Financial companies are refused outright rather than forced through a formula built for a different balance sheet, because the resulting number would be precise and meaningless. Separately, a quarter whose filing did not carry the working-capital and retained-earnings components returns partial rather than a headline score assembled from the components that happen to be there. The table names which of the two applies, because they mean opposite things about whether the number will ever exist.
The M-score, with its assumptions labelled
An earnings-manipulation flag built from eight ratios. Variables that can be computed from filings are computed; those that cannot are set neutrally and marked as such, so the output is never presented as more informed than its inputs.
The composite, computed as of a date
Profitability, safety, growth and ownership combined against a market-wide percentile, using only statement versions whose statutory filing date had passed. Ask about a past quarter and the answer is the score that was computable then, which is what makes it usable in a backtest.
Terms
Four scores, one line each:
Where the line is
All four models were fitted on other markets and other decades, and none of them were validated on Taiwan filings by their authors. They are useful as a systematic first pass over a universe — cheap, uniform, and free of the narrative that makes a familiar name feel safer than its statements say. They are not a substitute for reading the accounts, and the M-score in particular flags a statistical pattern rather than misconduct. The more valuable property here is the one the scores are wrapped in: because every input is gated on its statutory filing date, a score from a past quarter is one a research process could actually have acted on.
FAQ
What do the F, Z and M scores each measure?
The F-score is a nine-point checklist asking whether financial health improved year over year. The Z-score measures distance from distress, where higher is safer. The M-score flags a statistical pattern consistent with managed earnings — a prompt to read the notes, not an accusation.
Why is the F-score shown out of six rather than nine?
Because only six of the nine tests could be computed from the filings available for that company. Tests whose inputs are missing are reported as unavailable and removed from the denominator rather than counted as failures, which would understate a company for a disclosure gap that has nothing to do with its finances.
Can financial companies be scored with a Z-score?
No, and this site refuses rather than approximates. The Z-score has separate formulas for manufacturers and non-manufacturers, and a bank or insurer's balance sheet fits neither. Pushing one through anyway returns a precise number that means nothing. The table marks that case as not applicable, which is a different statement from a quarter whose filing did not carry the components — that one reads as inputs not filed, because the score may well exist next quarter.
Do past scores use information that was not public yet?
No. Every score is computed only from statement versions whose statutory filing date had already passed on the date asked about. That is the difference between a score you can backtest and one that quietly rewards you for knowing a result before it was filed.
Are these ratings, or advice?
Neither. They are published academic formulas applied to public filings, with the adaptations and missing inputs stated. They are not credit ratings, not a recommendation, and a high composite score has never prevented a share price from falling.
Run it yourself
The free tier needs no signup — call this feature's API straight away with the demo key. Unlock as-of history and full ticker coverage on a paid tier.