Fake Scientific Calculators: Six Tests That Expose a Broken One in Thirty Seconds

There is no register of "fake" calculators, and the interesting failures are not frauds anyway. They are ordinary tools built by someone who wired a JavaScript expression evaluator to a keypad and never checked the edges. The result looks exactly like a scientific calculator and quietly disagrees with the one in your exam. So rather than name and shame, this guide gives you a test battery you can run on any calculator, including ours, in about half a minute.

The six tests that work

These survive display rounding, which is what makes them useful. A scientific calculator shows about 10 significant digits, so any test whose failure hides in the 11th digit will pass on a careless tool. These do not hide.

Test 6 is the one people skip and it is arguably the most practical. A calculator that will not tell you whether it is in degrees or radians will eventually hand you a confidently wrong answer, and you will have no way to notice.

Why sin(180) catches so much

Computers store numbers in binary, in a fixed number of bits. The value stored for the constant pi is the nearest representable double precision number to pi, which is not pi. It is off by roughly 1.2 parts in 10¹⁶.

To compute sine in degrees, the naive approach multiplies by pi/180 and calls the built in sine function. So sin(180) becomes the sine of a number that is very slightly not equal to pi. Sine near pi is almost exactly linear, so the tiny input error passes straight through to the output: you get about 1.2246e-16 instead of zero.

The important part: this is not a bug in the sine function. It is the correct sine of the number it was actually given. The failure is that a calculator aimed at students passed a floating point artefact through to the display instead of recognising that 180 degrees is a quadrant boundary with an exact answer. Careful calculators special case those boundaries. Careless ones do not, and that single decision separates the two populations more reliably than any other test.

This is also why sin(pi) in radian mode is a bad test. There, you typed an approximation of pi yourself, so a residue is the mathematically honest response. Only the degree mode version is a fair test.

The famous tests that do not work

You will find these repeated everywhere. They are real phenomena and poor detectors.

• 0.1 + 0.2, which in raw binary floating point gives 0.30000000000000004. A genuine finding about computer arithmetic, but it only catches a tool that does no output rounding whatsoever. Every calculator that rounds to 10 digits passes, careful or not.
• (√2)² giving 2.0000000000000004, and 0.3 - 0.1 giving 0.19999999999999998. Same category, same weakness.
• Large integer cancellation such as 10^15 + 1 - 10^15. This one is actively misleading as a fake detector, because it can fail on the real hardware. A physical Casio works to 15 significant decimal digits internally, so it can return 0 here, while a browser calculator using binary doubles returns the correct 1. Marking the browser answer as evidence of fakery gets it exactly backwards.

If a guide leads with 0.1 + 0.2 and stops there, it has not tested anything a rounding call would not have hidden.

The one test with no right answer

You will see 6 ÷ 2(1+2) offered as a correctness test. It is not one. It is a convention test, and the expression is genuinely ambiguous.

Casio's own manual assigns implicit multiplication a higher priority than explicit division, so a physical fx-991ES PLUS returns 1. Most software, including JavaScript expression parsers, treats implicit multiplication at the same priority as division and evaluates left to right, returning 9. Mathematicians writing about this generally conclude the notation is ambiguous and should be avoided rather than adjudicated.

So do not call 9 wrong. But do know which convention your tool follows, because if you are practising for an exam on a physical Casio and your browser calculator disagrees, that difference will cost you marks on exactly this kind of expression. Type it once and find out. Ours returns 1, matching the hardware, and we say more about that below.

What Casio actually offers online

This matters because "official Casio online calculator" is a claim some sites make, and it is worth knowing the real position before you trust one.

• Casio runs ClassPad.net. ClassPad Math, its graphing and notes environment, is free. The ClassWiz emulator on that platform is paid.
• There is no free official browser emulator of the fx-991ES PLUS.
• Casio's older free online calculator service, keisan.casio.com, closed on 20 September 2023. Its notice states the site was closed on that date and user data was removed.

So a site presenting itself as the official free Casio fx-991ES PLUS online is misrepresenting something. Independent Casio-style tools are legitimate, ours included, but the honest ones say plainly that they are independent and unaffiliated. If a site does not say that anywhere, treat the omission as information. We cover the landscape in Casio fx-991ES PLUS online alternatives.

When it is worse than inaccuracy

Wrong answers are the common problem. A much smaller set of cases involves calculator branding used as cover, and it is worth being precise rather than alarming, because the evidence is narrower than the internet suggests.

Calculator branded apps that were not calculators. Security researchers at CYFIRMA reported in February 2025 on an Android app marketed as a finance calculator that served fake loan applications through a WebView instead of the advertised functionality, while harvesting contacts, call logs, SMS and clipboard contents. ThreatFabric reported in October 2022 on an Italian tax code calculator app with over 10,000 installs that showed a fake update page and installed a banking trojan. These are two documented cases, not a wave.

Calculator styled vault apps. This is the best evidenced strand. Zhang, Baggili and Breitinger, writing in Computers & Security in 2017, studied 18 Android vault applications together accounting for close to 220 million downloads, and found that 6 of the 18 did not encrypt stored photos, 8 did not encrypt videos, and 7 stored passwords in cleartext. The headline is not that these apps are malicious. It is that the app promising to hide your files largely does not encrypt them. One caveat worth stating: those are vault apps generally, and only one of the eighteen is calculator-styled, so treat this as evidence about the category the disguise borrows from rather than about calculators.

Deceptive download patterns on free utility sites. Malwarebytes documented a network of deceptive download sites where hovering over a download button previews the genuine developer URL, but a click is intercepted by JavaScript and routed elsewhere. Their summary is the useful lesson: the link looks safe when you hover, but the click says otherwise. Do not trust a hover preview on a download button.

The practical conclusion is mild and boring: a calculator that runs as a plain web page cannot read your other tabs, your cookies or your files, because the browser's origin model forbids it. That guarantee is stronger than any promise in a listing. Prefer a page over an install unless you specifically need a side panel, and if you do install something, read the permissions. We did that for thirteen Chrome extensions in the extension comparison, and one with 10,000 users asks for access to every website you visit.

The trust checklist

1. Run tests 1 to 3. Thirty seconds, and they catch the majority of careless builds.
2. Check for a mode indicator on the display. No D, R or G means you cannot verify your own trig.
3. Confirm it is HTTPS and that a privacy policy exists and loads.
4. Look for an independence statement if it imitates a known brand. Its absence tells you something.
5. Prefer no install. A web page cannot reach your other tabs. An extension might.
6. If you must install, read the permissions, and remember the store does not show them on the listing.
7. Distrust hover previews on download buttons.
8. Check whether it matches your exam hardware on 6 ÷ 2(1+2), so a convention difference does not surprise you later.

How our own calculator scores

It would be poor form to publish a test battery and not run it on ourselves. We ran all of it against WideMath on 6 September 2026. These are the actual displayed results.

Full disclosure: until this week WideMath failed two of its own tests. It returned 1.224646799e-16 for sin(180), exactly the failure described at the top of this article, and it followed the software convention on 6 ÷ 2(1+2) rather than Casio's. Both were found while writing this piece and both are now fixed, with a regression test covering the priority table, the display rounding and the range gates. We are pointing at them rather than quietly shipping, because a guide about calculators that do not check their edges should admit when its author did not either.

Frequently asked questions

Sources

Every external claim above, in the order it appears.

• Floating point and IEEE 754. David Goldberg, What Every Computer Scientist Should Know About Floating-Point Arithmetic, Oracle. Also the Python tutorial on floating point and 0.30000000000000004.com.
• Casio's documented calculation ranges and priority sequence. fx-115ES PLUS / fx-991ES PLUS User's Guide (PDF, Casio). This is the source for the factorial range of 0 to 69, the tangent domain exclusions and the priority-7 ranking of omitted multiplication signs.
• The ambiguity of 6 ÷ 2(1+2). Oliver Knill, Harvard, Ambiguous PEMDAS (PDF), and The PEMDAS Paradox in Plus Magazine.
• What Casio offers online. ClassPad.net, and Casio Education UK on emulator licence codes, which is the basis for describing the ClassWiz emulator as paid. The closure notice is on keisan.casio.com itself.
• The vault-app study. Xiaolu Zhang, Ibrahim Baggili and Frank Breitinger, "Breaking into the vault: Privacy, security and forensic analysis of Android vault applications", Computers & Security vol. 70 (2017), pp. 516 to 531. Open access copy. Note that these are Android vault apps generally, not calculator-styled ones specifically; only one of the eighteen is calculator-styled.
• The finance calculator app. CYFIRMA, 21 February 2025, SpyLend research report.
• The tax code calculator dropper. ThreatFabric, 28 October 2022, The Attack of the Droppers, and BleepingComputer's coverage.
• Download buttons that lie on click. Malwarebytes, August 2026, 41 deceptive download sites show a real link, then send you somewhere else.