Inverse Symbolic Calculator, Explained
Paste a decimal like 1.6449340668 and find the exact expression it came from, \pi^2/6. This guide explains how the search works, how to read a match, and when to trust it.
What the tool does
You measure or compute a number and get a decimal: 1.6449340668. That decimal is a shadow. The Inverse Symbolic Calculator tries to recover the object that cast it. In this case the answer is \pi^2/6 = \zeta(2), the sum of the reciprocals of the squares. Check it: \pi^2 = 9.8696044\ldots, and dividing by 6 gives 1.64493407\ldots, matching every digit you supplied.
The tool works in two modes. In number mode it takes one decimal and searches for a closed form built from known constants. In sequence mode it takes a list of integers and looks for a formula or recurrence that reproduces them, then predicts the next terms. Both modes share the same discipline: a candidate is only reported when your digits rule out a lucky accident.
When to use it, and when not
Use it when a decimal appeared from real mathematics and you suspect a clean form. A numerical integral that settles near 0.9159655942 is a good candidate: that is Catalan's constant G. A limit, a residue, a special value of a series, the volume of a polytope: these often have closed forms, and the tool exists to name them.
Do not use it on numbers with no reason to be special. A stock price of 142.37, a measured mass, a random sample mean: these have no closed form, and any "match" is coincidence dressed up as insight. The tool guards against this (see the scoring below), but the strongest guard is your own judgement about where the number came from.
Give the tool as many correct digits as you have. A match against 6 digits is a guess. A match against 15 digits is evidence. If you computed the number to only 4 places, the tool cannot tell \pi^2/6 = 1.6449 apart from thousands of nearby rationals.
The math: integer relations
Behind the tool is one idea. Suppose you have a small list of known constants, for example 1, x, \pi, \pi^2 where x is your input. An integer relation is a set of whole numbers a_i, not all zero, such that the weighted sum is zero:
Here a_0, a_1, a_2, a_3 are the unknown integers. If a small relation exists, it encodes a closed form. For x = 1.6449340668 the relation -6x + \pi^2 = 0 has coefficients a_0 = 0, a_1 = -6, a_2 = 0, a_3 = 1, which rearranges to x = \pi^2/6.
Finding such relations by hand is hopeless. The PSLQ algorithm and lattice reduction do it systematically: given the constants to high precision, they find the smallest integer coefficients that drive the combination below your precision floor. Continued fractions handle the simplest case, a single rational multiple, by expanding x as
When the expansion terminates or one term is huge, you have found a good rational approximation. For example 3.1428571429 expands to 3 + 1/7 = 22/7, the classic approximation to \pi. The fraction and continued fraction converter shows these expansions directly.
Why chance fits are rejected
Any decimal can be fit if you allow large enough integers. The rational 16449340668/10000000000 reproduces 1.6449340668 exactly, but it explains nothing: its numerator carries 11 digits, exactly as many as the input. That is the test. A candidate whose coefficients could reproduce your digits by luck is worthless.
The tool measures this with a complexity score. It counts the decimal digits a candidate's integer coefficients could explain by accident. A match is shown only when your input carries at least 3 more digits than that score, and it is called convincing at 6 or more spare digits.
Compare the two fits for 1.6449340668 (10 significant digits after the leading 1). The form \pi^2/6 uses the coefficient 6, worth about 1 digit of complexity, leaving 9 spare digits: convincing. The brute rational leaves 0 spare digits: rejected.
| Candidate | Complexity (digits) | Spare digits | Verdict |
|---|---|---|---|
| \pi^2/6 | 1 | 9 | Convincing |
| 16449340668/10^{10} | 10 | 0 | Rejected |
Reproducing the demo: 1.6449340668
- The input is
1.6449340668, carrying 11 significant digits. - A continued-fraction pass finds no short rational: the expansion is 1 + 1/(1 + 1/(1 + 1/(1 + \ldots))) with no terminating tail, so a plain fraction is ruled out.
- The integer-relation search over the basis tries multiples of named constants. It tests x against \pi^2 and finds 6x - \pi^2 \approx 0 to 10 digits.
- The coefficient 6 has complexity near 1 digit, leaving about 9 spare digits, well past the threshold of 6.
- The tool reports \pi^2/6, and notes the equivalent name \zeta(2), the value of the Riemann zeta function at 2.
You can verify the digits with high-precision arithmetic in the big number calculator: compute \pi to 30 places, square it, divide by 6.
Reading a result
A result gives you the closed form, an equivalent name where one exists, and the digit accounting. Read the spare-digit count first. If it says 3 to 5 spare digits, treat the form as a hypothesis to confirm: recompute your number to more places and run it again. If it says 6 or more, the form is very likely correct, though "likely" is not "proven". The tool finds relations; it does not prove them.
Several forms can name the same number. \pi^2/6, \zeta(2), and the sum \sum 1/n^2 are one value under three descriptions. The tool prefers the simplest, but knowing the alternatives helps you connect the number to whatever produced it.
A sequence: 1, 4, 9, 16, 25
Switch to sequence mode and enter 1 4 9 16 25. The tool fits a polynomial in n. The differences are 3, 5, 7, 9, and the second differences are constant at 2, so the fit is degree 2:
It predicts the next terms 36, 49, 64. Because constant second differences pin down a quadratic exactly, four terms already determine it; the fifth term confirms rather than adds information. A sequence like 1 1 2 3 5 8 instead matches the linear recurrence a_n = a_{n-1} + a_{n-2}, the Fibonacci rule, with next terms 13, 21.
Common mistakes
The frequent errors are all about precision and provenance.
- Too few digits
- Feeding
1.645instead of1.6449340668. Four digits admit far too many fits. Supply at least 8, ideally 12 or more. - Trailing rounding error
- If your last digit is wrong, no clean form matches.
1.6449340669(last digit off by one) may miss \pi^2/6 entirely. When a value you trust returns nothing, drop the final digit and retry. - Numbers with no closed form
- Physical measurements and random data have no exact expression. A reported match on such input is coincidence.
- Off-by-one in sequences
- A single wrong or transposed term breaks recurrence detection. Enter the exact integers, and give the tool at least 4 terms.
Related tools
Once you have a closed form, other tools help you work with it. For rational and continued-fraction views of a decimal, use the fraction converter. To evaluate a form to hundreds of digits and confirm a match, use the big number calculator. If a recovered sequence is combinatorial, the combinatorics tool computes binomial, Stirling and Bell numbers exactly. For questions about the integers in a result, factor them with the number theory tool, and convert between bases with the base converter. To round a recovered value cleanly, the scientific notation tool handles significant figures.
Frequently asked questions
How many digits should I enter?
At least 8, and 12 to 15 when you can. The spare-digit rule means a form with complexity 4 needs about 7 digits before it is even shown and 10 digits before it is called convincing. More digits never hurt.
Does a match prove my number equals that expression?
No. The tool finds a relation consistent with your digits to the precision you gave. That is strong evidence, not a proof. To prove it you still need mathematics, for example deriving \zeta(2) = \pi^2/6 from its series.
Why did I get no result for a number I think is special?
Three usual causes: too few digits, a rounding error in the last place, or a constant outside the search basis. Try adding digits, then dropping the final digit, before concluding no form exists.
What is in the search basis?
Common constants including \pi, e, \sqrt{2}, \ln 2, \varphi, \gamma, \zeta(3) and Catalan's G, algebraic numbers up to degree 4, elementary functions of small rationals, and about 40 named constants such as Feigenbaum's and Khinchin's. Sequences are matched against about 80 named sequences, polynomials, and linear recurrences.
Can it name any integer sequence?
Only ones that fit a polynomial in n, a linear recurrence with constant coefficients, or one of the named sequences. A sequence defined by an unusual rule may return nothing even when a pattern exists to human eyes.
Related tools
Number Base Converter
Binary, octal, decimal, hex and any base from 2 to 36.
Prime Factorization & Number Theory
Prime factors, divisors, GCD/LCM and modular arithmetic.
Scientific Notation & Significant Figures
Convert to scientific and engineering notation, round to sig figs.
Big Number Calculator
Exact arithmetic with thousands of digits: +, −, ×, ÷, ^, fact().
Fraction ↔ Decimal ↔ Continued Fraction
Exact conversions with repeating decimals and convergents.
Roman Numeral Converter
Numbers to Roman numerals and back, with validation.