Percentile to z-score calculator

Give it a percentile strictly between zero and one hundred, and it returns both the equivalent probability and z-score. It is useful for interpreting normal-distribution cutoffs, checking statistical tables, setting one-sided thresholds, and moving between percentile language and standard deviations from the mean without manually searching a table.

What a percentile means on the standard normal curve

A percentile describes cumulative area to the left of a point. For example, the 50th percentile divides the standard normal distribution in half, so its z-score is zero. A percentile above 50 produces a positive z-score, while one below 50 produces a negative z-score. This calculator interprets the entered value on a zero-to-one-hundred scale and divides it by one hundred to obtain a probability. It then solves the inverse standard normal cumulative distribution function: instead of starting with a z-score and finding an area, it starts with the area and finds the z-score. The result therefore answers a precise question: how many standard deviations above or below the normal mean marks the requested cumulative percentile? This direction matters because an ordinary z-score calculator usually starts with a raw observation, mean, and standard deviation. Here, no raw data or distribution parameters are needed because the reference distribution already has mean zero and standard deviation one. The returned probability is included so you can confirm the exact scale used in the conversion.

How to enter and interpret the result

Enter a finite numeric percentile strictly greater than 0 and strictly less than 100. Decimal percentiles are supported, so values such as 2.5, 90, 95, or 99.9 can represent common tail cutoffs. The response includes `percentile`, `probability`, and `z_score`. The sign of `z_score` tells you which side of the mean contains the cutoff, and its magnitude gives the distance from the mean in standard-deviation units. Symmetric percentiles have z-scores with opposite signs: the 5th and 95th percentiles, for example, sit equally far from zero on opposite sides. Do not interpret the result as the probability between zero and the z-score; it is based on cumulative probability from the far left tail. If your source uses right-tail probability instead, convert it to a left-tail percentile first by subtracting that probability from one, then multiplying by one hundred. Keeping the tail convention explicit prevents a plausible-looking answer with the wrong sign.

Accuracy, endpoints, and practical boundaries

The calculation uses a deterministic, piecewise rational approximation to the inverse standard normal distribution. Separate formulas handle the center and tails, where a single polynomial would lose useful accuracy. The returned z-score is rounded to twelve significant digits, which is more precision than most statistical tables and routine threshold calculations require while keeping the JSON result stable. The exact 0th and 100th percentiles are rejected even though they look like boundary values on the percentage scale. Mathematically, their standard normal quantiles are negative and positive infinity, not finite z-scores that can be represented as ordinary JSON numbers. Values outside the zero-to-one-hundred interval, nonnumeric text, missing fields, NaN, and infinity are also invalid. Extremely tail-heavy but finite percentiles may be limited by ordinary floating-point representation, so retain only the precision justified by your source percentile. The browser calculation and API handler share the same pure implementation, with no network lookup, random values, or time-dependent behavior. API use costs $0.002 per request.