Z-Score Calculator: Percentile, P-Value & Data Set

Your 78 in chemistry was the better result, even though your history test came back at 85. The chemistry class averaged 70 with a standard deviation of 4, so 78 sits 2 standard deviations above the mean. History averaged 80 with a standard deviation of 10, so 85 is only half a standard deviation above. That distance from the mean, counted in standard deviations, is the z-score.

The calculator below finds a z-score from a score, turns a z-score into a percentile or p-value, works backwards from a percentile, and scores a whole list of values at once. Every answer comes with the steps and a shaded normal curve.

Z-Score Calculator

Type the score, the mean and the standard deviation in the same units. Decimals and negatives are fine.

z-score –
Normal curve (shaded area)
Step by step

    78 in Chemistry or 85 in History? Putting Two Scores on One Scale

    The formula is short:

    z = (x − μ) ÷ σ

    Here x is the score, μ (mu) is the mean and σ (sigma) is the standard deviation. Subtracting the mean tells you how far the score is from average. Dividing by the standard deviation turns that distance into a count of standard deviations.

    ChemistryHistory
    Your score (x)7885
    Class mean (μ)7080
    Standard deviation (σ)410
    x − μ85
    z = (x − μ) ÷ σ8 ÷ 4 = 25 ÷ 10 = 0.5
    Share of the class below you (if scores are roughly normal)97.72%69.15%

    The raw scores point one way and the z-scores point the other. In chemistry, an 8-point gap is a lot: if the scores are bell-shaped, about two-thirds of the class would be within 4 points of 70. In history, the scores were spread out, so 5 points above average is nothing special.

    The formula also runs backwards. To match your chemistry result in history, you would have needed a z of 2 there too: x = μ + z × σ = 80 + 2 × 10 = 100.

    That is the whole point of a z-score: it strips away the units and the scale. Test points, kilograms, seconds and dollars all become “standard deviations from the mean”, so you can compare things that were measured differently.

    A z-score doesn’t have to be a whole number. With a mean of 80 and a standard deviation of 6, a 90 gives z = 10 ÷ 6 = 5/3 ≈ 1.6667. The calculator shows the fraction when the decimal repeats.

    What the Sign and the Size Tell You

    The sign says which side of the mean you are on. A positive z is above average, a negative z is below, and z = 0 is exactly average. The size says how far.

    For data that follow a normal (bell-shaped) distribution, the share of values within each distance is fixed:

    Distance from the meanShare of values insideShare outside (both tails)
    Within ±1 (−1 < z < 1)68.27%31.73%
    Within ±295.45%4.55%
    Within ±399.73%0.27%

    This is the 68–95–99.7 rule. It gives a quick way to judge a z-score:

    • |z| under 1: ordinary. About two values in three are this close to the mean.
    • |z| between 1 and 2: noticeable, but about 27% of values land here.
    • |z| between 2 and 3: unusual. Only about 4.3% of values are this far out on either side.
    • |z| of 3 or more: rare in a normal distribution (about 0.27%). Check for a typing error before treating it as real.

    The calculator places your z-score in one of these bands and says so just below the result.

    From a z-Score to a Percentile

    A percentile is the share of values below a score. On a normal curve, it is the area to the left of z.

    Wechsler IQ tests, for example, report full-scale scores on a scale with a mean of 100 and a standard deviation of 15. A score of 115 has z = (115 − 100) ÷ 15 = 1. The area to the left of z = 1 is 0.8413, so 115 is at about the 84th percentile. A score of 130 (z = 2) is at about the 98th percentile: only 2.28% of scores are higher. A score of 70 (z = −2) mirrors it, with 2.28% of scores lower.

    Use the z → area tab for this. Type the z-score, pick “to the left”, and the answer is the percentile as a decimal. The calculator also shows the area to the right, which is the share of values above you. With a list of actual values instead of a mean and SD, the percentile calculator ranks them directly.

    A percentile is not a percentage score. An 84th percentile means you beat 84% of the group, not that you earned 84% of the marks. The test score calculator handles the second kind.

    Reading a Z-Table: Three Tables, Three Different Numbers

    Before calculators, people looked areas up in a printed z-table. If your course still uses one, know that there are three common kinds, and each gives a different number for the same z.

    Take z = 1.12:

    Type of tableWhat it showsValue at z = 1.12
    Cumulative (“less than”)Area to the left of z0.8686
    Mean to z (“0 to z”)Area between 0 and z0.3686
    Right tail (“greater than”)Area to the right of z0.1314

    The three are tied together. For a positive z, mean-to-z is the cumulative value minus 0.5, and the right tail is always 1 minus the cumulative value. Check the first few rows of your table. If the value at z = 0.00 is 0.5000, it’s cumulative. If it’s 0.0000, it’s a mean-to-z table.

    To look up 1.12, go down to the row for 1.1 and across to the column for 0.02. The row gives the first decimal and the column gives the second.

    Negative z-scores use symmetry. A mean-to-z table only lists positive values, so for z = −0.85 look up 0.85 instead. For the area to the left of −0.85, take the right tail of +0.85: 1 − 0.8023 = 0.1977.

    A printed table has to stop somewhere, since the areas far out in the tails are tiny. The calculator keeps going and shows them in scientific notation. For z = 5, the area to the right is 2.867 × 10⁻⁷, about 3 in 10 million.

    Printed tables round z to two decimal places, so a table answer can be slightly off for z = 1.1247. The calculator works with the exact z. Its steps still tell you which row and column to use, so you can check your table work.

    One Tail or Two? z-Scores as p-Values

    In a hypothesis test, the z-score of your result becomes a p-value: the chance of a result at least that extreme if nothing unusual is going on. Which area you use depends on the question.

    For z = 2.1:

    • One-tailed (upper): area to the right = 0.0179. Use it when the question is “is it higher?”.
    • Two-tailed: both tails = 2 × 0.0179 = 0.0357. Use it when the question is “is it different?”, in either direction.

    Both are below 0.05, so the result counts as significant at the 5% level either way. With a z of 1.8, the one-tailed p would be under 0.05 and the two-tailed one over it, which is why you settle on the type of test before looking at the data. For chances that don’t come from a normal curve, such as cards, dice or repeated tries, use the probability calculator.

    The cut-offs worth remembering:

    LevelOne-tailed cut-offTwo-tailed cut-off
    10%1.2816±1.6449
    5%1.6449±1.9600
    1%2.3263±2.5758

    When you test a sample mean instead of a single value, divide by the standard error, σ ÷ √n, not by σ itself:

    z = (x̄ − μ) ÷ (σ ÷ √n)

    A sample of 36 people with a mean IQ of 103, against μ = 100 and σ = 15, gives a standard error of 15 ÷ 6 = 2.5 and z = 3 ÷ 2.5 = 1.2. The two-tailed p-value is 0.2301, far from significant. A single person at 103 would have z = 0.2. The sample mean has the bigger z, because averages of 36 people vary much less than individuals do. Work out z with the formula above, then use the z → area tab for the p-value.

    Working Backwards: From a Percentile to a Score

    Sometimes you know the share and need the cut-off. Say a course gives an A to the top 10% and scores have a mean of 70 and a standard deviation of 8.

    1. The top 10% means 90% of scores are below the cut-off. The Percentile → z tab turns 90% into z = 1.2816.
    2. Turn z back into a score with x = μ + z × σ = 70 + 1.2816 × 8 = 80.25.

    Using 1.645 here would be a mistake: that is the 95th percentile. It marks the top 5%, so it would set the cut-off at 83.16 and give A’s to half as many students. The 90th percentile is 1.2816.

    The same tab covers a middle range. For the middle 95% of IQ scores, pick “in the middle” and type 95%. The cut-offs are z = ±1.96, which is 100 ± 1.96 × 15: from 70.60 to 129.40. For a two-tailed significance level, pick “in both tails” and type 0.05. You get the same ±1.96.

    Sample or Population Standard Deviation?

    When the mean and standard deviation come from your own list of values, you calculate them first, and the choice of standard deviation changes every z-score.

    Take ten daily delivery times in minutes: 22, 25, 19, 31, 24, 27, 23, 26, 20 and 63.

    • The mean is 280 ÷ 10 = 28.
    • The sample standard deviation (divide the squared distances by n − 1 = 9) is 12.7802.
    • The population standard deviation (divide by n = 10) is 12.1244.

    The 63-minute delivery gets z = 35 ÷ 12.7802 = 2.7386 with the sample figure, and 35 ÷ 12.1244 = 2.8868 with the population figure.

    Use the sample version when your list is a sample from something bigger, such as ten days out of a year. Use the population version only when the list is the whole group, such as every student in one class with no others counted. The standard deviation calculator explains why the n − 1 version is the fair estimate.

    The Whole data set tab does all of this in one go. It scores every value, highlights the one furthest from the mean and flags anything 2 or more standard deviations out. You can switch between sample and population and watch the z-scores change.

    Two checks hold for any list, whichever standard deviation you choose. The z-scores add up to 0. Their squares add up to n − 1 with the sample standard deviation, or to n with the population one.

    Why the |z| > 3 Rule Can’t Catch an Outlier in a Short List

    The 63 in the delivery times is clearly suspicious: every other value is between 19 and 31. Yet its z-score is only 2.74, so a “flag anything above 3” rule lets it through.

    That isn’t bad luck. In a short list, no value can get a large z-score, because an extreme value pulls the mean towards itself and inflates the standard deviation at the same time. The most any single value can reach is fixed by the list length:

    Values in the list (n)Largest possible |z|, sample SD: (n − 1) ÷ √nLargest possible |z|, population SD: √(n − 1)
    51.792.00
    82.472.65
    102.853.00
    113.023.16
    204.254.36

    With ten values and the sample standard deviation, no value can ever exceed 2.85, whatever you type. You need at least 11 values before a z-score above 3 is even possible.

    For short lists, compare the value with the rest instead. Take the 63 out and the other nine have a mean of 24.11 and a sample standard deviation of 3.69. Measured against them, 63 is (63 − 24.11) ÷ 3.69 ≈ 10.5 standard deviations above the mean. The calculator states the limit for your list length, so you know when the rule can’t work.

    When a z-Score Doesn’t Give the Right Percentile

    You can calculate a z-score for any data. The step that needs care is turning it into a percentile or a p-value. That step uses the normal curve, so it only gives the right answer when the data are roughly bell-shaped.

    Skewed data break it. Delivery times, incomes and house prices tend to have long right tails: a few very large values. For data like that, the share of values above z = 2 can be much more than the normal curve’s 2.28%.

    There is one guarantee that holds for any data at all. Chebyshev’s inequality says that at least 1 − 1/k² of the values lie within k standard deviations of the mean. So at least 75% are within 2 standard deviations, and at least 88.9% are within 3. The normal curve promises much more (95.45% and 99.73%), but only for normal data.

    If your data look lopsided, rank them and count to find the real percentile. Treat the normal-curve figure as an approximation.

    z-Scores in Growth Charts and Bone Scans

    Two places where a z-score decides what happens next:

    • Children’s growth. The World Health Organization’s Child Growth Standards measure a child’s height against the median for their age and sex, in standard deviations. A height-for-age more than 2 standard deviations below the median counts as stunting. On a normal curve, z = −2 leaves about 2.3% of the reference group below it, and z = −3 leaves about 0.13%.
    • Bone density scans. A DXA scan reports two scores. The T-score compares your bone density with a healthy young adult; a T-score of −2.5 or lower is classified as osteoporosis. The Z-score compares you with people of your own age and sex. The International Society for Clinical Densitometry uses Z-scores for children, premenopausal women and men under 50, and calls a Z-score of −2.0 or lower “below the expected range for age”.

    In both, the z-score answers the same question as the chemistry test: how unusual is this measurement compared with the right group?

    Doing the Same Thing in Excel or Google Sheets

    Spreadsheets have the pieces built in:

    • =STANDARDIZE(x, mean, standard_dev) returns the z-score. Microsoft’s own example standardizes 42 with a mean of 40 and a standard deviation of 1.5, which gives 1.3333.
    • =NORM.S.DIST(z, TRUE) returns the area to the left of z, which is the percentile.
    • =NORM.S.INV(probability) goes backwards from an area to z. =NORM.S.INV(0.9) gives 1.2816.
    • =NORM.DIST(x, mean, standard_dev, TRUE) skips the z-score and gives the percentile of a raw score directly.

    For a whole column, put the mean in one cell with =AVERAGE(A2:A11) and the standard deviation in another with =STDEV.S(A2:A11) (or STDEV.P for a population). Then fill =STANDARDIZE(A2, $C$1, $C$2) down the column. Google Sheets accepts the same function names, so the formulas work there unchanged.

    Other Things Called a “Z-Score”

    The name turns up in places where it means something else, so check which one you need:

    • Altman Z-score is a credit-risk formula that combines five financial ratios to estimate the chance of a company going bankrupt. It is not a distance from a mean, and this calculator doesn’t compute it.
    • T-scores in bone density, mentioned above, are z-scores measured against a young-adult reference instead of your own age group. Some psychological tests also use “T-scores” on a different scale, with a mean of 50 and a standard deviation of 10, so a T of 60 is z = 1.
    • Standard scores on many school and IQ tests are z-scores rescaled to a friendlier mean and standard deviation (100 and 15 on Wechsler composites, 10 and 3 on its subtests). Divide back to get z: a subtest score of 13 is (13 − 10) ÷ 3 = 1.

    Whatever the label, if it is a distance from a mean measured in standard deviations, the formula and the calculator above apply.