Z-Score Calculator

Calculate Z-scores, convert probabilities to Z-values, find probabilities from Z-scores, analyze normal distributions, and visualize results using interactive statistical charts.

Z-Score from Raw Value

Raw score (x)

Population mean (μ)

Standard deviation (σ)must be > 0

Empirical Rule — 68 / 95 / 99.7

For any normal distribution, ~68.27% of values fall within ±1σ of the mean, ~95.45% within ±2σ, and ~99.73% within ±3σ — a quick sanity check whenever you read a Z-score.

Standard Normal CDF — Φ(z)

The cumulative distribution function maps every Z-score to a percentile. Each point shows the fraction of the population at or below that Z.

Common Z-Score Interpretations

Z-Score	Percentile	Interpretation
-3	0.135%	Extremely low — far below average
-2	2.275%	Very low — bottom 2–3%
-1	15.866%	Below average — bottom 16%
+0	50.000%	Exactly average — equal to the mean
+1	84.134%	Above average — top 16%
+2	97.725%	Very high — top 2–3%
+3	99.865%	Extremely high — top 0.13%

Probabilities are computed live using the Abramowitz & Stegun erf approximation (Φ accurate to ≈1.5×10⁻⁷) and the Wichura AS241 inverse-normal algorithm (Φ⁻¹ accurate to ≈1×10⁻⁷ across typical inputs). Read more in the methodology and editorial policy.

What Is a Z-Score?

A Z-score — also called a standard score — measures how many standard deviations a value sits away from the mean of its population. It rewrites every raw observation onto a single universal scale where the mean is 0 and the standard deviation is 1, so an exam mark, a body-temperature reading, a stock return, and an IQ score can all be compared on the same footing. Z = 0 sits at the centre of the distribution, positive Z values lie above the mean, and negative Z values lie below it.

This calculator combines three tools on one page: a Z-Score Calculator that derives Z and every standard-normal probability from a raw value, a Z ↔ Probability Converter that lets you start from any probability and recover the matching Z, and a Probability Between Two Z-Scores calculator for confidence-interval and tail-area questions. It pairs naturally with the probability calculator, the mean median mode range calculator, and the scientific calculator.

How the Z-Score Works

Standardise the value

Subtract the population mean μ from the raw value x, then divide by the population standard deviation σ. The result, Z = (x − μ) ÷ σ, measures the value's distance from the mean in units of σ — not in the original measurement units.

Look up Φ(z) on the standard normal table

Φ(z) — the cumulative distribution function of the standard normal — returns the area under the bell curve to the left of z. That area equals the probability of a randomly drawn value being less than x.

Derive every other probability

Once Φ(z) is known, the right tail is 1 − Φ(z), the area between the mean and x is |Φ(z) − 0.5|, the two-sided inside area is Φ(|z|) − Φ(−|z|), and the two-sided outside area is the complement of that. Every standard-normal question follows from a single CDF lookup.

Interpret as a percentile

Multiply Φ(z) by 100 to translate the Z-score into a percentile rank — the percentage of the population scoring at or below that value. Z = 1.00 corresponds to the 84.13th percentile; Z = −1.00 to the 15.87th.

6 Ways to Use This Z-Score Calculator

Standardise any raw score

Drop in a test mark, a measurement, or a financial return with its population mean and standard deviation to get the Z-score and percentile rank.

Read tail probabilities

Answer questions like ‘what fraction of observations exceed this value?’ or ‘what fraction fall below this cutoff?’ in one step.

Invert P → Z

Need the Z behind 0.975 cumulative probability or the 95% two-sided cutoff? The converter accepts any probability and returns the matching Z.

Build confidence intervals

Common confidence levels (90, 95, 99%) all reduce to a two-sided Z probability — enter 0.95 once and read ±1.96 directly.

Compare distributions

When two datasets have different means or units, comparing their raw values is meaningless — comparing their Z-scores is fair and immediate.

Run a hypothesis-test sketch

Compute the test statistic, look up its tail probability, and decide whether it's small enough to reject the null — without flipping through a Z-table.

Best Practices for Working With Z-Scores

A Z-score is only meaningful when the underlying variable is approximately normal. Heavy-skewed data, bimodal data, or data with a long right tail (income, real-estate prices, response times) will produce Z-scores whose ‘percentile’ interpretation is misleading. Inspect a histogram first; if the data looks bell-shaped, Z-scores are a clean summary — if not, prefer median and IQR, or transform the variable (a log transform often rescues right-skewed data).

Be explicit about which mean and standard deviation you are using. Standardising against the sample statistics of a small sample produces a noisy Z that drifts as more data arrives. Standardising against published population values (IQ μ = 100 σ = 15, SAT μ = 500 σ = 100, etc.) gives a stable comparison.

Choose one-tailed or two-tailed probabilities deliberately. A one-tailed Z = 1.96 corresponds to roughly 2.5% in one tail; the same Z corresponds to roughly 5% in both tails combined. Mixing them up is the most common mistake in introductory hypothesis testing.

Why Z-Scores Matter

Education & testing

Standardised tests — SAT, GRE, IQ batteries — report results as Z-scores or percentile ranks so cohorts taking different forms of the same exam can be compared fairly. The whole concept of a ‘percentile’ is the cumulative Z probability.

Finance & risk

Value-at-risk, Sharpe ratio, z-spreads, and credit-score models all depend on standardising returns against historical mean and standard deviation. A 3σ daily move marks a tail event worth investigating — for the same reason, options pricing leans on the standard normal.

Research & science

Lab measurements report results as Z-scores when comparing a sample to a reference population (bone-density T-scores, growth-chart Z-scores in pediatrics, soil-sample assays). Quality-control charts flag any sample beyond ±3σ.

Machine learning

Feature standardisation — subtract the column mean, divide by the column standard deviation — is built into every preprocessing pipeline (scikit-learn’s StandardScaler, the BatchNorm layer, etc.). Z-scores are the unit deep learning works in.

Where Z-Scores Get Tricky

Non-normal data

If the data is skewed, bimodal, or heavy-tailed, the Z-to-percentile mapping breaks down. A Z = 2 in a log-normal distribution is far less rare than the 2.28% the standard normal would suggest.

Sample vs population σ

When σ is estimated from a small sample, the proper sampling distribution is Student's t, not the standard normal. For n < 30, prefer a t-statistic and a t-table over a Z.

One-tailed vs two-tailed

‘95% confidence’ is two-sided (Z = 1.96), but ‘5% one-tail significance’ is one-sided (Z = 1.645). Always state which side you mean and quote the matching critical value.

Continuity correction

When using a normal approximation to a discrete distribution (binomial, Poisson), adding ±0.5 to the value before standardising improves accuracy — especially for small n or extreme probabilities.

The Core Z-Score Formulas

Z-score

z = (x − μ) ÷ σ

Standardise any raw value into σ-units around the mean.

Standard normal CDF

Φ(z) = ½[1 + erf(z/√2)]

Area under the bell curve to the left of z.

Left-tail probability

P(X < x) = Φ(z)

Probability of a value at or below x.

Right-tail probability

P(X > x) = 1 − Φ(z)

Probability of a value strictly above x.

Mean ↔ x area

P(μ < X < x) = |Φ(z) − 0.5|

Area between the centre line and your value.

Between two Z-scores

P(Z₁ < X < Z₂) = Φ(Z₂) − Φ(Z₁)

Used for confidence intervals and band probabilities.

Two-sided inside

P(−|z| < X < |z|) = 2Φ(|z|) − 1

Symmetric area around the mean.

Two-sided outside

P(X < −|z| or X > |z|) = 2(1 − Φ(|z|))

Combined area in both tails — the two-tail p-value.

Inverse normal (probit)

z = Φ⁻¹(p)

Recover the Z-score from any cumulative probability.

Percentile rank

percentile = Φ(z) × 100

Percentage of the population at or below x.

Empirical rule

P(|Z| ≤ 1, 2, 3) ≈ 68.3, 95.5, 99.7%

Quick sanity check for any normal distribution.

Raw value from Z

x = μ + z·σ

Useful for cutoff design — pick a percentile, get the score.

Common Z-Score Mistakes

1
Treating Z-scores as percentages
A Z-score is in σ-units, not percentage points. Z = 1.5 does not mean ‘1.5% above the mean’ — it means ‘1.5 standard deviations above the mean’, which corresponds to the 93.32nd percentile.
2
Mixing one-tailed and two-tailed cutoffs
‘95% confidence’ is two-sided (Z = ±1.96); ‘5% one-tail significance’ is one-sided (Z = 1.645). Pick the right one before reading the critical value.
3
Using sample SD with a small n
Estimating σ from fewer than 30 observations introduces enough noise that the Z-distribution is no longer accurate. Switch to Student's t-distribution.
4
Applying Z to skewed data
Z assumes an underlying normal distribution. For income, response times, or word counts, the Z-to-percentile mapping is wildly wrong — transform first, or use rank-based percentiles.
5
Negative σ
Standard deviation is always non-negative. If you compute a negative σ, you have a bug — the magnitude is correct, but be careful with sign elsewhere in the formula.
6
Confusing Z with t
Use Z when σ is known and n is large; use t when σ is estimated and n is small. Plugging the same value into both calculators gives different probabilities for a reason.

Real-Life Uses of Z-Scores

Standardised testing

Every percentile rank you see on a national exam (SAT, GRE, IQ, GMAT) is computed from a Z-score against a published reference population. Z makes scores from different test forms or different years directly comparable.

Clinical reference

Bone-density T-scores, paediatric growth charts, blood-pressure z-scores, and BMI percentiles for children all express a measurement as standard deviations from a normative reference — the cleanest way to flag clinically meaningful deviations.

Quality control

Six Sigma processes target a ±6σ tolerance around the specification mean — equivalent to 3.4 defects per million opportunities. Control charts flag any sample whose Z exceeds 3 as ‘out of control’.

Feature engineering

Most ML algorithms — linear models, neural nets, k-NN — work better when each input feature has mean 0 and σ 1. Pre-computing per-column Z is a standard first step before any modelling.

Methodology you can verify

Probabilities are computed live using the Abramowitz & Stegun erf approximation (Φ accurate to ≈1.5 × 10⁻⁷) and the Wichura AS241 inverse-normal algorithm (Φ⁻¹ accurate to ≈1 × 10⁻⁷ across typical inputs). No values are hard-coded from a printed Z-table. Read more on the methodology and editorial policy pages.

Frequently Asked Questions

A Z-score — also called a standard score — measures how many standard deviations a value sits away from the mean of its population. It is computed as Z = (x − μ) ÷ σ, where x is the raw value, μ is the population mean, and σ is the population standard deviation. A Z of 0 lies exactly on the mean; positive Z values lie above the mean and negative Z values lie below it. Because Z is a unit-less ratio it lets you compare values from completely different distributions — a test mark, a body-temperature reading, and a stock return — all on the same standardised scale.

Subtract the population mean μ from the raw value x, then divide the result by the population standard deviation σ: Z = (x − μ) ÷ σ. Example: with x = 5, μ = 3, and σ = 2, Z = (5 − 3) ÷ 2 = 1. A Z-score of 1 means the value is exactly one standard deviation above the mean. Look up Φ(z) on a standard normal table — or read it from this calculator — to translate Z into a probability, a percentile rank, and tail areas.

A positive Z-score means the raw value is above the population mean. The larger Z is, the further above the mean the observation lies, and the rarer it becomes in a normal distribution. Z = 1 corresponds to the 84.13th percentile (about 16% of the population sits above it); Z = 2 corresponds to the 97.72nd percentile (only about 2.3% above); Z = 3 corresponds to the 99.87th percentile (just 0.13% above). Positive Z-scores are good news on metrics where higher is better (test scores, income) and warning signs on metrics where lower is better (defect counts, hospital wait times).

A negative Z-score means the raw value is below the population mean. Z = −1 corresponds to the 15.87th percentile (about 16% of the population sits at or below it); Z = −2 corresponds to the 2.28th percentile; Z = −3 corresponds to the 0.13th percentile. The magnitude tells you how unusual the observation is, and the sign tells you which side of the mean it falls on. For two-sided questions, the absolute value of Z is what matters — Z = −1.96 and Z = +1.96 mark the boundaries of the central 95% of a standard normal distribution.

A Z-score of 1.00 corresponds to the 84.13th percentile of the standard normal distribution — meaning 84.13% of the population scores at or below that value, and 15.87% scores above it. This number comes from Φ(1.00) ≈ 0.8413, the cumulative area under the bell curve to the left of z = 1. The same Z works for any normal distribution: a Z of 1 always marks the 84.13th percentile, whether the raw value is an IQ score, an exam mark, or a stock return.

A Z-score of 2.00 corresponds to the 97.72nd percentile — 97.72% of the population scores at or below that value and only 2.28% scores above it. This is computed from Φ(2.00) ≈ 0.9772. A Z of 2 is the rough threshold many fields treat as "very high" or "statistically significant": clinical reference ranges, six-sigma quality control, and most introductory hypothesis tests all use ±2σ as a meaningful boundary. Two standard deviations from the mean captures about 95.45% of any normal distribution — the basis of the popular ‘95% confidence interval’ rule of thumb.

The standard normal distribution is a bell-shaped probability distribution with mean 0 and standard deviation 1, written Z ~ N(0, 1). Every normal distribution can be transformed into a standard normal by subtracting the mean and dividing by the standard deviation — exactly the Z-score formula. The standard normal is the reference distribution for Z-tables, p-values, confidence intervals, and a long list of inferential procedures. About 68% of its area lies within ±1 standard deviation, 95% within ±2, and 99.7% within ±3 — the famous empirical 68-95-99.7 rule.

A Z-table lists Φ(z), the cumulative area under the standard normal curve to the left of a given Z-score. To find a left-tail probability P(X < x), compute z = (x − μ) ÷ σ, then read Φ(z) from the table — for example, Φ(1.00) = 0.8413 means 84.13% of the distribution lies below z = 1. The right-tail probability is 1 − Φ(z), the area between the mean and x is |Φ(z) − 0.5|, and the two-sided inside area is Φ(|z|) − Φ(−|z|) = 2Φ(|z|) − 1. This calculator computes Φ live using the Abramowitz & Stegun erf approximation, so you do not have to consult a printed table.

Tail probability is the area under the bell curve outside a given Z-score. The left tail is P(X < z) = Φ(z); the right tail is P(X > z) = 1 − Φ(z); and the two-sided tail is P(|X| > |z|) = 2 · (1 − Φ(|z|)). Tail probabilities power one-tailed and two-tailed hypothesis tests — small tail probabilities (typically below 0.05 or 0.01) are taken as evidence that the observed value is too extreme to have come from the assumed distribution by chance. The Z ↔ Probability Converter on this page lets you start from any tail probability and recover the matching Z, which is exactly the inverse operation a Z-table performs.

A Z-score is a measurement expressed in units of standard deviation. Z = 1 means ‘one standard deviation above the mean’, Z = −2 means ‘two standard deviations below the mean’, and so on. The standard deviation σ does the unit conversion — it tells you how big a ‘unit’ of variability is for the underlying variable. For an IQ scale with σ = 15, a Z of 2 corresponds to a raw score of 100 + 2 × 15 = 130. To go the other way, the raw value behind a Z is x = μ + Z · σ — useful when you start with a target percentile and need the cutoff score.

Related Calculators

Pair the Z-score with other statistics, probability, and math tools.