Z-Score Calculator

About Z-Score Calculator

What Is a Z-Score?

A z-score (also called a standard score) tells you how far a value sits from the mean in units of standard deviation. The formula is:

z = (x - μ) / σ

where x is the raw value, μ is the population mean, and σ is the standard deviation. A positive z-score is above the mean, a negative one is below, and a z of 0 sits exactly at the mean. Because the units are standard deviations rather than raw points, z-scores let you compare values from distributions with completely different scales.

Worked example: On an exam with mean 72 and standard deviation 8, a student scores 88.

1. z = (88 - 72) / 8 = 16 / 8 = 2.0
2. The score is 2 standard deviations above the mean
3. This corresponds to about the 97.7th percentile - only 2.3% of students scored higher

The standard normal distribution itself has mean 0 and standard deviation 1, which is why every z-score is directly comparable. The cumulative distribution function (CDF), written Φ(z), returns the probability that a random variable from that distribution falls below z. This calculator uses the Abramowitz and Stegun 26.2.17 approximation, which is accurate to about 7 decimal places across the full range.

How Do You Read a Z-Score?

The table below shows the most common z-score landmarks, with the percentile (the share of data at or below the score) alongside.

Z-Score	Meaning	Percentile	% of Data Below
-3.0	3 SD below mean	0.13th	0.13%
-2.0	2 SD below mean	2.28th	2.28%
-1.0	1 SD below mean	15.87th	15.87%
0	At the mean	50th	50%
1.0	1 SD above mean	84.13th	84.13%
2.0	2 SD above mean	97.72nd	97.72%
3.0	3 SD above mean	99.87th	99.87%

A practical rule of thumb: a z between -1 and +1 is roughly average, between 1 and 2 is notably above (or below) average, and anything past ±2 is unusual. Past ±3, you are in the top or bottom 0.15% of the distribution.

The 68-95-99.7 Rule (Empirical Rule)

For any normally distributed data set, roughly 68% of values sit within one standard deviation of the mean, 95% within two, and 99.7% within three. This is often called the empirical rule and provides a quick sanity check when you do not have a full z-table to hand.

Range	Z-Scores	% of Data
μ ± 1σ	-1 to +1	68.27%
μ ± 2σ	-2 to +2	95.45%
μ ± 3σ	-3 to +3	99.73%

Values beyond z = ±3 are very rare - roughly 0.3% of data, or about 3 observations in every 1,000. The rule assumes the data is actually normal. For skewed or heavy-tailed data (income, insurance claims, internet traffic) these percentages break down and you need Chebyshev's inequality or a non-parametric approach. For summary statistics on any dataset, the standard deviation calculator shows mean, SD, variance, and range side by side.

How Does Z-Score Convert to Probability?

The probability mode of this calculator returns three quantities for any z-score, derived from the standard normal CDF Φ(z):

Type	Meaning	For z = 1.96
Left-tail P(Z < z)	Area to the left of z	0.9750
Right-tail P(Z > z)	Area to the right of z	0.0250
Two-tailed P(|Z| > z)	Area in both tails beyond ±z	0.0500

The value z = 1.96 is the most cited number in applied statistics because it marks the boundaries of a 95% confidence interval, with 2.5% of the distribution in each tail. In hypothesis testing, a calculated test statistic exceeding ±1.96 in a two-tailed test lets you reject the null at α = 0.05. For the full hypothesis testing workflow, use the p-value calculator, which interprets the tail probability directly against a chosen significance level.

Critical Z-Values for Common Confidence Levels

Confidence intervals and significance tests rely on fixed critical z-values. Memorise 1.96 (95%), 2.576 (99%), and 1.645 (one-tailed 95%) - these three cover the majority of applied statistical work.

Confidence Level	Alpha (α)	z* (two-tailed)	z* (one-tailed)
80%	0.20	1.282	0.842
90%	0.10	1.645	1.282
95%	0.05	1.960	1.645
99%	0.01	2.576	2.326
99.9%	0.001	3.291	3.090

A note on when to use z versus t: z-critical values assume a known population standard deviation, or a sample size large enough (n ≥ 30 is a common cut-off) that the t-distribution has effectively converged to normal. For smaller samples where σ is estimated from the data, use the Student's t-distribution instead.

Comparing Values Across Different Scales

The real power of z-scores is that they let you compare values from distributions with different means and spreads. Raw numbers alone are meaningless when the scales differ - a 78 in English does not mean the same thing as a 78 in maths if the classes have different averages and spreads.

Example: Who performed better relative to their class?

	Student A (Maths)	Student B (English)
Raw score	82	78
Class mean	70	65
Class SD	10	6
Z-score	(82-70)/10 = 1.2	(78-65)/6 = 2.17
Percentile	88.5th	98.5th

Student B finished higher relative to their class (98.5th percentile) even though their raw score was lower. This is exactly how standardised tests work. The SAT, for instance, had a mean total score of 1024 and a standard deviation of 229 for 2024 high-school graduates (per the College Board), so a 1253 sits at roughly z = 1.0, about the 84th percentile.

Reverse Z-Score: Finding the Raw Value

The reverse mode answers the opposite question: given a z-score (or target percentile), mean, and standard deviation, what raw value does that correspond to? The formula is simply the z-score equation rearranged:

x = μ + z × σ

Example: What score sits at the 90th percentile on a test with mean 500 and standard deviation 100?

1. 90th percentile corresponds to z ≈ 1.282 (from the z-table)
2. x = 500 + 1.282 × 100 = 628.2

The calculator uses a rational approximation of the inverse normal CDF (Beasley-Springer-Moro style) so you can type the desired probability directly without consulting a table. This is the mode to use for setting cut-offs: pass marks at the 40th percentile, medical reference ranges at the 2.5th and 97.5th percentiles, or quality control tolerances at ±3σ.

Where Are Z-Scores Used in Practice?

Beyond statistics classrooms, z-scores appear across medicine, finance, education, and industry. The table below lists the most common applications along with the scale typically used.

Field	Application	Typical Scale
Education	Standardised test scoring (SAT, GRE, IQ tests)	SAT mean 1024, SD 229
Medicine	Growth charts for children (height, weight, BMI for age)	WHO/CDC percentile curves
Bone density	T-scores and Z-scores for osteoporosis screening	T < -2.5 = osteoporosis
Finance	Altman Z-score for bankruptcy risk	Z > 2.99 = safe, < 1.81 = distress
Quality control	Six Sigma process capability	6σ = 3.4 defects per million
Research	Hypothesis testing and meta-analysis	|z| > 1.96 for α = 0.05
Psychometrics	IQ scores standardised to mean 100, SD 15	z = (IQ - 100) / 15

Six Sigma is an interesting case: the widely quoted 3.4 defects per million opportunities assumes a process mean that has drifted 1.5 sigma from the specification centre, so the defect rate corresponds to z = 4.5 rather than a pure 6σ (which would give about 2 defects per billion). This "1.5 sigma shift" is Motorola's empirical allowance for long-term process variation.

Common Mistakes With Z-Scores

A handful of mistakes recur in applied work. Most are easy to avoid once flagged.

• Using sample SD when you need population SD. The basic z-score formula assumes σ is known (population). For a sample, you typically want a t-statistic instead, especially when n is small.
• Assuming normality without checking. Z-scores and normal-distribution percentiles are only meaningful if the underlying data is approximately normal. A QQ plot or Shapiro-Wilk test is worth running on any new dataset before quoting percentiles.
• Confusing one-tailed and two-tailed critical values. 1.645 is the one-tailed critical z for α = 0.05, but the two-tailed value is 1.96. Using the wrong one roughly doubles your type I error rate.
• Reading the z-table wrongly. Most tables give P(Z < z) (left-tail cumulative). For right-tail probability, you want 1 - Φ(z), and for two-tailed, 2 × (1 - Φ(|z|)).
• Rounding too aggressively. A z of 1.96 gives p = 0.0500, but 1.95 gives 0.0512 and 1.97 gives 0.0488. Near critical boundaries, carry at least three decimal places through your working.

For probability calculations on other distributions (binomial, Poisson, simple event probabilities), see the probability calculator.

All calculations run entirely in your browser. No data is sent to any server.

Sources

• Wikipedia - Standard Score (Z-Score)
• Wikipedia - Normal Distribution
• Wikipedia - 68-95-99.7 Rule
• Abramowitz & Stegun - Handbook of Mathematical Functions (Normal CDF approximation 26.2.17)
• NCES - SAT Mean Scores and Standard Deviations
• Wikipedia - Six Sigma (3.4 DPMO and 1.5 sigma shift)
• NIST/SEMATECH e-Handbook - Critical Values of the Normal Distribution