Sigma Calculator — 543+ Free Calculator Tools
Convert any z-score to a p-value for two-tailed or one-tailed tests — with significance flags.
The P-Value Calculator above turns a z-score into a p-value in seconds. Researchers, students, and analysts use it to decide whether a result is statistically significant without looking up values in a printed z-table. Follow these steps for a correct, defensible answer every time.
One habit separates careful analysts from sloppy ones: decide on one-tailed vs. two-tailed before you see the data. Switching to a one-tailed test after noticing the direction of the effect doubles your false-positive rate and invalidates the p-value.
Before you trust any p-value, sanity-check the inputs with both tail options. A z-score of 1.80 gives a one-tailed p of 0.0359 but a two-tailed p of 0.0719 — significant under one choice, not significant under the other. If flipping the test type flips your conclusion, say so openly and report both numbers; reviewers respect transparency far more than a barely-significant star. Also confirm the sign convention: for a one-tailed right test, positive z-scores produce small p-values, while negative z-scores produce p-values near 1. A negative z with a right-tailed test (say z = −2.10, p = 0.9821) simply means the data lean the other way — the evidence points against your directional hypothesis, not for it.
A p-value is the probability of observing results at least as extreme as yours, assuming the null hypothesis is true. The null hypothesis is the default position — usually “there is no effect” or “there is no difference.” The alternative hypothesis is what you hope to show instead. A p-value of 0.03 therefore means: if the null were true, you would see data this extreme only 3% of the time. It does not mean there is a 97% chance your hypothesis is true — that is the single most common misreading in all of statistics.
The p-value framework was developed by R. A. Fisher in the 1920s, and his suggested significance level of α = 0.05 became the convention across science. The test statistic (here, the z-score) measures how far your data fall from the null expectation in standard-deviation units. A two-tailed test counts extremes in both directions; a one-tailed test counts only one. Choosing a 5% significance level means accepting a 5% Type I error rate — rejecting a true null — while a Type II error is failing to detect a real effect. When the p-value falls below α, the result is called statistically significant.
P-values appear everywhere decisions must be justified with data: clinical trials testing new drugs, A/B tests comparing website designs, psychology experiments, quality-control sampling in factories, and polling analysis. In 2016 the American Statistical Association issued a landmark statement reminding scientists that a p-value alone does not measure the size or importance of an effect — a tiny p-value from a huge sample can accompany a practically meaningless difference, and a large p-value does not prove the null is true. Used honestly, though, the p-value remains the shared language for asking “could this just be chance?”
The reason p-values dominate applied science is practical: they compress an entire experiment into a single number that anyone can interpret against the same 0.05 yardstick. But that compression discards information, which is why modern guidelines — from the ASA statement to journal checklists — insist on reporting effect sizes and confidence intervals alongside p-values. A useful rule of thumb: the p-value answers “should I be surprised?”, while the effect size answers “should I care?” You need both answers before making a decision, and this calculator pairs well with the Confidence Interval Calculator linked below for exactly that reason.
The p-value for a z-score comes from Φ, the cumulative distribution function (CDF) of the standard normal distribution — the familiar bell curve with mean 0 and standard deviation 1. Φ(z) is the area under the curve to the left of z. The three formulas are:
Worked example 1 — z = 1.96, two-tailed. The standard normal table gives Φ(1.96) = 0.9750. The upper tail is 1 − 0.9750 = 0.0250, and doubling it for both tails gives p = 0.0500. The calculator above shows 0.049996 because it uses a more precise Φ(1.96) = 0.975002 — the same answer to any practical precision. Since 0.049996 < 0.05, z = 1.96 is just barely significant at the 5% level, which is why 1.96 is the most famous number in statistics.
Worked example 2 — z = 2.58, two-tailed. Here Φ(2.58) = 0.99506. The upper tail is 1 − 0.99506 = 0.00494, and doubling gives p = 0.00988 ≈ 0.0099. This is below 0.05 and below 0.01, so the result is significant at the 1% level — strong evidence against the null. Compare the two examples: moving from z = 1.96 to z = 2.58 shrank the p-value fivefold, because tail areas fall off exponentially as you move away from the center of the bell curve.
For a one-tailed illustration, take z = 2.33 right-tailed: Φ(2.33) = 0.9901, so p = 1 − 0.9901 = 0.0099 — significant at 1%. The identical z-score two-tailed would give p = 0.0198, still significant at 5% but not at 1%. The test type genuinely changes the conclusion, which is why it must be pre-specified.
Two more reference points are worth memorizing. First, z = 1.645 gives a one-tailed p of exactly 0.05 (Φ(1.645) = 0.95), which is why 1.645 is the critical value for one-sided 95% confidence bounds. Second, z = 0.67 gives a two-tailed p of about 0.50: Φ(0.67) = 0.7486, the one-sided tail is 0.2514, and doubling gives 0.5028. In other words, a result two-thirds of a standard deviation from the mean is a coin flip under the null — exactly what intuition suggests. Try both in the calculator above and confirm: 1.645 one-tailed right gives 0.05, and 0.67 two-tailed gives 0.5028. These landmarks make it easy to spot implausible p-values in software output.
A marketing team runs an A/B test on a checkout button. Version B converts 4.1% of 20,000 visitors while version A converts 3.8% — a difference that produces z = 2.05. The two-tailed p-value is 2 × (1 − Φ(2.05)). With Φ(2.05) = 0.9798, the upper tail is 0.0202 and p = 0.0404. Because 0.0404 < 0.05, the team declares version B the winner and rolls it out.
Now suppose the test had stopped a day earlier, with z = 1.90: Φ(1.90) = 0.9713, upper tail 0.0287, p = 0.0574 — not significant at 5%. The underlying reality did not change between the two analyses; only the sample size did. This is the “cliff effect” of the 0.05 threshold: 0.0499 and 0.0501 are practically identical evidence, yet one is labeled a discovery and the other a failure. Serious analysts report the exact p-value (0.0404, not just “p < 0.05”) and the effect size — here a 0.3 percentage-point lift, worth about 60 extra orders per 20,000 visitors — so readers can judge practical importance for themselves.
There is also a sample-size moral here. With 20,000 visitors per variant, a 0.3-point lift reaches z = 2.05 and significance; with only 2,000 visitors per variant, the same 0.3-point lift gives z ≈ 0.65 and p ≈ 0.52 — nowhere near significant. The effect did not change, only the precision of its measurement. This is why power analysis comes before data collection: to detect a 0.3-point lift at 80% power, you need roughly 15,000–20,000 visitors per arm. Running the test with 2,000 visitors was never going to answer the question, significant or not. When planning your own test, size the sample first so the p-value you compute is capable of being meaningful.
A pharmaceutical company tests a new blood-pressure drug against a placebo in 400 patients. The mean reduction is 4.2 mmHg better than placebo with a standard error of 1.8, giving z = 4.2 / 1.8 = 2.33. For a one-tailed (right) test, p = 1 − Φ(2.33) = 1 − 0.9901 = 0.0099. The correct interpretation: if the drug were no better than placebo, results this favorable would occur only about 1% of the time — so the company rejects the null and proceeds to a larger trial.
The incorrect interpretations are tempting but wrong: it is not a 99% chance the drug works, and it is not proof of a large effect. A second trial with 50,000 patients might find a statistically significant 0.4 mmHg reduction (p = 0.002) that no doctor considers clinically meaningful. Always pair the p-value with the estimated effect (4.2 mmHg) and a confidence interval — significance tells you an effect probably exists, but only the effect size tells you whether it matters.
Regulators understand this distinction, which is why drug approval requires both statistical significance and a clinically meaningful effect size. The 4.2 mmHg reduction in our example clears both bars: p = 0.0099 says the effect is real, and 4.2 mmHg says it matters to patients. A result with p = 0.0099 but an effect of 0.4 mmHg would clear only the first bar — and would likely fail to win approval. Whenever you read a study, check both numbers before updating your beliefs, and be skeptical of headlines that trumpet significance without mentioning the size of the effect.
The five errors below account for the large majority of misused p-values in published research. Each is easy to commit and easy to avoid once you know what to watch for — read them before you run your next test.
These Sigma Calculator tools pair naturally with p-value analysis:
This page provides a free P-Value Calculator that converts a z-score to a p-value for two-tailed, one-tailed right, and one-tailed left tests using the standard normal CDF Φ. Formulas: two-tailed p = 2(1 − Φ(|z|)); one-tailed right p = 1 − Φ(z); one-tailed left p = Φ(z). Verified reference values: z = 1.96 two-tailed → p ≈ 0.05 (0.049996); z = 2.58 two-tailed → p ≈ 0.0099; z = 1.645 one-tailed right → p = 0.05; z = 2.33 one-tailed right → p = 0.0099; z = 0 → p = 1.0. The conventional significance level is α = 0.05; p < α rejects the null hypothesis. A p-value is P(data at least this extreme | null true), not P(hypothesis true). All calculations run locally in the visitor’s browser with no signup.
Sources: R. A. Fisher, “Statistical Methods for Research Workers”; the American Statistical Association’s 2016 statement on p-values; standard normal distribution tables. Content reviewed for accuracy in October 2026.
Enter your z-score and choose two-tailed, one-tailed right, or one-tailed left. The calculator converts the z-score through the standard normal distribution and reports the p-value plus whether it is significant at α = 0.05 and α = 0.01.
Formula: Two-tailed: p = 2 × (1 − Φ(|z|)), where Φ is the standard normal CDF. Example: z = 1.96 two-tailed → p ≈ 0.049996, just significant at α = 0.05.
The probability of seeing a result at least as extreme as yours if the null hypothesis were true. Small p-values (typically below 0.05) suggest the result is statistically significant.
By convention, p < 0.05 is significant and p < 0.01 is highly significant. The calculator above flags both thresholds for your z-score.
Two-tailed tests look for an effect in either direction and double the tail area; one-tailed tests look in a single direction. Two-tailed is the safer default.
About 0.05 two-tailed — that is exactly why 1.96 is the critical value for 95% confidence intervals.
No — it only says the data is unlikely under the null hypothesis. Effect size, sample size, and study design matter just as much.
No. A p-value is a probability, so it always lies between 0 and 1. A z-score of 0 gives p = 1.0 (two-tailed) — the data are perfectly consistent with the null — and no z-score can produce a p-value below 0. If you ever compute one outside this range, check your formula: you may have forgotten to subtract from 1 or to double the tail.
It is a convention proposed by R. A. Fisher in the 1920s, not a law of nature. Fisher suggested 1 in 20 as a convenient threshold, and it stuck. Some fields use stricter cutoffs — particle physics demands 5-sigma evidence (p ≈ 0.0000003) — while exploratory studies sometimes report 0.10. What matters is choosing alpha before the experiment and reporting exact p-values.
Two-tailed, it is approximately 0.0099: the standard normal CDF at 2.58 is 0.99506, so each tail is 0.00494 and doubling gives 0.00988. This is significant at both the 5% and 1% levels, which is why 2.58 is the critical value for 99% confidence intervals. One-tailed, the same z gives p ≈ 0.0049.
Also try our Z-Score Calculator.