After you run a hypothesis test — whether it is an A/B experiment, a regression coefficient, a survey comparison, or a screening model — you end up with a single standardized number: the z-score. That z-score by itself does not tell a decision maker much. What people actually want to know is how unlikely is this result if the null hypothesis is true. That probability is the p-value. This p-value calculator converts any z-score into the tail probability under the standard normal distribution, then compares it to your chosen significance level so you can decide whether to reject the null hypothesis.
Move the sliders and the result updates instantly. You can switch between a one-tailed and a two-tailed test and watch the p-value halve or double, which is the single most common source of mistakes when people read statistical output.
How to Use
- Enter your Z-Score. This is your test statistic after standardizing by its standard error — the number that comes out of most spreadsheet stats functions, regression summaries, and experiment dashboards. It can be positive or negative; the sign only points the direction of the effect.
- Choose the Test Type. Pick Two-tailed if you are checking whether an effect differs in either direction. Pick One-tailed only if you predicted a specific direction ahead of time (for example, "the new page raises conversion").
- Set your Significance Level (alpha), usually 5%. This is the pre-agreed bar: if the p-value lands below alpha, you treat the result as statistically significant.
- Read the result card. It shows the p-value for the test type you chose, the two-tailed and one-tailed values side by side, the cumulative probability up to your z-score, and a plain-language verdict.
The rule is simple: small p-value = surprising under the null = evidence against the null. A p-value of 0.05 means you would see a result this extreme only about 5% of the time if nothing were really going on.
The Formula Behind the Number
The p-value comes from the standard normal (Z) distribution, whose cumulative distribution function is:
Φ(z) = 0.5 × (1 + erf(z / √2))
From that CDF, the two-tailed p-value is the probability of landing outside ±|z|:
p(two-tailed) = 2 × (1 − Φ(|z|))
And the right-tail (one-tailed) p-value is:
p(one-tailed) = 1 − Φ(z)
Worked example: at z = 1.96, Φ(1.96) ≈ 0.9750, so the two-tailed p-value is 2 × (1 − 0.9750) = 0.05. That is why 1.96 is the canonical 5% critical value. At z = 2.58, Φ(2.58) ≈ 0.9951, giving a two-tailed p of about 0.0099 — the familiar 1% threshold. At z = 0, you are dead center of the distribution, so the two-tailed p is exactly 1.0 (nothing is surprising).
The calculator evaluates the standard normal CDF using the Abramowitz-Stegun 7.1.26 polynomial approximation, the same closed form referenced by the NIST/SEMATECH statistical handbook. Its absolute error is below 7.5e-8, so the numbers match textbook z-tables to four decimals and agree with SciPy stats.norm to display precision. Note that this closed form applies to the normal (Z) distribution only; p-values from a t-distribution require a separate t-table lookup by degrees of freedom and are out of scope here.
Two-Tailed vs One-Tailed
This choice changes your answer, so choose it before you look at the data.
- Two-tailed splits alpha across both ends of the bell curve. At a 5% threshold, each tail gets 2.5%, which is why the critical z is 1.96. Use it when you only care whether an effect exists, not which way it points.
- One-tailed puts the whole 5% in one tail, so the critical z drops to about 1.645. It is more sensitive, but it is only legitimate if you predicted the direction in advance. Running a one-tailed test after seeing the sign of your result is a well-documented way to manufacture significance.
Notice the symmetry: z = +1.96 and z = −1.96 give the same two-tailed p-value (0.05). The sign tells you the direction; the p-value tells you the strength of evidence.
Examples
Example 1 — The textbook 5% threshold. Your A/B test returns z = 1.96 at a two-tailed alpha of 5%. The p-value is 0.0500, right on the boundary. Because 0.05 is not strictly less than 0.05, you should treat this as borderline and not celebrate it as a clear win.
Example 2 — A strong result. A regression coefficient comes back at z = 2.58. The two-tailed p-value is 0.0099, comfortably below 1% and well below your 5% alpha. You reject the null: the effect is statistically significant.
Example 3 — No effect. Your experiment returns z = 0.00. The two-tailed p-value is 1.0000. There is no evidence against the null whatsoever; the treatment and control are indistinguishable.
Example 4 — A negative effect. A supplier change lowers error rates with z = −1.96. With a two-tailed test the p-value is still 0.0500 — the minus sign only tells you the change moved in the "better" direction, not that it is weaker.
Example 5 — One-tailed sensitivity. If you pre-registered a one-tailed test at 5% and got z = 1.96, the right-tail p-value is 0.0250, now comfortably significant. The same data crosses the bar only because you committed to the direction in advance.
FAQ
What is a p-value in simple terms?
A p-value is the probability of seeing a result as extreme as yours — or more extreme — if the null hypothesis (the idea that there is no real effect) were true. A small p-value means your data would be surprising under "no effect," which is evidence that the null is wrong. It is not the probability that the null is true, and it is not the size of the effect.
How do I convert a z-score to a p-value?
For a two-tailed test, compute p = 2 × (1 − Φ(|z|)), where Φ is the standard normal cumulative distribution. For a right-tailed test, p = 1 − Φ(z); for a left-tailed test, p = Φ(z). Plug your z-score into this calculator and it returns the exact value instantly.
Is z = 1.96 significant at 5%?
It is right on the boundary. The two-tailed p-value for z = 1.96 is approximately 0.05. By the strict rule "reject if p < alpha," p = 0.05 does not beat alpha = 0.05, so most statisticians treat z = 1.96 as borderline rather than a clean significant result.
Should I use a one-tailed or two-tailed test?
Use a two-tailed test unless you predicted a specific direction before collecting data. Two-tailed is the default in most research and A/B testing because it detects effects in either direction. One-tailed tests are more powerful but must be planned in advance; choosing them after you see the sign of the result inflates your false-positive rate.
What does p < 0.05 actually mean?
It means the observed data would occur less than 5% of the time under the null model, so you reject the null at the 5% significance level. It does not mean there is a 95% chance your effect is real, nor does it tell you how large or important the effect is. Always report the effect size alongside the p-value.
Why is the p-value the same for z = +1.96 and z = −1.96?
Because a two-tailed test looks at both ends of the symmetric normal distribution. The absolute value |z| drives the tail probability, so the sign only tells you the direction of the effect, not its strength. For a two-tailed test, z = +1.96 and z = −1.96 both give p ≈ 0.05.
What if my p-value is exactly 1.0?
That happens when z = 0, meaning your observed statistic sits exactly at the center of the null distribution. There is zero evidence against the null — your result is perfectly consistent with "no effect."
Disclaimer
This p-value calculator is provided for informational purposes only and does not constitute statistical, scientific, medical, or financial advice. It computes tail probabilities under the assumption that your test statistic follows a standard normal (Z) distribution; if your statistic actually follows a t-distribution, a chi-squared distribution, or another model, the results will not apply. Always confirm your test type, sample size, and significance threshold with a qualified statistician before making real decisions.