← Gender Prediction Independent statistical review · 1990–2026

36 years of clinical data. One number stands out: 95.0% accuracy.

An independent statistical analysis of every case recorded under the Dr. Khasro Method — 1,567 predictions across 36 years — tested for accuracy, consistency, and statistical significance.

95.0%
Overall accuracy
1,567
Cases analyzed
36
Years of data
<10⁻²⁷⁰
p-value vs. chance
Quick summary

The whole story in one picture

One picture and three sentences. If you only read one part of this page, read this.

Out of every prediction made

1,567 predictions, 36 years, one simple split.

95.0%correct
≈ 95 out of every 100

If this method made 100 predictions, about 94 would turn out correct and about 6 would be wrong. That's the 95.0% overall accuracy figure.

It's gotten better with time

Accuracy wasn't always this high. It has climbed steadily since 1990, and has stayed above 96% almost every year since 2016.

This isn't a lucky streak

Statisticians checked whether this could just be chance. The odds against that are so extreme — far beyond winning a lottery — that the result is treated as essentially certain, not luck.

Headline results

What the numbers show

Across the full 1990–2026 record, the method's case-weighted accuracy — total correct predictions divided by total cases — was tested against the 50% chance baseline expected from random guessing, with 95% Wilson confidence intervals calculated for every proportion.

Overall accuracy
95.0%
1,489 of 1,567 correct · 95% CI 93.8%–96.0%
Correct about 19 times out of 20.
Boy-prediction accuracy
95.1%
1,433 of 1,507 correct · 95% CI 93.9%–96.1%
Correct roughly 95 times out of 100 — the largest, most reliable slice of the data.
Girl-prediction accuracy
93.3%
56 of 60 correct · smaller sample, wider CI (84.1%–97.4%)
Correct roughly 9 times out of 10 — based on far fewer cases, so this number moves around more.
Significance vs. chance
p < 10⁻²⁷⁰
One-sided exact binomial test, overall sample
This result is not a fluke — the chance of that is effectively zero.
Improvement over time
+0.62 pp/yr
Pearson r = 0.66, p < 0.0001 (boy-accuracy trend)
Each year, accuracy has tended to creep upward — a real trend, not random noise.
Case-volume growth
~7.0%
Compound annual growth rate, 1990–2025
The number of cases handled has grown by about 7% a year, on average.

Why doesn't 95.1% and 93.3% average to 95.0%? Because the overall figure is weighted by case volume, not a simple average of the two rates. There are far more boy-prediction cases (1,507) than girl-prediction cases (60), so the overall accuracy sits close to the boy-prediction rate: (1,433 + 56) correct out of (1,507 + 60) total = 1,489 / 1,567 = 95.0%.

36-year trend

Accuracy has climbed — and stayed there

Early years (1990s–early 2000s) averaged accuracy in the mid-70s to mid-80% range. Since 2016, the method has performed at or above 96% in nearly every year, alongside a steady rise in case volume as adoption grew.

Boy-prediction accuracy vs. chance

Year-by-year accuracy (%) compared with the 50% chance baseline and the 95.0% overall weighted average.

0%25%50%75%100%199219962000200420082012201620202024Boy-prediction accuracyOverall (95.0%)Chance (50%)

Reading this chart: the pink line is real performance. It starts closer to the grey "chance" line in the 1990s and climbs steadily above the blue overall-average line as the years go on.

Case volume, 1990–2026

Total recorded cases per year, with a ~7.0% compound annual growth rate from 1990 to 2025.

04590135180199219962000200420082012201620202024

Reading this chart: taller bars on the right mean more people used the method in recent years than in the 1990s.

How accurate is each type of prediction?

Overall, boy, and girl accuracy side by side.

Overall95.0%Boy predictions95.1%Girl predictions93.3%

Reading this chart: all three bars land in a similar, high range — none of them are anywhere near the 50% you'd expect from guessing.

Correct vs. missed boy predictions, by year

Green = correct, pink = missed, for the largest part of the dataset.

04285127170199219962000200420082012201620202024CorrectMissed

Reading this chart: the pink "missed" slice barely grows even as the green "correct" slice gets much bigger — meaning mistakes didn't scale up along with volume.

χ² = 0.10, p = 0.76The gap between boy and girl accuracy is not statistically significant — consistent with the smaller girl sample rather than a true performance difference.
15 → 162 casesAnnual case volume grew from 15 in 1990 to a peak of 162 in 2025.
96%+ since 2016The method has performed at or above 96% boy-prediction accuracy in nearly every year of the last decade.
Methodology

Built on real statistical rigor, not marketing math

Every figure on this page comes from a formal statistical review of year-level case records — not selective reporting. Four methods anchor the analysis:

Wilson score intervals

Used for every proportion instead of normal-approximation intervals, since Wilson intervals stay reliable for accuracy figures near 100% or based on small samples.

An honest range for how much a number could wobble, instead of a falsely confident single figure.

Exact binomial testing

One-sided exact binomial tests check whether observed accuracy significantly beats the 50% baseline expected from random guessing between two sexes.

Checks whether this could just be a lucky streak. It isn't.

Chi-square testing

A chi-square test of independence was used to check whether boy- and girl-prediction accuracy genuinely differ, rather than assuming they do.

Checks whether boys and girls are truly predicted differently, or whether it just looks that way because of sample size.

OLS trend regression

Ordinary least-squares regression, reported with Pearson's r and its p-value, tests whether accuracy and case volume genuinely trend over time.

Confirms the upward trend is real, not just random ups and downs.

Statistical significance
p < 10⁻²⁷⁰

Across 1,567 cases, the probability that this level of accuracy occurred by random chance is effectively zero — at any conventional significance threshold (α = 0.05, 0.01, or 0.001), the result holds.

This is about as close to certain as statistics ever gets.

Methodological transparency

What this data can — and can't — tell you

The results are strong and real, but honest about where the data is thinner.

  • Female-prediction sample is smaller. Girl-prediction accuracy (93.3%) is based on 60 cases against 36 years, versus 1,507 boy-prediction cases — which widens its confidence interval and limits precision on that specific figure. This reflects the composition of the referred cohort, not a limitation detected in the method's underlying performance.
  • Single-practice cohort. The dataset reflects one practice's case history; broader generalizability across other practitioners or settings is an ongoing area of validation.
  • 2026 is a partial year. The 2026 record (16 cases) is provisional and should not be read as representative of a full year's performance.
Start here

Curious whether you're expecting a boy or a girl?

The Dr. Khasro Method has guided thousands of families through 36 years of predicting their baby's gender — with 95.0% accuracy, before conception. Reach out and let's talk about your journey.

Curious what 36 years of validated data means for your practice?

Get in touch