The chart said boy. Someone in the family has already bought something blue, the anatomy scan is eleven weeks away, and sitting in the back of your mind is the line you read two tabs ago — up to 90% accurate.
That figure has a sourcing problem. Chinese gender predictor accuracy has been measured three separate times, by research teams with no connection to one another, in Sweden, Massachusetts and Turkey, against a combined 2,896,729 births. All three landed in the same place. It is not 90%.
Who this is for: Anyone holding a prediction and wondering what it is worth, anyone being sent chart screenshots by relatives, and anyone working out why two sites read the same chart and hand back different babies.
We read the three papers rather than the coverage of them. One published its full results table but never converted it into a plain accuracy figure, so we ran that arithmetic ourselves. The number it produced appears nowhere else online.
- It is a coin toss, and that is measured rather than assumed. Three peer-reviewed studies covering 2,896,729 births put accuracy at 50.2%, 50.3% and 50.4%.
- The 93–99% claim has no study behind it. It survives in the medical literature only as a claim the chart makes about itself.
- In one dataset, answering "boy" every time beat the chart. 52.5% of those 17,580 babies were boys. The chart scored 50.3%.
- There is no 2026 edition of the chart. It is one grid, relabelled each year. What changes annually is your lunar age.
- Chinese New Year 2026 fell on 17 February — unusually late. Conceive before that date and you are still in the previous lunar year, which most calculators get wrong.
- Two methods have real numbers behind them: NIPT from around 10 weeks, and the 18–22 week anatomy scan.
How Accurate Is the Chinese Gender Predictor?
The Chinese gender predictor is about 50% accurate. Three peer-reviewed studies covering 2,896,729 births measured it at 50.2%, 50.3% and 50.4% — statistically identical to a coin toss. Promoters of the chart claim 93 to 99%. No published study has found it beating chance.
Two of the three report a kappa statistic rather than a percentage, which is the more honest measure. Kappa asks how much better than luck a method does, on a scale where zero is no skill at all. A forecaster who says "sunny" every day in a desert is right 90% of the time and still scores zero. The chart scored 0.0002 and 0.002.
The three studies, side by side
These are the only large tests in the published literature. Most articles on this topic cite one of them — usually the Swedish one — and stop there.
| Study | Where and when | Births tested | Result |
|---|---|---|---|
| Villamor et al., 2010 | Sweden, 1973–2006 | 2,840,755 | kappa 0.0002 — indistinguishable from guessing |
| Katz & Wylie, 2009 | Massachusetts General, 1995–2008 | 38,394 | 50.4% by western age, 50.2% by lunar age |
| Jafarzade et al., 2023 | Ankara, Turkey, 2017–2022 | 17,580 | kappa 0.002, p = 0.797 — no agreement |
The Swedish team, based at the University of Michigan School of Public Health, did not stop at a total. They checked whether accuracy improved for any subgroup of mothers — by age, education, body mass index, previous children, year of conception. Nothing moved the number. There is no type of woman for whom this chart works better.
The number the 2023 study never printed
The Turkish team published a kappa and a p-value and left it there. They also printed their full contingency table, though — every prediction against every outcome — so the plain accuracy figure sits in the paper waiting to be worked out. Across their 17,580 births the chart called 3,906 girls and 4,931 boys correctly: 8,837 right, or 50.27%. Running their own kappa formula back over those four cells returns 0.0020, matching the 0.002 they published, so the arithmetic checks against the authors' own statistic.
The chart that loses to guessing "boy"
Here the same table turns genuinely unkind. Of those 17,580 babies, 9,230 were boys — 52.5%, a mildly male-tilted sample, which is normal. The chart answered "boy" for 9,375 of them, or 53.3% of the time. It leans male, in a population that happens to lean male.
So run the comparison. Someone who ignored the chart and simply said "boy" to all 17,580 mothers would have been right 52.5% of the time, against the chart's 50.3%. A centuries-old grid performed 2.2 points worse than a stranger with no information and one word in their vocabulary. That is subtraction on a published table, not a rhetorical flourish — and it is the cleanest evidence available that the chart's occasional correctness has nothing to do with the chart.
Where the "90% Accurate" Claim Actually Comes From
Chase it and it leads nowhere. The figure appears on hundreds of sites, always unattributed, usually as "some say." Its one appearance in medical literature is as the thing being tested: the Massachusetts General team opened their paper by noting that the Chinese birth calendar claims 93 to 99% accuracy, then spent 38,394 deliveries finding otherwise.
Healthline still tells readers the method has "a reputation for amazing accuracy — up to 90%," on a page published in 2018 that cites nothing at all.
The claim never dies because nobody has to lie to keep it alive. A 50% method produces a permanent supply of true stories. Half of everyone who uses the chart gets a correct answer, and those are the people who screenshot it and tell their sister; the other half quietly stop mentioning it. Read any forum thread and the pattern shows up within a screen — wins written up in detail, misses buried in a one-line reply. The asymmetry does not require the chart to work. It only requires the chart to exist.
Gender Prediction Methods Ranked by Measured Accuracy
Plenty of pages will tell you a blood test is "very accurate" and the chart is "just for fun." Far fewer attach numbers and sources to each row. Every week count below runs from your last period, so the dates only hold if your dating does — our guide to when to take a pregnancy test covers how early a result means anything, and Nurturepedia's pregnancy due date calculator turns it into weeks.
| Method | Earliest | Measured accuracy | Evidence base |
|---|---|---|---|
| Anatomy scan | 18–22 weeks | 99% in the second trimester | 28 studies, 13,384 women |
| NIPT (cell-free DNA) | ~10 weeks | 95.4% sensitivity, 98.6% specificity | 57 studies, 6,541 pregnancies |
| First-trimester scan / nub | 11–13 weeks | 69% at week 11, 89% at 12, 96% at 13 | Same 28-study review |
| Chinese gender chart | From conception | 50.3% | 2,896,729 births, three studies |
| Fetal heart rate | Any scan | No usable difference exists | Boys 167.3 bpm, girls 167.0 bpm |
| Ring on a string | Any time | Untested and untestable | None |
The heart rate row is the myth most likely to be repeated to you by someone wearing scrubs. Compared in the first trimester, male fetuses averaged 167.3 bpm and female fetuses 167.0 — a gap of three tenths of one beat. Researchers calculated that confirming a difference that small would take over 12,000 participants, and stopped.
The JAMA meta-analysis behind the NIPT row pooled 57 studies, and found the thing clinics rarely mention: testing too early ruins it. Before 7 weeks, sensitivity fell to 74.5%. After 20 weeks it reached 99%. The test is excellent at the right moment, which is not the same as the test being excellent. ACOG's guidance on prenatal genetic screening covers how it is offered.
Why the "99% at 13 weeks" nub claim needs an asterisk
Nub theory sites lean on a single 1999 study reporting 100% accuracy at 13 weeks — one team, a few hundred scans. The 2025 review pooling 28 studies and 13,384 women puts week 13 at 96% and week 11 at 69%, with individual studies ranging from 56% to 98%. That spread is the honest picture, and the number on the leaflet came from conditions your own 12-week scan may not reproduce.
What Lunar Age Means, and Why Two Sites Give You Different Answers
If two Chinese calendar baby gender tools have given you contradictory results, neither is broken. They are disagreeing about your inputs, not the grid.
How to work out your lunar age
- Count how many Chinese New Years have passed since the day you were born.
- Add one, because traditional reckoning makes you a year old on arrival rather than zero.
- Read that total as your lunar age, which usually sits one to two years above your western age.
- Convert the conception date to a lunar month as well — this is the step most calculators skip.
Step four is where 2026 gets interesting. Chinese New Year fell on 17 February 2026, one of the latest dates it can land on. Anyone who conceived between January and mid-February this year is still inside the previous lunar year, in the eleventh or twelfth lunar month. A calculator reading the month off your wall calendar drops them into column one or two instead — a ten-column error on a grid twelve columns wide.
Which explains the steady searches for "lunar age meaning in Urdu" and "lunar age in Pakistan." That is not a translation problem. Someone raised on the Hijri calendar understands lunar months perfectly and still gets this wrong, because Chinese nominal age counts you as one at birth and then advances everyone together at New Year rather than on birthdays. Two babies born six weeks apart can share a lunar age.
For the conversion done properly rather than approximated, Nurturepedia's Chinese Gender Predictor resolves the actual lunar year, handles leap months, and shows the grid cell your answer came from. It will still be a coin toss. It will at least be the right coin toss.
Can the Chinese Gender Predictor Be Wrong?
Yes, and the expected rate is roughly half. Not "occasionally wrong" or "wrong if you mistyped a date" — wrong about as often as it is right, every time, for everyone, regardless of how many friends it got correct.
Worth saying plainly, because most parenting sites step around it: in families where a son is hoped for more openly than a daughter, a chart that says "boy" is not neutral entertainment. It starts a countdown somebody has to absorb later. If the prediction is already circulating in your family as though it were information, the studies above are useful to have — a 50.3% method has no standing to set anyone's expectations.
What to Do With the Answer You Already Got
Two questions settle this faster than any amount of reading.
Would you be disappointed if it turned out wrong? If yes, you are not treating it as a game, and the honest move is to ask your provider about NIPT timing instead. If no, enjoy it — folk traditions owe nobody predictive validity, and this one entertained families for generations without ever claiming to be a test.
How long until you actually know? With the anatomy scan four weeks out, the chart is a way to pass four weeks. At week 7 there is a real wait ahead, and filling it with a 50.3% answer tends to make the wait longer rather than shorter. Better hours go into baby name trends for 2026, where both shortlists stay useful whichever way the scan goes.
One last thing the three studies agree on, quietly. Across 2,896,729 births in three countries over fifty years, nobody has found a single subgroup — no age band, no BMI range, no parity, no decade — where this chart beats chance. That is a thorough failure, and the strongest argument for using it exactly as it deserves: for fun, out loud, and without letting anyone write it down.
