The zodiac sign of champions is Aquarius, Libra or Taurus: 730,728 athletes, and the rulebook that decides

Key findings

• In countries where youth football is grouped by calendar year, a boy born under Aquarius is 2.4 times as likely as a Sagittarius to become a professional footballer. Among 50,649 men born 1990–2007 in 32 such countries, Aquarius has 1.44× the expected number of players and Sagittarius 0.60×: the oldest children of each age group are born in January and February, the youngest in November and December.
• The “champion sign” is wherever the cutoff puts it. English youth football follows the school year from 1 September: the top sign is Libra (1.52×). Japan’s school year starts on 2 April: Taurus (1.52×). Under 1 January it is Aquarius.
• When the rule changed, the sign changed. German footballers born before 1980 were, on average across three-year cohorts, 69% more often born in August–October than in May–July; those born after 1989 are 133% more often born in January–March than in October–December. The Netherlands and Belgium flipped the same way. These countries were in our first look at the data, so the switch is an observation, not a prediction; the one switch we predicted blind, Australia’s, did not show (only 199 players in the early cohorts).
• We wrote 56 predictions down before counting; 44 of 51 directional predictions came true, and 40 of 47 named the winning sign. Football in all 24 unopened calendar-year countries, from Argentina to Ukraine, came out as predicted (once the direction holds, the winning sign nearly follows, so that is one result, not two). The misses are listed below.
• Famous people outside sport barely show it. Against official births, politicians, writers, actors, painters, musicians and researchers are within 4% of the expected count for a 1 January or a 1 September cutoff — a small fraction of the effect in football.
• Sorting by cutoff, not by sign, makes the stars unnecessary. Lined up by months since their own cutoff date, all 66 sport-and-country groups with at least 200 players fall on one curve, from 1.43× in the first month to 0.66× in the last. For the 16 of them whose cutoff is not 1 January, the real dates give 1.75×; wrong dates drawn at random, 10,000 times, gave at most 1.33×. Some of those cutoffs were chosen after our first look; the 9 predicted blind alone give 1.45× against at most 1.32×.

One sport, four champion signs

Children compete in age groups, and every group has a cutoff date. A child born just after it is the oldest of the group, almost a year older than one born just before it — at eight, that is an eighth of a lifetime of growth. Coaches pick the bigger, faster, more practised children for the better teams, those get more coaching, and the head start compounds. Psychologists call it the relative age effect; Malcolm Gladwell made it famous with Canadian hockey players in Outliers. A Sun sign is a birth date range, so it inherits the effect — and the “sign of champions” depends on where each country puts its cutoff.

Men only; football except the US baseball wheel. “Calendar-year countries” pools 32 countries, players born 1990–2007. A sign straddling the cutoff is mixed: Capricorn (22 December – 19 January) holds both the youngest children of one group and the oldest of the next, which is why Aquarius, wholly after 1 January, wins.

When the rule changed, the sign changed

A pattern that follows signs would stay put. A pattern that follows the rulebook should move when the rulebook does. In Germany and Belgium the federations moved the date in 1997, and the peak moved with it; Dutch players show the same flip, though we found no published date for the Dutch change. The chart compares two ways of being “old for your group”: the gold line follows the old cutoff, the blue line the new one. Each point is three birth years.

In Germany the two lines cross in the birth years of the switch: before 1980 the August–October children averaged 1.69× the May–July children and January–March was at 0.72×; after 1989 January–March averages 2.33× and August–October 0.76×. A Virgo born in 1975 had the head start; a Virgo born in 1995 had the handicap. The crossover comes with players born in the mid-1980s, who were 11 to 14 when the rule changed: the selection that decides careers — youth academies, regional and national youth teams — happens in the teens, under whichever date is in force by then.

Wikidata knows 7,888 US-born baseball players born 1950–1993; the Lahman database, which lists every major leaguer, has 6,478. Their month-by-month patterns correlate at r = 0.97, and both show the same August advantage (1.34× in Wikidata, 1.34× in Lahman). For players born 1994–2005 the August advantage is gone and May–July is slightly ahead (1.16× and 1.17×, with intervals that reach down to about 1, on 1,782 and 1,228 players). Little League’s 2006 move would explain it only for the youngest of them, so we read US baseball as a blurrier echo of the German switch, not as a second proof.

Line them up by cutoff, and the zodiac disappears

Here are 58 groups of players with a documented cutoff date (sport, country and birth years; at least 500 players each). By calendar month they disagree: some peak in January, some in April, some in September. Count the months from each group’s own cutoff instead, and they fall on one line.

The fake-cutoff test asks how special the real dates are: each group is given a wrong cutoff month at random, and the oldest-over-youngest quarter ratio of all groups together is computed again, 10,000 times. Across all 66 groups with at least 200 players the real cutoffs give 1.90× and the random ones at most 1.13× — but 50 of those groups use 1 January, and an excess of January–March births among athletes for any other reason would beat random dates too. The test that matters is the 16 groups with another cutoff (England, Japan, US baseball, cricket and rugby on 1 September, Germany and Belgium before the switch): 1.75× for the real dates, at most 1.33× for random ones. 7 of those groups were in our first look, and their cutoffs were chosen with the data seen; the 9 groups whose cutoff we predicted blind still give 1.45× against at most 1.32× for random dates, with no draw out of 10,000 reaching the real ones. There, January–March is not the advantage — English footballers born in it are under-represented — and the peak sits wherever the rulebook put the cutoff. A zodiac sign is a fixed calendar window; nothing about it knows which country’s rulebook a child grew up under.

We wrote our predictions down first

A first look at the data covered men’s football in ten countries (Germany, the Netherlands, the UK, Japan, France, Brazil, Belgium, Italy, Spain and Russia), baseball in the USA and Japan and hockey in Canada and Russia. Everything else stayed closed until the predictions below were written into the script and committed, with their cutoff dates and sources. Each country is its own prediction; the table shows every one, the misses too. Judge the pre-registration for what it is: the commit sits in our own repository, not on a public timestamping service; the raw files had been downloaded (not analysed) before it; and the baseline, the thresholds and the scoring code were written after it. The predictions themselves are unchanged — the published script carries them as committed.

44 of 51 directional predictions held (more players born in the three months after the cutoff than in the three before, one-sided z > 1.645), and 40 of 47 named the top sign. Of the five other predictions, 4 held. With every day of the year as the baseline instead of official births, 44 of 51 directional predictions hold; with famous non-athletes as the baseline, 44. One check flips: P14 (famous non-athletes stay flat) fails against a flat calendar, which ignores the real seasons of births, and holds against the non-athlete baseline.

P1. Men’s football, calendar-year youth leagues (born 1987–2007): more players born January–March than October–December, and the top sign Capricorn, Aquarius or Pisces

P2. Cricket, 1 September age groups: more players born September–November than June–August

P3. Rugby union in the home nations, school year from 1 September

P4. Rugby union, calendar-year age groups (born 1975–2007)

P5. Ice hockey, calendar-year age groups

P6. English rugby league, school year from 1 September

P8. Baseball, Dominican Republic and Venezuela: MLB’s international signing age runs from 1 September

P9. US soccer before the 2016 reform: 1 August age groups (born 1975–2004)

P10. South Korea, school grade from 1 March (born before 2002; late school entry blurs it)

The other five

• P7, Australia (miss). Australian football is documented as moving to 1 August in 1988 and back to 1 January in 1995, so players born 1981–1987 should lean to August. They do not: their January–March over October–December ratio is 3.01× against 1.68× for those born 1990–2007 (199 and 642 players). Either the documented switch did not reach the clubs that produce professionals, or Wikidata’s Australians are too few; we count it as a miss.
• P11, women’s football is weaker (held). 1.31× for 22,107 women against 2.17× for men in the same countries and years.
• P12, rowing is weaker (held). Rowers start late, as teenagers: 1.17× against 2.17× for footballers.
• P13, darts, snooker, jockeys and golf are weaker (held). No age-group pyramid to climb: 1.04× for 5,690 players against 1.40× for footballers from the same countries. It held, but says less than it seems: these players come mostly from the USA, Japan and England, which do not use 1 January, while the footballers of the same list are mostly from calendar-year countries.
• P14, no cutoff outside sport (held). Every one of the six groups of famous non-athletes is within 4% of expected for both a 1 January and a 1 September cutoff with the baseline used everywhere else; with official births only, within 4% (next section). The main baseline borrows from these same people where official births are missing, so the second figure is the honest one.

Directional misses: rugby union in Wales (1.12×, 572 players), rugby union in Scotland (0.81×, 308 players), rugby union in Ireland (1.19×, 462 players), rugby union in Argentina (1.27×, 313 players), rugby union in Italy (1.24×, 335 players), ice hockey in Austria (1.09×, 287 players) and baseball in Venezuela (1.05×, 542 players). Direction held but another sign won: cricket in India (Sagittarius), cricket in Australia (Sagittarius), rugby union in Australia (Gemini), ice hockey in Latvia (Aries) and rugby league in England (Sagittarius). The Scottish prediction used the wrong date: Scotland’s school year starts on 1 March, not 1 September, and Scottish rugby players do lean to March–May (top sign Aries) — it stays a miss. Indian cricket peaks in Sagittarius, two to three months after the official 1 September; that fits a later effective date or the misreported ages Indian cricket has fought for years, and its expected counts rest on famous Indians outside sport rather than on official births (see Limits). We did not predict it.

Who should show nothing

If the skew came from the sky, or from the way Wikidata records birthdays, it would show up in people who never played youth sport. The table compares famous non-athletes with official births only (people from countries and years without them are left out), so the check does not lean on the substitute baseline built from these same people.

Oldest over youngest quarter for each cutoff. The largest departures are a few per cent (politicians 1.04× for 1 January, z = 4.3; the smaller groups swing more on August), against 2.17× for men’s football in calendar-year countries. Some of it may be real: studies have found small relative-age effects among managers and politicians too.

One more pattern, described rather than tested. Among men born 1990–2007 in calendar-year football countries, the January–March over October–December ratio is 2.28× at its highest for players with a handful of Wikipedia articles and 1.54× for the 516 most famous (40 or more language editions). It looks like the “underdog effect” of the research literature — relatively young players who survive the filter turning out better — but the groups differ in country mix (the top one is mostly Spanish, French and German players, the bottom one has many from Turkey and Russia), so it is not evidence of it.

Look up your sport, country and sign

The “sign” of every sport

For completeness, each sport with every country pooled. Read it with care: pooling mixes cutoffs (England’s September with Japan’s April and everyone else’s January), so a pooled table understates the effect inside each country and says nothing a horoscope could use.

Men, except women’s football; sports with at least 800 players with a country. Ice hockey, basketball and tennis lead because almost every country runs them by calendar year; baseball and American football are dominated by the USA, where school grades and Little League’s 31 July split the year differently.

Method

• People. Wikidata (via the QLever endpoint, 3,013,074 occupation rows) for 25 sports and six groups of famous non-athletes. A person counts in every group of their occupations; sports use men only, women’s football is its own group. Only birth dates given to the day are used — a date known only to the year is stored as 1 January and would land exactly where a cutoff effect shows. 611,259 people without a day-precise date and 38,387 with two different ones were dropped; 1,245,868 remain with births in 1950–2007.
• The 1st of every month is left out. Even among day-precise dates, 1 January holds 2.0 times an ordinary January day’s births, and other 1sts up to 1.16 times. Observed and expected counts both skip every 1st.
• Country. Wikidata’s “country for sport” if it names one country, otherwise the country of the place of birth (England, Scotland, Wales and Northern Ireland separately), otherwise a single citizenship: 249,064, 601,831 and 310,848 people; 84,125 had none and are left out of the country tables.
• Expected births. For every person, the monthly births of their country in their birth year, spread evenly over the days of the month (Eurostat from 1960, Human Fertility Database from 2000, daily US and French births). 55% of athletes are matched to official births of their own year or a year within ten; 13% to the birthdays of famous non-athletes from the same country and decade; 32% (Latin America, Africa, Asia outside Japan and Korea, the USSR before 1990) to the official birth season of their hemisphere. Seasons of births move a sign’s expected count by under 10% in most groups and by up to about 16% in a few single-country cells; the cutoff effect is several times larger, and the predictions come out the same with any of the baselines (44 and 44 of 51 against 44).
• Signs. Conventional newspaper dates (Aries 21 March – 19 April and so on). The Sun changes sign at a different hour each year, so a person born on a cusp day can belong to either neighbour; with birthdays only, the table date is used.
• Quarter ratio. Players born in the three months from the cutoff over those born in the three months before it, each divided by its expected count; the 95% interval and z come from Poisson counts.
• Fake cutoffs. Each documented group got a wrong cutoff month at random (uniformly among the eleven others), 10,000 times; the statistic is the pooled quarter ratio.

Sources

• Wikidata (CC0), queried through QLever↗ on 28 October 2026.
• Lahman Baseball Database (SABR, CC BY-SA 3.0), the check of Wikidata’s coverage; only the aggregates quoted on this page are used, not redistributed.
• Official births by month: Eurostat demo_fmonth; Human Fertility Database, Short-Term Fertility Fluctuations (CC BY 4.0); NCHS and SSA daily US births; INSEE daily French births.
• Cutoff dates, as used in the predictions:
  • Helsen et al. 2005, J Sports Sci 23:629 (ten European countries); Musch & Hay 1999
  • FA / ECB / RFU age groups follow the school year from 1 September; Kelly et al. 2021
  • BCCI Age Verification Programme rules 2025-26 (cutoff 1 September)
  • Cricket Australia age groups, PeerJ 2019 7:e6867
  • Barnsley et al. 1985 (Canada); IIHF age categories by calendar year
  • U20 World Championship RAE study (UiTM, 2019): calendar-year cohorts
  • van den Honert 2012, J Sports Sci; Musch & Hay 1999: 1 Jan -> 1 Aug (1988) -> 1 Jan (1995)
  • MLB Rule 3(a)(1)(B): an international amateur must be 17 by 1 September of the first season
  • Korean school year from 1 March until the 2002 birth cohort (Hangyo 2006)
  • Smith et al. 2021, Front Sports Act Living 3:635195: 1 Aug until the 2016 reform
  • Germany and Belgium: 1 August until 1997 (Helsen et al. 2000; Ostapczuk & Musch 2013)
  • Japanese school year from 2 April (J-League RAE studies, IJSSC 2022)
  • Little League: 31 July until 2006, then 30 April (littleleague.org; Slate 2008)

Limits

• The 730,728 athletes are everyone with an exact birth date in the 25 sports, women included; 62,229 of them have no country and enter only the pooled sport tables. The sign wheels by country are men, except women’s football.
• Wikidata is not a census of athletes: it is richer for Europe and the Americas, for men and for top leagues. The Lahman check covers one sport in one country; elsewhere, coverage could bend a single cell, though not the agreement across dozens of them.
• A birthplace is not always where a player grew up; migrants trained under another country’s cutoff are counted where they were born unless Wikidata names their sporting country.
• The cutoff dates come from federations and published studies; where a country ran several dates side by side (clubs against schools), one is used.
• Outside Europe, the USA and recent years, the expected counts lean on substitutes: Japan’s football wheel rests about half on famous Japanese outside sport (52% of its expected count), India’s cricket almost wholly, and Latin America, Turkey and Ukraine on the official season of their hemisphere. Seasons of births move a sign’s expected count by under 10% in most groups, up to about 16% in a few; effects of 1.5× and more do not hinge on them, small ones can. With every day of the year alike as the baseline, 44 of 51 directional predictions still hold.
• Intervals and z-scores treat counts as Poisson. Real counts vary more — the baseline has its own error — so a borderline z (around 2) is borderline, not proof.
• Wikidata stores some birthdays on suspicious days: besides 1 January, days like 2 February or 10 October (day and month alike) are more common than their neighbours. Every 1st of the month is left out; the rest shift a sign by about 1%.
• This is about Sun signs as date ranges. Nothing here tests the Moon, the rising sign or the rest of a birth chart — birth times are almost never known for athletes.