Side Projects
Six Years of My Great British Bake Off Fantasy League, in Data
October 9, 2026
In 2021 I started a fantasy league for The Great British Bake Off. The rules are simple: after the first episode airs (and before the second), you draft three bakers, one of them your captain, and you rack up points every week for Star Baker, Paul Hollywood handshakes, technical placings and simply surviving. A captain who wins the whole thing is the jackpot.
Six seasons later, the league has grown from 65 mostly family and friends to 170 people. I've kept every spreadsheet and Google Form along the way, so this year I finally did what any data scientist would do with a hobby: I over-analyzed it.
289
different people have played since 2021
29
have played all six seasons
81%
of 2025 players came back for 2026
0
times the league's favourite captain won
A league that keeps growing
The growth comes from people coming back and bringing friends. 2026 is the first season where returning players clearly outnumber newcomers: 135 returning, 35 new.
Each season's league, split by the year each player first joined. The dark base is the 2021 originals who keep coming back.
Some players changed their names over the years, so a few returning players may be counted as new here.
The crowd is good at spotting finalists, terrible at spotting winners
Here's the pattern I didn't expect. Collectively, the league is a decent forecaster: bakers who end up on more teams tend to last longer in the tent. But the single most popular captain has never won. Not once in five finished seasons.
A big reason is timing. Picks lock after episode one, so the league leans hard on first impressions, and the week-1 Star Baker is almost always a top captain pick. Week-1 Star Bakers have finished 4th, 4th, runner-up, 10th and 7th.
Each dot is a baker, placed by the share of the league that made them captain. Gold marks the eventual winner.
Picks lock after episode 1 airs and before episode 2, so the league chooses with very little to go on. 2026 is still airing.
My favourite stat in this whole project: in 2023, exactly one entry out of 100 made Matty their captain. He won. Congratulations, Phil & Steph G., you are the league's only true oracle. In 2024, three people captained the eventual winner, Georgie.
Sometimes the whole league piles onto one baker. In 2023, 45% of players made Abbi their captain; she went home in week 3. This year is the second-biggest pile-on: 41% made week-1 Star Baker Mo their captain. 2026 does set one record, though. 39 people submitted the exact same Mo / Tom / Gary team, the largest identical bloc ever. (The last time a big group shared one team, in 2021, eight of them tied for first.) History suggests it's a bold strategy.
Share of the league that picked the single most popular captain, by season.
2023 had the biggest pile-on (45% for Abbi, out in week 3). 2026 has the largest group with an identical team: 39 people picked Mo, Tom and Gary.
Does experience help? No.
I fully expected long-time players to have an edge. They don't. First-time players finish, on average, at the 49th percentile; people on their fourth season or later finish at the 52nd. It's a coin flip with frosting.
Average finishing percentile in 2022–2025, by how many seasons a player had already played. 50% is a coin flip.
And if you're wondering whether running the league gives me some kind of advantage: please enjoy my results.
Where I finished in my own league each year (right = better).
In 2025 I added a benched baker, a fourth pick who subs in once one of your starters goes home. Mostly it created regret: 80% of benched bakers outlasted at least one of their team's starters, and 22 people benched the eventual winner.
2025 introduced a benched baker. How many teams benched each baker, with the winner in gold.
80% of benched bakers outlasted at least one of that team's starters; 19% outlasted all three.
What about the bakers?
Each series has 12 bakers, split six women and six men, so five finished seasons gives us 60 bakers: fun to look at, far too few to draw conclusions from. A few things still jump out. No baker over 50 has made a final in the five finished series. Londoners reach the final at more than twice the rate of everyone else. Men have taken more of the finalist spots, but women have won three of the five titles. Two of the five winners worked in healthcare.
Share of 2021–2025 bakers who reached the final (top 3), by trait. 🏆 = series winners. Only 60 bakers, so treat this as trivia, not science.
Age
Job
Hometown
Gender
This year's league has made its own bet. In every previous season, picks were split roughly evenly between men and women. In 2026 the three most-picked bakers are all men, and the average woman in the tent is on fewer than 1 in 10 teams. Given everything above, I'd keep an eye on the women.
Share of 2026 teams that include each baker in any role. The three most-picked bakers are all men; across the five previous seasons, picks were split evenly between men and women.
How 170 people found a Bake Off spreadsheet
This year I added a “who recruited you?” question to the sign-up form. The answers ranged from full names to “my hot wife,” “I'm an OG, duh,” and “the ‘surprise me’ button on Google.” (That last one turned out to be me.) After some cleanup, 152 of this year's 170 players trace back to me, but only 46 directly. The rest came through friends of friends of friends, up to seven links away.
Every 2026 player linked to the person they said recruited them. Bigger circles recruited more people. Search for anyone, or click a circle.
Top recruiters (besides me)
- 8
- 6
- 6
- 5
- 5
Names are first name + last initial. Lots of sign-ups left off last names or used nicknames, which made matching people tricky, so a few links may be missing. 5 players' recruiters couldn't be identified and float on their own.
How I built this
Everything here came from six years of Google Form exports and scoring spreadsheets, plus each series' baker details (age, hometown, occupation) from Wikipedia and the show's “Meet the Class” pages. The hardest part wasn't the charts; it was working out who's who. Families share email addresses, people change last names, and someone signs up as “Hiphopapotamus.” A few Python scripts match players across years, and I re-run them every season. I'll update this post when 2026 wraps up and we find out whether Mo breaks the curse.