Fair Machine
Train a machine to pick a cricket team, catch it being unfair, and fix it. Everything runs in your browser, zero coins.
A school wants a machine to pick the cricket team
The coaches are busy. They say: "Machine, look at last year's list. Learn who we picked. Then pick this year's team the same way."
Here is last year's list. Notice anything?
📋 See all 40 players from last year
Now let it pick this year's team
12 new players tried out. The machine must pick 6. Tap the button and watch.
The machine learned that "town" matters, because last year's coaches were unfair. It copied them.
Fix it
Try one fix, or both. Then retrain and see who gets picked.
Fairness meter
Grey = how much of the talent comes from each town. Purple = how much of the team the machine picked from there. Fair means they are close.
What this means in real life
Real machines pick people for jobs, decide who gets a loan, and even unlock phones by looking at faces.
They learn from old lists made by people, and people are sometimes unfair without noticing.
So a machine can say "no" to a great person just because of their town, their name, or the colour of their face.
That is why grown-ups must check the list the machine learned from, and test it on everyone, before they trust it.
The machine is a tiny "logistic regression". It has one number (a weight) for each thing it can see. Big positive weight = "this helps you get picked". Big negative = "this gets you rejected".
Train the machine to see its weights.
How it decides: score = batting × w1 + bowling × w2 + town weight + a starting number. Then it turns the score into a "chance of being picked" between 0% and 100%. The 6 highest chances make the team.