
This one is almost all matchday: we analysed three real games with tracking data and measured the broadcast against them, instead of tuning it by feel. On top of that, a rebuild of the youth setup, where the AI had taken an advantage for itself.
Who gets the pass is now decided by a model built from 1,763 real passing decisions: distance, space, open passing lane. It matches 43.3% of the real choices, blind guessing gets 10.0%. The old formula never looked at a single opponent.
A foul in the box is now a penalty. Distance to goal used to decide on its own: 30 out of 35 offences inside the area gave a free kick.
Headers get lost now, the defender clears 46.5% of aerial duels. And a third of corners are played short instead of swung in; we were at 4.1%.
Attacks now break down against the defence rather than a dice roll. Plus: no more rubber band on the ball, no more creeping around for seconds, no more keeper chasing the ball to the touchline.
An AI club used to get nine prospects from the whole pool every four weeks, while you without ScoutNet only saw your own youth. Now the same rule applies to them.
No more training penalty. Level 1 without a youth coach sat at minus 16%, that is the zero point now. Upgrading costs 14 million instead of 35, but comes with upkeep of 4,000 to 44,000 per week.
An AI club promotes at most two graduates per season, capped at squad strength plus five: they used to come out 13.7 points above their club's strength, now it is 2.3.
They pay 6 to 18% above the going rate - the easier the difficulty, the more.
In return they can fight back: a club living beyond its means sells a player, a club short on cash takes out a loan.
The starting balance is yours alone now, AI clubs begin with a third of it.
Below ten frames per second the game speaks up and offers to turn the detail down, the assistant saves you two clicks, and 21 pieces of text were still German in all five languages.
Tell me what you notice. The penalty and the creeping ball carrier both came from your reports.