NOTE: This project is in the ideation phase and currently there's no code here.
Basically, the general idea is to make a game where players draw simple neural networks using the inputs below for simple amoeba to follow. Each amoeba has breedable traits. The amoeba are then trained to play a simplified version of American Football.
Game is played on a 256x128 grid Game is 60 minutes Players can occupy tenths of a grid unit
Stats: Speed - number of tenths a player can move per turn [default 1] Shake - when grabbed, the liklihood they can shake off a player of equal weight per turn [default 0.05] Catch - odds a player catches a perfect pass [default 0.95] Throw - odds a player throws a ball perfectly straight [default 0.975] Block - odds a player blocks a perfect pass they are in line of [default 0.1] Int - odds a player intercepts a perfect pass they are in line of [default 0.01] Weight - effects how slowed down a dragged player is, +200 [default 75] Arm - average pass distance, normal distribution past this number [default 30]
Player Input: Play call number [int] Relative positions of every player on the field [array of pairs of int] 0 is on top of player, 1-4 are literal field units, 5 is within 32 field units, 6 is within 128 field units, 7 is past that Position on the field, where x=0 is your own goal line, x=255 is the opposing endzone [pair of int] Can ball be thrown, becomes false if ball passes line of scrimmage or is passed forward [bool]
Player Output: Direction - direction to move in Speed - 0-8 how much the player should move as a fraction of their speed Pass - 0-10 which player to pass or hand the ball to (pass is > 1 field unit, forward pass advances the ball in the x direction)
Coach Input: Score delta [int] Game time in minutes [int]
Coach Output: Play call number [int]