Tip: Versatility is important.
Tip: Stay on track.
Tip: Try using as high a TPS as possible.
Tip: Choose the neural network inputs thoughtfully.
Tip: Try tweaking the natural selection controls in-between each
generation.
Tip: Try changing the stadium track in-between each generation.
Tip: Use the garage to change the car's probe angles.
Tip: Use the racing stadium to draw your own tracks.
Tip: Export good cars to save them for later.
Tip: Think about what each control does and what implications it has.
Tip: Having more/larger layers, probes, and inputs do not guarantee better
performance.
Tip: Use the leaderboard to compare car performance.
Tip: Use the car peeker to inspect car internals.
Tip: Use the import/export functionality to share cars with others.
Tip: Your settings are saved across refreshes.
Tip: Enable race mode to get a larger stadium while disabling natural
selection.
Tip: Use the reset settings button to reset all settings to their
defaults.
Tip: Use the reset leaderboard button to reset the leaderboard.
Tip: Use the clear stadium button to clear the racing stadium.
Tip: You can drag and drop car JSON files to import them.
Racing
Stadium
Stadium
LMB: Draw Track
RMB: Erase Track
RMB: Erase Track
Garage
Front (0°)
Back
Neural
Network
Network
Input Layer
Input
1
Hidden
Output
1
Natural Selection
Gen
Population
Survival
Best Score
Leaderboard
Rank
ID
Gen
Score
Lap
Speed
On Track
?
Road Score:
Every tick,
onTrack ⇒ +0.2, offTrack ⇒ -5
speed[-1,0,1,5] ⇒ [-3,-2,+1,+4]
Performance Score: At the end,
lap (clockwise) ⇒ +100×(onTrack%)²
averageSpeed[0,1] ⇒ [-100,+100]
Total Score: Road Score + Performance Score
onTrack ⇒ +0.2, offTrack ⇒ -5
speed[-1,0,1,5] ⇒ [-3,-2,+1,+4]
Performance Score: At the end,
lap (clockwise) ⇒ +100×(onTrack%)²
averageSpeed[0,1] ⇒ [-100,+100]
Total Score: Road Score + Performance Score
Max TPS (est.):
0
Current TPS:
0
Target TPS:
Drop car JSON files here to import them