I am stuck in wood 1 league, I’ve beat the boss, I even made more efficient code for what was given but it still won’t pass to the next level.
Promotion to the next league not depend only of beating the boss, but to be ranked better than him. That imply to also beat the top players.
Look at the games you’ve lost, and try to find out what went wrong. Then improve.
Good luck
Is it possible to switch between leagues? I was enjoying watching races in the Silver League and then double-pods broke it all.
Can I somehow experiement in the Silver League a little bit more and then switch back to the Golden one?
No it is not possible unless you create a new account.
Hi, I’m trying to simulate the game with python but i’m unable to reproduce collisions. My simulation of the trajectories without collision work perfectly, I’m able to detect collision, i can find the exact point and the exact date of the collision, i have tested brute force, point to line and few method to apply bounce (like the Magus one) and i always find the same wrong result wich drift slightly compared to the game logs. Don’t you have any tips that can help me ?
Why magus apply collision force twice ? (if i apply it only once, i find the same result than with my own calculation…)
Friction should be applied to the collision force ?
PS: If i use a minimal collision impulsion of 130, the first collision is good but not the second 
…
Or should i reverse the game level seed with checkpoints positions to apply rand() somewhere lol ^^
I need help i do not know how to do any thing pls send tutorial vid
‘Why magus apply collision force twice ? (if i apply it only once, i find the same result than with my own calculation…)’ - what if you collide with more than one pod or checkpoints
‘Friction should be applied to the collision force ?’ - apply friction only after playing one turn on velocities
The sim does it twice, so Magus does too. It’s a programming trick to make the collisions look right (though not technically correct for real-life calculations) while avoiding a lot of extra math and weird side-effects you get when floating point numbers reach their accuracy limits.
1 Like
Hello,
I am trying to predict the position of a pod next turn based on inputs, target and thrust. My prediction can be up to 150 different from the one given (norm 2 of the segment [predictedDestination, realDestination]). Is it possible to make it better? or even perfect?
Thank you
That’s nice. I can get faster to the point I want to code.
1 Like
I feel good. That game is awsom!
2 Likes
I am confused in wood 1 to bronze it says replace the thrust value by boost key word when i do that nothing happen it shows you are better than boss when i test it arena i don’t go to next league after i resubmit it boss win and when i again submit it. It said i am better than boss.
Can anyone Help me.
It would be neat if you still had access to your 1-pod bot after graduating to gold league and still watch 1-pod battles. I would like to see how my bot would have fared against other bots that also made silver league champion.
Maybe you could add a legacy racing circuit where champion silver submissions compete with each other?
1 Like
After recent challenge I was determined to find out why this is happening, as I increase the turns count for the simulation my pods start to play worst. But unfortunately I couldn’t, so I’m asking for help.
I’ve stripped the task to only one pod, just looping checkpoints alone.
My evaluation function, for the pod, is very simple:
(50000 * passedCheckpointsCount) - distanceToNextCheckPoint
I want to optimize the commands for may pod to pass as much CPs as possible and be nearest to the next CP.
Here is a simulation(2 CPs, 1 pod) for 8 turns in the future:
Seems right, the path is optimized to be nearer to the next CP
But here is a simulation(2 CPs, 1 pod) for 32 turns in the future:
It seems like something is broken, when a CP is crossed during a simulation, and the pod gets stuck on the border of the CP instead of turning towards the next one. And I cannot understand why…
Have this happened to anyone or do you see a problem in my logic?
Amazing Multiplayer Game 
Easy to learn, hard to master…
1 Like
In your test case, how do you select a solution that takes the minimum steps to go from start to finish ?
I’m only checking the evaluation of each chromosome:
(50000 * passedCheckpointsCount) - distanceToNextCheckPoint
Based on that I choose parents (using roulette wheel) for creating the new generation. At the end of the simulation I get the first command of the chromosome with the best evaluation.
Looking at the length of the lines, my guess would be that in 32 turns it can reach the checkpoint but then can’t turn round quickly enough to reduce the back distance to CP0. In this case the optimum strategy for your eval function would be to use all 32 turns to come into CP1 as slowly as possible and end up just inside CP1 (to minimise the distance back to CP0). Often this can be fixed by making the evaluation take into account the turn the CP was passed so getting there earlier is better.
2 Likes