https://www.codingame.com/training/medium/can-you-save-the-forest---episode-1
Send your feedback or ask for help here!
Created by @chabes,validated by @Alex-1,@slempel and @Nagato_Uzumaki.
If you have any issues, feel free to ping them.
Excellent! I made a mistake in my decision, it cost me gray hair)
Thanks to the author for an interesting mini-game!
1 Like
Nice puzzle, congrats!
I used beam search for the first time ever.
There are way too many puzzles with “Episode 1” or “Part 1” in the title with missing follow-up.
Please avoid this fate, this puzzle deserves a second episode!
2 Likes
And the episode 3 is well advanced although I paused development since the Spring Challenge and did not restart it since.
My first beam search too 
Hi all. I’m passing all the tests except for “several outbreaks”. My code is not finding a single “winner branch” of the possibility-Tree. Very shortly: i’m using a “weighted” Monte Carlo Tree Search , so i have the standard 4-component : select → find an expandable leaf, expand → actually expand the leaf found, simulate → from the node just created play randomly a game until the end , backtrack → keep track of the score, but i think this is not a beam search: i do not have a cut-off for the number of children that a given node can have. Last precisation is: i’m not expanding and simulating randomly, becouse i give a weight to the action to choose such that it prioritize, increasing the probability to play that move in that state, the level-3 fire squares (i’m weighting in this way 1->1 , 2 ->2 , 3 with 0 neighbours where to spread ->3 , 3 with 1 neighbours available to spread → 3+1 , …, 3 with 4 neighbours available → 3 +4 = 7 , and then normalized to the sum ofc).
That said my question is: if i should prefer to put a cut-off on the number of child that a node can have, how to choose such number of max_child ? How to calculate it ? And should it be variable from node to another ? Or maybe the problem is in the way i am prioritizing the search ?
P.S. theoretically, if i play A LOT of games (eventually exploring all the state-space) i should be able to find at least 1 winning-tree-path, but i cannot becouse of timeout (at best i can search for 1000 child, for every 1 of these i have played 1 “random” game till the end and lost everytime). I have no idea of the magnitude-order of the number of possible state for the game, so i cannot say if it is a waaaaay to big than 1000.
Thank you very much for the help! really nice puzzle
I litteraly tried this puzzle for 2 weeks, and 2h after i wrote this i found a solution. Now i strongly play non-randomly.