Coding Games and Programming Challenges to Code Better
Send your feedback or ask for help here!
Created by @PhOeNyX,validated by @qimingzhenfeijin,@TheCheezLord and @20Ll37.
If you have any issues, feel free to ping them.
Hello @Wazemo and @alexcercos!
Following a discussion I had with Alex, I updated the contribution with the following changes:
-
The statement has been updated to better describe the glyph spawning process.
- This should allow players to more easily anticipate where glyphs can spawn, while still preserving some uncertainty if not all flames have spawned yet.
-
A new test case using a random seed has been added.
-
40 random validators have been added to provide more accurate scores for programs that generalize better to the problem.
-
Previously, glyph spawning would be canceled if the character was standing on the targeted cell, but not if the cell was occupied by a flame. I’ve now made glyph spawning independent of cell occupancy, so glyphs will appear even if the selected cell already contains the character or a flame. Flame spawning itself is unchanged and still requires an empty cell.
The leaderboard has not been automatically updated with the new validators. Don’t forget to resubmit your code to receive the points awarded by the new validators.
For reference, the theoretical maximum score has increased significantly:
- Before: 10 validators Ă— 40 points for a perfect game = 400 maximum theoretical points.
- After: 50 validators (including 40 new random ones) Ă— 40 points for a perfect game = 2000 maximum theoretical points.
In practice, this maximum is likely unattainable, as some points will almost always be lost due to unavoidable movements that don’t extinguish flames.
2 Likes
Hi @PhOeNyX!
Thank you for implementing these changes, they’ll make the game more interesting.
It’s good that it happened early on – you could say @alexcercos and I were in the testing phase 
It didn’t take many submits to figure out the glyph layout.
3 Likes
I’m glad too it was implemented (didn’t want to disrupt too much in the forum with private debates/suggestions 
)! It would have been a pretty short lived game, with just some improvements in the heuristic, I was able to find offline the optimal already with basic beam search and it would have been pretty easy to hardcode or “overfit” the 10 validators to reconstruct it deterministically, now it’ll be more of a challenge for us and for next players too, and not a too late change.
2 Likes