The Cube Enigma

justhalf Unnamed CodinGamer

about 2 years ago

Hey CodinGamer! Your contribution has unfortunately not been moderated on time and will temporarily be removed from the system... But no worries! You can republish it anytime to give it another chance! :)

about 2 years ago

As darkhorse64 mentioned 2 months ago, this seems to be about finding a reachable configuration that gives the maximum score, not about the moves themselves. In that sense, this is not a beam search problem. You can just use heuristics to find a reachable configuration with maximum points (i.e., preserving the parity), then use a solver algorithm to bring the starting position to the found configuration. Also what was the procedure used to guarantee that the starting positions given in the test cases are not reachable from the solved cube?

Coco3244 Unnamed CodinGamer

about 2 years ago

seems hard but interesting

Leszczon Unnamed CodinGamer

about 2 years ago

Pog

Pablito1424 Unnamed CodinGamer

about 2 years ago

about 2 years ago

Thanks for the contribution ! It's very interesting !

Hiro-GG Unnamed CodinGamer

about 2 years ago

Jaden38 Unnamed CodinGamer

about 2 years ago

thanks nice contribution

Sentius Unnamed CodinGamer

about 2 years ago

I like the idea, but as previously stated, running the test will yeld 100% with the default code, i don't know if it is intended.

GuillaumeLAURIE Unnamed CodinGamer

about 2 years ago

This is not a solo game i mean puzzle in which the objective is to get 100%, this is an optim in where you have to optimize your code and goal is to score more points!

about 2 years ago

By letting the default code I get 100%. I think there is a problem in your code

eulerscheZahl Unnamed CodinGamer

about 2 years ago

Not a problem. This is an optimization task. You get 100% correctness but with a really low score compared to some other players. This is also the case for some official optimization puzzles such as the CodinGame sponsored contest or A*Craft.

about 2 years ago

The mechanics of the optim are greatly improved with these new additions but the statement is still a total mess. To prevent hardcoding, would not it be possible to generate random starting positions ?

about 2 years ago

it will lead to some lucky submissions.

about 2 years ago

I think your optim has a major flaw. Once you have computed the best scoring arrangement during the first second, all you have to do is reach that position and cycle on it so that the 300th move reach that very position. This is trivial if you are left with an even number of moves; for odd number of moves, a cycle with an odd period is needed (I am no Cube expert, I don't know if it is possible). So you will spend hundreds of moves doing nothing. How unattractive ... Scoring with the sum of scores for each move looks no that much better to me. Once the best arrangement is reached, there is no better move than cycling between that position and the least inferior one. Therefore, I am very tempted to reject it unless proven wrong.

about 2 years ago

But finding that best scoring arrangement is itself a big task, what changes are you looking for can you tell me in brief?

about 2 years ago

The statement still does not say how the score is computed: from the state at the end of the game, sum of score states at each turn ?

c0d1mus Unnamed CodinGamer

about 2 years ago

Added this in the statement - ```NOTE - THE SCORE FOR EACH TURN IS THE SCORE OF CUBE AT THAT TURN AND THE FINAL SCORE OF THE GAME IS THE CUBE'S SCORE AT LAST TURN```

about 2 years ago

Very interesting one! Great Job!

trictrac Unnamed CodinGamer

about 2 years ago

✅ Approved the contribution. Comment: Great job !

trictrac Unnamed CodinGamer

about 2 years ago

Hello jrke, Thanks for this "optimisation puzzle" proposal. The order of the "face" given by the input is "wygbor" which seems not consistent with the statement and the different picture saying "wryobg".

about 2 years ago

didn't get what are you saying, can you please explain it?

about 2 years ago

Changes in statement, Another example added, images changed. PLEASE LET ME KNOW FOR ERRORS :)

GeffSwayze Unnamed CodinGamer

over 2 years ago

over 2 years ago

Very interesting, thanx!

BartholomewIII Unnamed CodinGamer

over 2 years ago

The asymmetrical images were annoying me. Feel free to use these: https://imgur.com/... I neglected to remove the white background, but that should be relatively easy to remedy. If you're wondering, they were done in Penpa+ (https://swaroopg92.github.io/...)

leojean890 Unnamed CodinGamer

over 2 years ago

There are still too many typos and unclear things in the statement to approve this optim. I do not understand how the final score is computed. It looks like the last turn score but I do not get the point. Summing the score for each turn makes more sense to me. The formula for scoring on a face is so badly written that I had to read the referee to understand that you score the square of number of sides matching the center color. IMHO, you should score points on a face only if you have at least 2 sides from the color of the center side.

over 2 years ago

I think that he uses this formula with squares in order to boost scores when a face has many corresponding tiles. So to score it's better to have many tiles well placed fastly. An approach that solves the rubik using around 300 iterations without having many well placed tiles at intermediary steps can score lower than another approach that doesn't solve the rubik during the authorized steps but having more well placed tiles faster :) maybe he should clarify the statement a bit yes^^

Lliane Unnamed CodinGamer

over 2 years ago

over 2 years ago

That's a good one

ge_pet Unnamed CodinGamer

over 2 years ago

Very very interesting!! Well done!

MoonAir Unnamed CodinGamer

over 2 years ago

Top :)

FAngelo94 Unnamed CodinGamer

over 2 years ago

Hard to find best solution but it seems very nice interesting, I hope to play with it :)

Toss_A_Coin Unnamed CodinGamer

over 2 years ago

Very interesting puzzle!

eulerscheZahl Unnamed CodinGamer

over 2 years ago

I finally found some time and motivation to test the game (just the basics so far, my submit score is around 10500). Mechanics look fine to me, as I made my bot crash is the received and expected state differ. My implementation agrees with yours. The statement could take a bit of cleanup. I edited a few sentences on github, but this is by no means complete. The schematic representation of the cube looks like it's done in Paint, no 2 tiles have the same size. It might help the game to add a command to terminate the game early and also give a penalty for taking many turns. But I would have to work harder on my solver to have a strong opinion on this. The viewer doesn't show the last action of the game.

leojean890 Unnamed CodinGamer

over 2 years ago

I hope you received enough feedback from the community to improve your contribution. Once you have made the necessary updates, publish your contribution!

over 2 years ago

Interesting ! I can implement solving method I learnt 15 years ago :P Reached 4429 using some easy 10 lines local randomized heuristic ^^'

over 2 years ago

Interesting idea. * The score/turn are difficult to read. Am I getting too old ? *"NOTE - TEST CASE PROVIDED CAN'T BE SOLVED SO JUST SCORE POINTS!". W00t ? From the comments, I understand that the cubes (all of them ?) cannot be solved but it should be explained much more clearly in the statement

over 2 years ago

Changed color for score and turn tell me if its not understandable for the note i will add the info. Thanks. :)

TheProCode Unnamed CodinGamer

almost 3 years ago

I hope you received enough feedback from the community to improve your contribution. Once you have made the necessary updates, publish your contribution!

almost 3 years ago

bissaka Unnamed CodinGamer

about 3 years ago

I can't help it :)

Astrobytes Unnamed CodinGamer

about 3 years ago

Won't comment on the contrib for now (other than agreeing with euler's idea) but this has nothing to do with The Lost Child Episode 1! Call it something else :)

eulerscheZahl Unnamed CodinGamer

about 3 years ago

The way it is, I see a risk of hardcoding while using public resources such as online solvers. Maybe make an impossible-to-solve cube and turn it into a real optim? Like 10 tiles of 1 color or just an impossible to solve rotation (not every state is reachable) This way you can't just rely on public literature either and have to find your own way. A win-win :)

about 3 years ago

hey euler, i read your comment and also agreed that hardcoding can be done brutually on this game and over night thinking on this i got an idea neither random nor pre-defined states i don't know is that a good idea or not but still sharing it - as of now there are pre defined states and i am thinking to play one random move before sending first turn input data to user and in this case there will be different cases for same test every time it runs, For the point of spam submitting for randomness there will be maximum increase lenght in solution of 1 to -1 also i can also make that one random move to two random move so total cases will be nearly 40(total no. of test cases) * 18(moves of cube)^2(depth) = 12960 and even if you take 10k states and take 10 moves to solve an state on an average which very to then it will cross 100k which is not the code limit for CG also i will remove test cases json from git repo whats your opinion over this?