[Community Puzzle] 2048

As far as I’m aware the hardcoding message is given whenever you submit and don’t get a 100% result. Either your program isn’t providing a move (time limit, runtime error etc) or it’s providing a completely invalid move (printing something other than u/d/l/r).

1 Like

Ty guy, but all my validator are in the green, all my grid score are comprised between ~600.000 and 1m3, and i have no problem until i submit.

After submiting, then, i get between one and four red hardcoding message. Meaning i end between 97% and around 8x%, and can’t push my score into the leaderboard.

Something you said made me think that maybe by submiting the server check again and have some timeout i didn’t have before? That would be surprising… but i will try soon to run my code with half the time alocated to check this.

Thanks

I know the added score from grid 2-31. Why would i need this random seed to have an expectation of my final score, when i can even have a good expectation of this random seed?

More, you saying that what’s happening to me can’t be, when it is… how can that be helpfull?

We have a big misunderstanding here, or you don’t make any effort.

Forget about the “possible hardcoding” message. It just means that when submitting, you did not pass one of the validators.

Usually, validators are different from IDE tests. In this particular puzzle they are the same, but there is a random validator. Not passing a validator can come from wrong/invalid answer or simply from a timeout issue. Maybe measure your response time locally or in IDE to check if you are very close to the threshold (which is a few seconds).

In the leaderboard you are #222 with 100% and 1.4M points. It might be an earlier submit.
When you resubmit with a better code, better points that PASSES 100% you can go up. But if you fail a validator and get <100%, then your new submit will not replace the old one, even if the new submit has more points.

1 Like

Okay, everything you said make sens. I still have a problem:

My code give the same answer for the grid 2/31-31, every time. The random seed is the only one that never failed.

I’ll try another way to keep time, but for now i’m just tired of this puzzle. :sweat_smile:

This game has 30 validators with a fixed seed, that are also visible in the IDE as testcase 2-31.
Testcase 1 (the random seed) is only a testcase and not a validator on submit. I know that CodinGame shows that differently - but that’s caused by a bug on their end, the random testcase is not run on submit.

Are you sure, that all testcases are shown in green on submit? I would expect at least 1 to be red, maybe a different one every time if you are close to timing out.

1 Like

I have had the same problem (direction values order and spawnTile needed adjustments)

Running time seems a bit random here…
This is a screenshot of 4 submissions for the exact same code:

image

Tha absolute technique would be some Reinforcement Learning with Python library, and recreate his own forward pass. The score would improve very huge. But I don’t know how to do more than 168Ko on one test with my online solver. I run a beam search. What is the way to do 1.74M online on one test ?

I run a (unoptimized) Python beam search with k = depth = 10 and it runs correctly with 1.5M score on submit.