Add game: EinStein würfelt nicht!

[giogix2]

Here's my implementation of the game EinStein würfelt nicht!

I've implemented only the basic version of the game with a 5x5 board, as described in the wiki https://en.wikipedia.org/wiki/EinStein_w%C3%BCrfelt_nicht!. I'm not familiar with the other versions, so for now I decided to stick with the simplest version.

There is a common variant where a player wins the game after winning a total of 3 games. In this implementation, winning a single match means winning the game.

Feedbacks are appreciated. Hopefully this game is a fit for open_spiel.

[lanctot]

Nice!! Thanks!

[lanctot]

Error looks unrelated to your changes and broken in some of our code. We'll look to get this fixed and let you know when master is up to date with fixes!

[giogix2]

Thank you for the heads up. I was planning to dive deeper into this in the next few days, but I was also confused, because tests pass for macos-14, but not for the others.
I'll wait for the changes in the master branch.

[lanctot]

Thank you for the heads up. I was planning to dive deeper into this in the next few days, but I was also confused, because tests pass for macos-14, but not for the others. I'll wait for the changes in the master branch.

Ok fixes applied. Can you pull the most recent changes from master and push the merge commit?

[giogix2]

Thank you for the heads up. I was planning to dive deeper into this in the next few days, but I was also confused, because tests pass for macos-14, but not for the others. I'll wait for the changes in the master branch.

Ok fixes applied. Can you pull the most recent changes from master and push the merge commit?

Done.

[giogix2]

Some jobs are still failing. Could this still be problem unrelated to these changes? It's a bit difficult to understand what is the root cause from the logs.

Though, there is one test failing for which the reason is clear, specifically: build_and_test / build (ubuntu-22.04, 3.10, OFF, OFF, OFF). This test fails because the initial board setup differs from the one in the playthrough file. In this game implementation, the position of the cubes on the board is initialized randomly, controlled by a random seed passed as an argument. If the seed is set to -1, it’s taken from the clock; otherwise, the default value is 42. This is done before the first move, at the constructor level.
When running a playthrough with the default parameters, I’d expect the initial setup to be the same (the playthrough tests always pass locally, note: I am using macOS and Python 3.10.15). However, the playthrough tests fail on Ubuntu 22.04 with Python 3.10. Could the random generator produce different values on different operating systems?
And maybe all the tests are failing because of this even if not specifically mentioned in the logs...

If this is the case, I might need to modify the implementation so that the board starts empty, and the first random action is to set up the board. This concept seems similar to what you're doing in Backgammon here.