Add Xiangqi (Chinese Chess) implementation to OpenSpiel by Arahan-kujur · Pull Request #1512 · google-deepmind/open_spiel

Arahan-kujur · 2026-03-30T16:59:16Z

This PR adds an implementation of Xiangqi (Chinese Chess) to OpenSpiel.

Features:

10x9 board with standard starting position
movement rules for all pieces
cannon capture mechanics
horse blocking rule
flying general rule
observation tensor support
undo functionality

Includes unit tests for move generation, rule enforcement, and state transitions.

lanctot · 2026-03-30T20:26:20Z

Thanks!

lanctot

Can you add the game to docs/games.md ?

open_spiel/games/xiangqi/xiangqi.h

lanctot

Can you:

Add the game to pyspiel_test.py
Add a playthrough

See steps 5 & 11 of the "Adding a Game" guidelines here: https://github.com/google-deepmind/open_spiel/blob/master/docs/developer_guide.md#adding-a-game

…integration tests Made-with: Cursor

Arahan-kujur · 2026-03-31T16:02:13Z

Can you:

Add the game to pyspiel_test.py

Add a playthrough

See steps 5 & 11 of the "Adding a Game" guidelines here: https://github.com/google-deepmind/open_spiel/blob/master/docs/developer_guide.md#adding-a-game

Added Xiangqi to docs/games.md, updated pyspiel_test.py, and added the playthrough file for integration testing. Let me know if anything else needs to be adjusted.

lanctot · 2026-03-31T16:14:56Z

Added Xiangqi to docs/games.md, updated pyspiel_test.py, and added the playthrough file for integration testing. Let me know if anything else needs to be adjusted.

Great, thanks.

Did you see this comment? #1512 (comment)

Arahan-kujur · 2026-03-31T16:59:03Z

Added Xiangqi to docs/games.md, updated pyspiel_test.py, and added the playthrough file for integration testing. Let me know if anything else needs to be adjusted.

Great, thanks.

Did you see this comment? #1512 (comment)

Yes, thanks — I’ve addressed that comment in the latest commit.

lanctot · 2026-04-01T09:55:48Z

Something seems wrong with the usual tests, but the wheels tests passed. So it's probably fine.

Can you make a one-line change (maybe add a comment somewhere) and push it so it can trigger a new invocation of the tests?

Arahan-kujur · 2026-04-01T10:53:14Z

Something seems wrong with the usual tests, but the wheels tests passed. So it's probably fine.

Can you make a one-line change (maybe add a comment somewhere) and push it so it can trigger a new invocation of the tests?

I’ve pushed a small change to trigger a new test run.

lanctot · 2026-04-01T12:11:43Z

Ok, I ran into the same issue again, and each time I could not see a log. So this time I'm watching it.

https://github.com/google-deepmind/open_spiel/actions/runs/23845020407/job/69518259486

If it happens again, I suspect one of the simulation tests might be stuck in an infinite loop. Especially if it's a game where uniform random policy could go on forever, then it's likely just getting stuck in an endless game.

(To fix these cases, we usually put a limit on the number of moves and call it a draw after that number of moves have been applied without a winner. You'd have to add that to the implementation.)

lanctot · 2026-04-01T12:15:56Z

Hypothesis confirmed. This is the last part of the output:

player 0
Rewards: 0 0
Returns: 0 0
Sum Rewards: 0 0
chose action: 6907 (G(8,4)-(7,4))
State: 
...g.....
.........
.........
.........
.........
.........
.........
....G....
.........
.........
player 1
Rewards: 0 0
Returns: 0 0
Sum Rewards: 0 0
chose action: 282 (g(0,3)-(1,3))
State: 
.........
...g.....
.........
.........
.........
.........
.........
....G....
.........
.........
player 0
Rewards: 0 0
Returns: 0 0
Sum Rewards: 0 0
chose action: 6106 (G(7,4)-(8,4))
State: 
.........
...g.....
.........
.........
.........
.........
.........
.........
....G....
.........
player 1
Rewards: 0 0
Returns: 0 0
Sum Rewards: 0 0
chose action: 1101 (g(1,3)-(2,3))
State: 
.........
.........
...g.....
.........
.........
.........
.........
.........
....G....
.........
player 0
Rewards: 0 0
Returns: 0 0
Sum Rewards: 0 0
chose action: 6925 (G(8,4)-(9,4))
State: 
.........
.........
...g.....
.........
.........
.........
.........
.........
.........
....G....
player 1
Rewards: 0 0
Returns: 0 0
Sum Rewards: 0 0
chose action: 1902 (g(2,3)-(1,3))
State: 
.........
...g.....
.........
.........
.........
.........
.........
.........
.........
....G....
player 0
Rewards: 0 0
Returns: 0 0
Sum Rewards: 0 0
chose action: 7726 (G(9,4)-(8,4))
State: 
.........
...g.....
.........
.........
.........
.........
.........
.........
....G....
.........
player 1
Rewards: 0 0
Returns: 0 0
Sum Rewards: 0 0
chose action: 1083 (g(1,3)-(0,3))
State: 
...g.....
.........
.........
.........
.........
.........
.........
.........
....G....
.........
player 0
Rewards: 0 0
Returns: 0 0
Sum Rewards: 0 0
chose action: 6925 (G(8,4)-(9,4))
State: 
...g.....
.........
.........
.........
.........
.........
.........
.........
.........
....G....
player 1
Rewards: 0 0
Returns: 0 0
Sum Rewards: 0 0
chose action: 282 (g(0,3)-(1,3))
State: 
.........
...g.....
.........
.........
.........
.........
.........
.........
.........
....G....
player 0
Rewards: 0 0
Returns: 0 0
Sum Rewards: 0 0
chose action: 7726 (G(9,4)-(8,4))
State: 
.........
...g.....
.........
.........
.........
.........
.........
.........
....G....
.........
player 1
Rewards: 0 0
Returns: 0 0
Sum Rewards: 0 0
chose action: 1101 (g(1,3)-(2,3))
State: 
.........
.........
...g.....
.........
.........
.........
.........
.........
....G....
.........
player 0
Rewards: 0 0
Returns: 0 0
Sum Rewards: 0 0
chose action: 6907 (G(8,4)-(7,4))
State: 
.........
.........
...g.....
.........
.........
.........
.........
....G....
.........
.........
player 1
Rewards: 0 0
Returns: 0 0
Sum Rewards: 0 0
chose action: 1902 (g(2,3)-(1,3))
State: 
.........
...g.....
.........
.........
.........
.........
.........
....G....
.........
.........
player 0
Rewards: 0 0
Returns: 0 0
Sum Rewards: 0 0
chose action: 6106 (G(7,4)-(8,4))
State: 
.........
...g.....
.........
.........
.........
.........
.........
.........
....G....
.........
player 1
Rewards: 0 0
Returns: 0 0
Sum Rewards: 0 0
chose action: 1101 (g(1,3)-(2,3))
State: 
.........
.........
...g.....
.........
.........
.........
.........
.........
....G....
.........
player 0
Rewards: 0 0
Returns: 0 0
Sum Rewards: 0 0
chose action: 6907 (G(8,4)-(7,4))
State: 
.........
.........
...g.....
.........
.........
.........
.........
....G....
.........
.........
player 1
Rewards: 0 0
Returns: 0 0
Sum Rewards: 0 0
chose action: 1902 (g(2,3)-(1,3))
State: 
.........
...g.....
.........
.........
.........
.........
.........
....G....
.........
.........
player 0
Rewards: 0 0
Returns: 0 0
Sum Rewards: 0 0
chose action: 6106 (G(7,4)-(8,4))
State: 
.........
...g.....
.........
.........
.........
.........
.........
.........
....G....
.........
player 1
Rewards: 0 0
Returns: 0 0
Sum Rewards: 0 0
chose action: 1083 (g(1,3)-(0,3))
State: 
...g.....
.........
.........
.........
.........
.........
.........
.........
....G....
.........
player 0
Rewards: 0 0
Returns: 0 0
Sum Rewards: 0 0
chose action: 6925 (G(8,4)-(9,4))
State: 
...g.....
.........
.........
.........
.........
.........
.........
.........
.........
....G....
player 1
Rewards: 0 0
Returns: 0 0
Sum Rewards: 0 0
chose action: 282 (g(0,3)-(1,3))
State: 
.........
312/312 Test #271: python/pytorch/deep_cfr_pytorch_test.py ....................   Passed  166.26 sec
99% tests passed, 1 tests failed out of 312
Total Test time (real) = 379.87 sec
The following tests FAILED:
	136 - xiangqi_test (Subprocess killed)
Errors while running CTest
+ print_tests_failed
+ echo -e '\033[31mAt least one test failed.\e[0m'
At least one test failed.\e[0m
+ echo 'If this is the first time you have run these tests, try:'
+ echo 'python3 -m pip install -r requirements.txt'
+ echo 'Note that outside a virtualenv, you will need to install the '
+ echo 'system-wide matplotlib: sudo apt-get install python-matplotlib'
+ exit 1
If this is the first time you have run these tests, try:
python3 -m pip install -r requirements.txt
Note that outside a virtualenv, you will need to install the 
system-wide matplotlib: sudo apt-get install python-matplotlib
+ cleanup
+ [[ false == \t\r\u\e ]]

lanctot · 2026-04-01T12:19:36Z

Ok so the fix is to keep a variable in XianqiState that counts the number of total moves (see coop_box_pushing for an example: https://github.com/google-deepmind/open_spiel/tree/master/open_spiel/games/coop_box_pushing) and then ends the game in a draw after the maximum number of moves has been played.

You don't need to make the maximum a parameter to the game; you can just put a sensible constant at the top of the header file. But then you also need to return that number in MaxGameLength.

lanctot · 2026-04-01T12:24:29Z

move_number_ works too, nice - thanks.

Arahan-kujur · 2026-04-01T12:24:46Z

Ok so the fix is to keep a variable in XianqiState that counts the number of total moves (see coop_box_pushing for an example: https://github.com/google-deepmind/open_spiel/tree/master/open_spiel/games/coop_box_pushing) and then ends the game in a draw after the maximum number of moves has been played.

You don't need to make the maximum a parameter to the game; you can just put a sensible constant at the top of the header file. But then you also need to return that number in MaxGameLength.

Thanks for the clarification. I currently added the move limit check using the existing move_number_ from the base State class. Would you prefer that I instead add a separate move counter in XiangqiState, similar to coop_box_pushing, and return the same constant from MaxGameLength()?

lanctot · 2026-04-01T12:27:36Z

Nope, no need -- this is perfrect.

Arahan-kujur · 2026-04-01T13:28:38Z

Nope, no need -- this is perfrect.

Great, thanks! Is everything good to merge now, or is there anything else you'd like me to adjust?

lanctot · 2026-04-01T13:49:38Z

Great, thanks! Is everything good to merge now, or is there anything else you'd like me to adjust?

I only took a quick look, but it seems mostly fine. I'm currently on vacation, so I'll import it when I get back to work (after April 14th). There may be 1-2 more things to do if there's something major in the import, but that's rare. So expect it to be merged mid to late April.

Arahan-kujur · 2026-04-01T13:52:09Z

Great, thanks! Is everything good to merge now, or is there anything else you'd like me to adjust?

I only took a quick look, but it seems mostly fine. I'm currently on vacation, so I'll import it when I get back to work (after April 14th). There may be 1-2 more things to do if there's something major in the import, but that's rare. So expect it to be merged mid to late April.

Thanks for taking a look, and enjoy your vacation! Happy to make any adjustments if something comes up during the import.

Arahan-kujur added 2 commits March 30, 2026 22:25

Add Xiangqi (Chinese Chess) implementation to OpenSpiel

6ae8034

rbc disabled fix

889d053

lanctot requested changes Mar 30, 2026

View reviewed changes

open_spiel/games/xiangqi/xiangqi.h Show resolved Hide resolved

lanctot reviewed Mar 30, 2026

View reviewed changes

lanctot added the waiting Waiting to hear back from contributor (tests failed or thread reply / code update required) label Mar 31, 2026

Arahan-kujur added 2 commits March 31, 2026 21:27

Add Xiangqi documentation, test entry, and playthrough for OpenSpiel …

78ffc4b

…integration tests Made-with: Cursor

Playthrough + games.md

7f0886b

lanctot removed the waiting Waiting to hear back from contributor (tests failed or thread reply / code update required) label Mar 31, 2026

Trigger CI tests

b375c51

lanctot approved these changes Apr 1, 2026

View reviewed changes

Add max move limit to prevent infinite Xiangqi simulations

8e4b822

Conversation

Arahan-kujur commented Mar 30, 2026

Uh oh!

lanctot commented Mar 30, 2026

Uh oh!

lanctot left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

lanctot left a comment

Choose a reason for hiding this comment

Uh oh!

Arahan-kujur commented Mar 31, 2026

Uh oh!

lanctot commented Mar 31, 2026

Uh oh!

Arahan-kujur commented Mar 31, 2026

Uh oh!

lanctot commented Apr 1, 2026

Uh oh!

Arahan-kujur commented Apr 1, 2026

Uh oh!

lanctot commented Apr 1, 2026

Uh oh!

lanctot commented Apr 1, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

lanctot commented Apr 1, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

lanctot commented Apr 1, 2026

Uh oh!

Arahan-kujur commented Apr 1, 2026

Uh oh!

lanctot commented Apr 1, 2026

Uh oh!

Arahan-kujur commented Apr 1, 2026

Uh oh!

lanctot commented Apr 1, 2026

Uh oh!

Arahan-kujur commented Apr 1, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

lanctot commented Apr 1, 2026 •

edited

Loading

lanctot commented Apr 1, 2026 •

edited

Loading