> surprisingly readable and surprisingly strong chess engine in 111 lines of Python
Link I get shows 500 lines and it starts with 50 lines of piece-square tables. Maybe it's obvious when you are into the domain but otherwise... that's pretty much of opposite of what I would call "readable".
> I got 111 by deleting the tables in the top, and the UI code in the bottom, and then running 'cloc' on the result. That gave 20 blanks, 56 comments and 111 lines of code. ;-)
Bug report: I got to a late-middlegame position and then the bot crashed because it tried to castle illegally (after its queenside rook had already moved):
```
Uncaught Error: Invalid move: {"from":"e8","to":"c8"}
at Chess.move (chess.js:2530:23)
at cm-chessboard/:75:23
```
There didn't seem to be a way to recover besides refreshing.
Would maybe benefit from a "Checkmate!" or other finish screen, but cool project! Though I agree that it beating you may be more based on your skill level than its, given how it keeps pushing pawns.
The ai source code doesn't seem to have alpha-beta pruning yet. If you add that plus some basic move-ordering (killer move heuristic is easy, powerful, and has a cool name) that will let it search a lot deeper. One nice thing about such improvements is they converge to the same result, just faster, so unlike tweaking the evaluation function you can be pretty sure that you're improving it.
Yes. I did the exact same thing a few weeks ago (in Java, but the idea is to port it to my own programming language). I wrote a terminal UI. Lots of small challenges. Many things to learn, like minimax, bitboards, immutable vs mutable.
I was surprised how quickly I created a chess engine with Claude code a few months ago. It worked quite well. It was bad at recognizing certain chess patterns so I created a lot of tests for the chess engine.
It is terrible in the opening and a little better in the middle game (if you didn't win already, I mean if you played 4-5 random moves).
I think you can improve it a lot by taking into account its own position. I mean, moving forward isn't very important want you are under the menace of a checkmate somewhere else.
I tried playing against it, I didn't have many expectations, but even though I blundered a bishop on move 3 due to a mouse-slip, I could still checkmate it in 6 moves. To me it seemed like it makes random moves.
Chose the most aggressive move (in term of pieces value and check-mate), if none is aggressive, it takes on of the move equally non aggressive.
Didn't remember the depth of the algorithm but it was very simple C code, could check quickly. It should be able to find a mate in 2 or 3 if it was in position of having one.
I didn't check the correctness of the algorithm, just the intention.
I thought it played worse than random moves and couldn't understand how it could beat anyone (no offence to OP).
But if you intentionally hang your pieces, it tends to take them. And it will try to promote pawns in the endgame. So it is possible for it to stumble upon a checkmate, though in my effort where I gave away all my pieces, it instead found the only move to stalemate once it had K+Q+R vs K.
It’s terrible everywhere tbh, it has zero awareness of how dangerous the f7 square is in the opening, you can pretty much just mate it by landing the queen on f7 with a piece covering it every time.
I applaud the effort but not sure I see the point, I had a more capable chess playing program than this on a zx speccy in the 80’s.
It's a fun little AI, but I have another bug report. This time my move is the one that made it crash!
Uncaught Failed to apply player move! cm-chessboard:69:55
promotion result
Object { type: "pieceSelected", square: "c8", piece: "wq" }
cm-chessboard:111:37
Uncaught Error: Invalid move: {"from":"d7","to":"c8","promotion":"q"}
I had forked a rook and king with a pawn. Pawn takes rook, and was going to be offered a promotion, but taking and promoting at once seems to make it crash.
https://www.chessprogramming.org/ is also really interesting, see e.g. https://www.chessprogramming.org/Sunfish and https://www.chessprogramming.org/Quiescence_Search
Link I get shows 500 lines and it starts with 50 lines of piece-square tables. Maybe it's obvious when you are into the domain but otherwise... that's pretty much of opposite of what I would call "readable".
> I got 111 by deleting the tables in the top, and the UI code in the bottom, and then running 'cloc' on the result. That gave 20 blanks, 56 comments and 111 lines of code. ;-)
The file is 500 lines because of the piece square tables (as you mentioned), comments, and the CLI interface logic
Previous HN thread here: https://news.ycombinator.com/item?id=20068651
I always hated that video mode, but it looks good here when the magenta and cyan are limited just to links and headers!
``` Uncaught Error: Invalid move: {"from":"e8","to":"c8"} at Chess.move (chess.js:2530:23) at cm-chessboard/:75:23 ```
There didn't seem to be a way to recover besides refreshing.
I think you can improve it a lot by taking into account its own position. I mean, moving forward isn't very important want you are under the menace of a checkmate somewhere else.
Chose the most aggressive move (in term of pieces value and check-mate), if none is aggressive, it takes on of the move equally non aggressive.
Didn't remember the depth of the algorithm but it was very simple C code, could check quickly. It should be able to find a mate in 2 or 3 if it was in position of having one.
I didn't check the correctness of the algorithm, just the intention.
But if you intentionally hang your pieces, it tends to take them. And it will try to promote pawns in the endgame. So it is possible for it to stumble upon a checkmate, though in my effort where I gave away all my pieces, it instead found the only move to stalemate once it had K+Q+R vs K.
I applaud the effort but not sure I see the point, I had a more capable chess playing program than this on a zx speccy in the 80’s.