New Engine: Scoria

[ianmtv]

Hello everyone! I'm pretty new into the chess programming world and I've been on and off of it for the last handful of months, but I think I've finally been able to developed my first (somewhat decent) chess engine. Introducing Scoria 3, a chess engine written in Java. (In the latest release I have attached .jar and .exe binaries)

https://github.com/iannathan-k/scoria

The most recent version is Scoria 3, (v3.8.51) and I estimate it plays at around ~2000 ELO based on my individual (but definitely amateur) testing. It was just around 200 games against the different Stockfish 17.1 skill levels, and it landed between levels 4 and 5 which is where I got the estimate from.

I'm hoping to further improve by switching to bitboards in Scoria 4, but there is still a lot of learning to be done. My goal is to one day reach ~3000, but I think this is a good starting point. If anyone has any tips or suggestions, I am all ears. Thanks for reading!

[Gabor Szots]

Hello Ian,

I am testing 3.8.51, using the jar file, and first results show that its rating will be around 1700.

[Sylwy]

Hello Ian,

........ first results show that its rating will be around 1700.

Over Blank 1.3.0 (1642 Elo) ?

[ianmtv]

Hello,

I must have significantly overestimated the rating of Scoria 3, but I'm not sure whether it can be as low as below 1600. But either way it just goes to show how much room there is to improve. I'm now working on Scoria 4 with bitboards and I'm able to generate moves ~3x faster than before. Through my personal testing, I ran around 200 games at movetime 1000ms and got this result.

Code: Select all

Rank Name             Elo    +    - games score oppo. draws
   1 master-skill-5   142   70   69    50   52%   127   12%
   2 Scoria_v3.8.51   127   38   37   200   70%   -32   13%
   3 master-skill-4    17   68   71    50   35%   127   18%
   4 master-skill-3   -85   71   80    50   22%   127   16%
   5 master-skill-2  -201   83  105    50   13%   127    6%

Also, checking CCRL 40/2 Archive I noticed that KasparovChess 1.0.0 x64 has a rating of approximately 1618, which would put scoria at 1618 + 382 or 2000. And on the Blitz page Cupcake 1.1c 64-bit is rating around 2008 which would put scoria at 2008 - 301 or 1707. I'm quite new to this forum, so this is probably a lack of understanding on my part like about how Blitz vs 40/15 is calculated, so any clarification would be greatly appreciated. Apologies in advance for my naivety.

[Sylwy]

Hello,

I must have significantly overestimated the rating of Scoria 3, but I'm not sure whether it can be as low as below 1600. But either way it just goes to show how much room there is to improve. I'm now working on Scoria 4 with bitboards and I'm able to generate moves ~3x faster than before. Through my personal testing, I ran around 200 games at movetime 1000ms and got this result.

Code: Select all

Rank Name             Elo    +    - games score oppo. draws
   1 master-skill-5   142   70   69    50   52%   127   12%
   2 Scoria_v3.8.51   127   38   37   200   70%   -32   13%
   3 master-skill-4    17   68   71    50   35%   127   18%
   4 master-skill-3   -85   71   80    50   22%   127   16%
   5 master-skill-2  -201   83  105    50   13%   127    6%

Also, checking CCRL 40/2 Archive I noticed that KasparovChess 1.0.0 x64 has a rating of approximately 1618, which would put scoria at 1618 + 382 or 2000. And on the Blitz page Cupcake 1.1c 64-bit is rating around 2008 which would put scoria at 2008 - 301 or 1707. I'm quite new to this forum, so this is probably a lack of understanding on my part like about how Blitz vs 40/15 is calculated, so any clarification would be greatly appreciated. Apologies in advance for my naivety.

There is no need to apologize. It's just a matter of statistical dispersion. A correct rating requires many games and many opponents. My micro tournament with five chess engines written in Java has no relevance regarding the rating. I'm just curious if your engine (in version 3.8.52_ja-so the Windows executable) can match BLANK 1.3.0 (an engine tested a lot by me). Personally I think Scoria is a bit weaker (I watched some games). It's about the evaluation between the two. Sometimes even a weaker engine can be inconvenient for a relatively better engine (the game of architectures in the structure of the engines). That's all. Thank you very much for the 88th free chess engine project written in Java!