Hello from Zalazer

[Zalazer]

Hello community.
Not long ago I started vibe coding using html and have some results. With help of gpt and perplexity AI, i created bunch of games placed in my TG channel https://t.me/ZalazerBot
My last game Chess is playing pretty good against me and that's why i joined this board to compare it to yours versions.
The engine is based on alfabeta and book positions at the start. The most amazing thing for me that the size of the program is about 110kb in one html file, where 50kb is a book, and 60 kb is the engine.
I would like to create the engine stronger to let him beat other engines))))) I believe it now plays near 1800 level. How to check this?

[hgm]

Does your engine support a standard protocol for communicating its moves, such as UCI or CECP? If so, this would allow you to automatically play games against other engines supporting these protocols (mediated by a GUI). The way to measure the Elo rating is to let a play a few hundred games against computer opponents of known rating (e.g. from the CCRL list). This would be tedious if you had to do it by hand.

Also note that computer ratings are not really comparable to human ratings.

[ehenkes]

Also note that computer ratings are not really comparable to human ratings.

Can you explain to me what the connection is between computer ratings (CCRL) and human ratings?

[Zalazer]

Thank you for the answer. I don't have UCI right now but i understood what i need to do with my engine. I will create a new js file that will be ready to play with other engines. I believe i will be able to do that

[ehenkes]

https://github.com/billjordanchess/JavaScript-Chess
A simple chess program for teaching about chess programming for the web. Includes the html code, the javascript code and graphics needed to create a chess engine that can be played on a web page. ...

[hgm]

Can you explain to me what the connection is between computer ratings (CCRL) and human ratings?

Not in detail, because it isn't really known.

But they are measured in virtually non-overlapping pools of players. So even if they are good for comparing strength of players from the same pool, they might not say anything about comparing players from different pools. In particular the scales could be shifted by an unknown amount, which can only be determined by playing hundreds of man-vs-machine games. And this has not been done.

There also is another concern; it is known from determination of engine ratings that results between more similar players (such as different versions of the same engine) tends to exaggerate the rating difference as determined from a larger variety of opponents. And since almost all engines work pretty much in the same way, they are a much more similar group than humans. So basing a rating scale only on engine-engine matches could expand it compared to measuring their rating in a combined pool of engines and humans, dominated by humans.

[Zalazer]

I tried this link to start the game but couldn't

https://billjordanchess.github.io/JavaS ... index.html

I think mine engine is something very similar to it based on the functions listed

[ehenkes]

Perhaps AI, e.g. ChatGPT 5.4, can help you to find logical errors in your code.