tomitankChess 5.1

• network is trained with 4M example instead of 2.7M as in the previous version. (same size 768x16x1)

• better hash table usage

• around 20 elo better than previous version

Estimated elo: 2900.

Since JavaScript is about 4-5x slower than C, it would be around 3100 or above in C. (With AVX).

JavaScript uses only one core. With 4 cores, in C it would be near to 3300.

Have fun with tomitankChess!

tomitankChess 5.0

This is quite different from the SF solution.

• network is only 768x16x1. Smaller than I've seen before.

• it does not replace evaluation, it only compensates that. (*)

• trained only with 2.7M example.

• The network is not compatible with other engines (*) and it's integrated into the code. (hardcoded)

• I wrote the network from scratch. I don't use machine learning platform.

• I have around 10 elo better net (768x32x1) in fix depth test, but JavaScript is too slow, and don’t have AVX or similar methode in JavaScript (so i use vanilla approach). Essentially the smaller net has nearly the same strength in both short and long TC tests.

The training method is similar to that described by Andrew, thanks to him for the explanation.
But the implementation is unique and I have not received any help.
http://talkchess.com/forum3/viewtopic.p ... 72#p874960

This net is not perfect, and HCE slows that down further, but i win around 60-70 or even more. (i don't tested deeply.) In C it would be around 140-150.

I can try to use more examples later, but it’s pretty boring for me.

I also plan to turn off HCE and use only NN, but (probably) the authors of Halogen do this better.

I tried to go my own way, so feel free to send this version to the CCRL list.

Estimated elo: near to 2900.

Since JavaScript is about 4-5x slower than C, it would be around 3100 or above in C. (With AVX).

JavaScript uses only one core. With 4 cores, in C it would be near to 3300.

I did not tune the search. It could probably be even stronger.

Have fun with tomitankChess!

tomitankChess 4.2

tomitankChess 4.2

• revisited code
• rewrite pawn eval
• fixed a bug that's caused problem in IE
• Around 10 elo stronger than previous version.

tomitankChess 4.0

tomitankChess 4.0

• see pruning
• better lmr
• rewrited pawn eval
• new pawn shield code
• new see (add discovered checks)
• lots of little things (revisited code)
• Around 40 elo stronger than previous version.

tomitankChess 3.0

tomitankChess 3.0

• new pawn shield code (idea from Senpai 2, but a bit better)
• use 3x3 squares king ring (idea from Senpai 2)
• new king safety code (simple, but bit better)
• all parameters scoring by logistic regression
• better time management (take more time when unsure)
• 4-way transposition table
• lots of little things (revisited code)
• Around 100 elo stronger than previous version.

tomitankChess 2.1

tomitankChess 2.1

• Revisited code (threats, move gen, some bug-fix, etc..)
• Store eval score in hash
• Smaller king-ring
• Added logo.png
• Strength: similar to the previous version, maybe a bit stronger

tomitankChess 2.0

tomitankChess 2.0

• the first full BitBoard JavaScript chess engine on the World
• Better evaluation (pawn-safe mobility and king safety, threats, better passed pawn, etc..)
• Better search (better aspiration window and delta pruning etc..)
• Typed Arrays Hash table for better memory usage
• Pawn hash
• lots of little things
• Around 100 elo stronger than previous version.

tomitankChess 1.5

tomitankChess 1.5 (Last Mailbox Version)

• New UI interface code (now working with JSUCI as well)
• Tuned Search algorithm
• Use TT at depth == 0
• ~2590 elo (Strongest JavaScript Chess Engine /2017.12.03/)