2.0

What changed?

NNUE

Sentinel uses a small neural network ^{[1] [2]} to evaluate the positions.
[1] (768->128)x2->1
[2] trained on selfplay data

Search

Many features have been added to the search, all of which are written in the TODO.md file.

Elo

I assume the ELO is between 2700 - 2900.

1.02

• UCI fix
• Added UCI go depth x
• Output is without debug logs.

binaries will be added soon

1.01

• Windows compilation fixes
• Fixed movepick
• Fixed segfault
• Added countermoves

1.0

It all began with the 1024 tokens challenge by SebLague. I tried to participate (though I didn’t submit it), but it sparked my interest in chess engine programming. Two years ago, I attempted to create a chess engine in C#, but it didn’t go well [I couldn’t even get a functional move generator working.]

Fast forward to today, and I'm pleased to announce the first version of my engine. I believe it’s quite strong, with an estimated rating of around 2300-2500 Elo blitz, based on local tests against Smallbrain 1.0 and Simbelmyne 1.3.

For more details about the engine’s search and evaluation methods, please check out the README!

As mentioned in the README, I’d like to thank everyone in the Chess Engine Programming Discord community for their support throughout this journey.