Since you’re here...

We hope you will consider supporting us today. We need your support to continue to exist, because good entries are more and more work time. Every reader contribution, however big or small, is so valuable. Support "Chess Engines Diary" even a small amount– and it only takes a minute. Thank you.
============================== My email: jotes@go2.pl



Titan 1.1.0 JA - new version chess engine


Titan - UCI chess engine
Rating Chess Engines Diary CEDR=3645

This project is a chess engine that was developed as a learning experience and passion project. It is capable of using the UCI interface to communicate with most chess GUIs, and well past capable of beating me for better or for worse...

The Makefile supports two options. The first entry is capable of being built on stable, and utilizes compiler autovectorization for neural network updates and evaluation. The second option requires both an AVX512 capable cpu and the nightly compiler, as SIMD intrinsics in rust have not stabilized. I expect this will segfault on a non-AVX512 capable cpu.

 I've made somewhat sporadic and inconsistent progress on this project over the past year or so, and while it doesn't implement anything terribly original, (what it did was superseded -- RIP screlu181), I've learned a lot. It likely has quite a few bugs, and there's several areas of entirely illogical behavior that I know of, though others almost certainly exist. I lack the time and hardware to want to continue developing at the moment.
Titan has a fairly average set of features for the modern era of chess engines. It uses a single-layer, king bucketed, horizontally mirrored NNUE trained on Leela data for evaluation.
I would conservatively estimate Titan's strength to be around 3450 Elo on a single core, but as mentioned before, I don't currently have the time or hardware to test that with any level of statistical certainty.

v.1.1.0:
This release includes fixes to a few very crucial bugs

No longer crashes if given negative time as part of a UCI command (other UCI values will likely still crash, but they shouldn't ever occur like time can). Thanks to Cosmo for bringing this to my attention
ucinewgame no longer resets the thread UCI option
SMP actually works now

Titan 1.1.0 vs other engines:

Berserk 1353.5/137-30137 Games
Ethereal 14.2560.5/124-3124 Games
Alexandria 7.0.057/123-9123 Games
Chess-System-Tal-254/108+0108 Games
Uralochka 3.41a JA48/94+294 Games
PlentyChess 3.0.038/91-1591 Games
Obsidian 13.035.5/86-1586 Games
Clover 6.236.5/72+172 Games
Caissa 1.2031/71-971 Games
RubiChess 2024011228.5/62-562 Games
RubiChess 2024081724.5/53-453 Games
Dragon 3.319.5/53-1453 Games
Viridithas 15.0.021.5/48-548 Games
Caissa 1.2119/44-644 Games
Velvet 8.0.023.5/43+443 Games
Motor 0.7.021/43-143 Games
Lizard 10.422/41+341 Games
Peacekeeper 3.0122/40+440 Games
Motor 0.8.0 JA18.5/40-340 Games
Alexandria 7.1.017.5/40-540 Games
Obsidian 12.2916/40-840 Games
PlentyChess 2.0.016.5/38-538 Games
Caissa 1.1918/36+036 Games
Seer 2.8.018/36+036 Games
Integral v617/36-236 Games
Uralochka 3.42 dev716.5/36-336 Games
Berserk 2024111915.5/36-536 Games
PlentyChess 3.0.113/32-632 Games
Stockfish 1713/31-531 Games
Peacekeeper 3.0017.5/30+530 Games
Velvet 8.1.1 JA14/30-230 Games
Caissa 1.1813.5/30-330 Games
Obsidian 14.23 SE12.5/30-530 Games
Lizard 11.115.5/29+229 Games
Stormphrax 1.0.015/28+228 Games
Viridithas 13.0.015/28+228 Games
Viridithas 14.0.012.5/26-126 Games
Clover 8.1 JA12/26-226 Games
Stormphrax 6.0.011/25-325 Games
Starzix 6.011.5/24-124 Games
Clover 7.0 JA11/24-224 Games
Yuliana 5.08/24-824 Games
Lizard 10.59/22-422 Games
Stockfish 16.17/22-822 Games
Clover 8.0.110/21-121 Games
Wasp 7.0016/20+1220 Games


Windows/Linux & Android builds - Jim Ablett compiles:


Comments