Chal - UCI Chess Engine
Author: Naman Thanki
Chal 1.4.1 what's new?
999 lines of C99
Strength
v1.4.1 is estimated at ~2750 Elo, a solid improvement over v1.4.0 (~2700 Elo).
SPRT test vs. v1.4.0 (8+0.08, UHO_Lichess_4852_v1.epd):
Elo: 51.06 +/- 21.82, | LOS: 100% | Games: 610
Notes
It's been more than 3 weeks since I made a release. I've been trying to reduce lines but more than that trying to actually make Chal stronger - and honestly those two things are starting to fight each other in a way that just doesn't feel right. If I keep cutting lines just to compensate for new strength code, it goes against the whole point of this project. So I'd rather sit on it than ship something that doesn't feel like Chal anymore.
When I started this I genuinely had no idea it would go this far. 6 years away from chess programming and somehow ended up here. That means a lot more than any Elo number does.
I'm a little worried it'll be a long time before the next release - I haven't quite figured out how to move forward without compromising what makes this project what it is. But this version, 999 lines, 2750 Elo - I'm really proud of this one. Probably one of the best things I've done since coming back.
Thanks to the chess programming community for being genuinely wonderful, and to the CCRL testers - Gabor, Andres and Graham - for caring enough to test Chal.
And I have to say this properly - none of this would exist today without nescitus (Pawel Koziol). He's been there from the very beginning, as a co-author, as a mentor, and honestly just as someone who believed in this project when it was nothing. Chal is as much his as it is mine.
Strength
v1.4.1 is estimated at ~2750 Elo, a solid improvement over v1.4.0 (~2700 Elo).
SPRT test vs. v1.4.0 (8+0.08, UHO_Lichess_4852_v1.epd):
Elo: 51.06 +/- 21.82, | LOS: 100% | Games: 610
Notes
It's been more than 3 weeks since I made a release. I've been trying to reduce lines but more than that trying to actually make Chal stronger - and honestly those two things are starting to fight each other in a way that just doesn't feel right. If I keep cutting lines just to compensate for new strength code, it goes against the whole point of this project. So I'd rather sit on it than ship something that doesn't feel like Chal anymore.
When I started this I genuinely had no idea it would go this far. 6 years away from chess programming and somehow ended up here. That means a lot more than any Elo number does.
I'm a little worried it'll be a long time before the next release - I haven't quite figured out how to move forward without compromising what makes this project what it is. But this version, 999 lines, 2750 Elo - I'm really proud of this one. Probably one of the best things I've done since coming back.
Thanks to the chess programming community for being genuinely wonderful, and to the CCRL testers - Gabor, Andres and Graham - for caring enough to test Chal.
And I have to say this properly - none of this would exist today without nescitus (Pawel Koziol). He's been there from the very beginning, as a co-author, as a mentor, and honestly just as someone who believed in this project when it was nothing. Chal is as much his as it is mine.
Chal 1.3.2 vs other engines:
| BenBot 1.7.5 | 6/6 | +6 | 6 Partii |
| Kociolek 2.1 | 6/6 | +6 | 6 Partii |
| Zeppelin 1.4.2 | 5/6 | +4 | 6 Partii |
| Kreveta 2.2.5 | 4/6 | +2 | 6 Partii |
| OxideChessEngine 1.0.0 | 3/6 | +0 | 6 Partii |
| Fatalii 0.10.0 | 2.5/6 | -1 | 6 Partii |
| Tunguska 2.0 | 1/6 | -4 | 6 Partii |
| Leaf dev-5382 JA | 1/6 | -4 | 6 Partii |
| Critter 1.6a | 0.5/6 | -5 | 6 Partii |
| 4k-dot-c 8.0 | 0/6 | -6 | 6 Partii |
| Ynode 0234 | 0/6 | -6 | 6 Partii |
| Allaya 1.5 | 4/4 | +4 | 4 Partii |
| FrankyCPP 1.6.0 | 3/4 | +2 | 4 Partii |
| Rusty Rival 1.0.35-rc8 | 3/4 | +2 | 4 Partii |
| Solo 1.6.0 | 1.5/4 | -1 | 4 Partii |
| Halogen 16.4.0 | 0.5/4 | -3 | 4 Partii |
| Raphael 3.2.0 | 0/4 | -4 | 4 Partii |
| Alexander 8.1 | 0/4 | -4 | 4 Partii |
| AllayaChess 1.5 | 2/2 | +2 | 2 Partii |
| Beast 3.3.1 | 2/2 | +2 | 2 Partii |
| Radiance 4.3 | 2/2 | +2 | 2 Partii |
Comments
Post a Comment