Cheng 4.39 release

mar · Post by **mar** » Mon Jul 20, 2015 11:44 pm

New version (temporarily) available here: http://www.crabaware.com/cheng4/cheng4_39.zip
The hosting is a bit lousy so you may have to retry if the file is corrupt.
I guess Emil is on holidays so it should be available soon at the usual http://www.vlasak.biz/cheng

The estimated gain should be around 30 elo (hopefully).
Roughly half of the gain is due to retuning eval using Peter's method.

I changed the versionning a bit so now it's Cheng 4.39 instead of the usual weird cheng4 0.39.

I also fixed a minor bug related to smp (minimum qs depth); I have no idea how (if at all) this will impact smp, we shall see.

For more details see readme.txt.

Graham Banks · Post by **Graham Banks** » Tue Jul 21, 2015 4:16 am

mar wrote:New version (temporarily) available here: http://www.crabaware.com/cheng4/cheng4_39.zip
The hosting is a bit lousy so you may have to retry if the file is corrupt.
I guess Emil is on holidays so it should be available soon at the usual http://www.vlasak.biz/cheng

The estimated gain should be around 30 elo (hopefully).
Roughly half of the gain is due to retuning eval using Peter's method.

I changed the versionning a bit so now it's Cheng 4.39 instead of the usual weird cheng4 0.39.

I also fixed a minor bug related to smp (minimum qs depth); I have no idea how (if at all) this will impact smp, we shall see.

For more details see readme.txt.

Good news.

Thanks Martin.

Frank Quisinsky · Post by **Frank Quisinsky** » Tue Jul 21, 2015 7:20 am

Hi Martin,

good news!
At the moment I am testing "older engines" again. Cheng4 0.39 will be follow after (today your engine is one of my favorits).

Thanks for your work!

Best
Frank

JVMerlino · Post by **JVMerlino** » Tue Jul 21, 2015 7:37 am

mar wrote:New version (temporarily) available here: http://www.crabaware.com/cheng4/cheng4_39.zip
The hosting is a bit lousy so you may have to retry if the file is corrupt.
I guess Emil is on holidays so it should be available soon at the usual http://www.vlasak.biz/cheng

The estimated gain should be around 30 elo (hopefully).
Roughly half of the gain is due to retuning eval using Peter's method.

I changed the versionning a bit so now it's Cheng 4.39 instead of the usual weird cheng4 0.39.

I also fixed a minor bug related to smp (minimum qs depth); I have no idea how (if at all) this will impact smp, we shall see.

For more details see readme.txt.

Congratulations!

jm

cdani · Post by **cdani** » Tue Jul 21, 2015 12:30 pm

Nice!

Ralf Müller · Post by **Ralf Müller** » Tue Jul 21, 2015 1:05 pm

Wow, thanks! I played several games and I'm impressed by it's sensible evaluation and playing style, especially for closed positions. I like it's preference for c4-positions/openings (English Opening/Sicilian...)

Manza · Post by **Manza** » Tue Jul 21, 2015 1:47 pm

Good news!!

petero2 · Post by **petero2** » Tue Jul 21, 2015 2:45 pm

mar wrote:The estimated gain should be around 30 elo (hopefully).

Congratulations!

I measured +61 elo against 0.38 in hyper bullet self play, so my guess is it will be a bit more than +30 in rating lists.

Engin · Post by **Engin** » Tue Jul 21, 2015 3:40 pm

yes, congratulation for the huge improvement each version you made with the texel method, it seems thats work in your engine very well.

badly didnt working on me so much. maybe its there a bug anywhere i didnt find it.

mar · Post by **mar** » Tue Jul 21, 2015 4:20 pm

Thanks guys.
I always divide self-play by 2, I actually measured ~+70 but that was bullet.
I've no idea if/how it will scale at longer TC plus error bars on various lists are quite large so I hope it won't be a fiasco

Cheng 4.39 release

Cheng 4.39 release

Re: Cheng 4.39 release

Re: Cheng 4.39 release

Re: Cheng 4.39 release

Re: Cheng 4.39 release

Re: Cheng 4.39 release

Re: Cheng 4.39 release

Re: Cheng 4.39 release

Re: Cheng 4.39 release

Re: Cheng 4.39 release