Kohai 2.0 coming soon...

[MikeB]

Kohai is a Senpai based engine. Senpai was authored by Fabien Letouzey. Fabien, as many of you know, was also the author Fruit exactly 10 years prior - which at the time of its release ,was the strongest open source engine available and it was a milestone in chess programming as it brought in many more programmers into the field and one could argue that without the release of the open source Fruit, chess engines would not be at the strength as they are now.

Fabien also authored the original polyglot adapter, and what you may not know, he also authored the Othello programs Turtle, Snail and Piloth and the checker programs , Toy and Scan. Scan won the draughts competition at 18th Computer Olympiad held at Leiden in 2015.

Suffice to say ,Fabien is an expert programmer and his Senpai code is a work of art to say the least. With that said, I am deeply humbled that I actually came up with a few lines of code for Senpai , in my fork of Senpai, which I labeled Kohai. In fact , the name Kohai was derived from Senpai and it was out of my deep respect for Fabien that my fork was named Kohai.

https://en.wikipedia.org/wiki/Senpai_and_kōhai

current results:

Code: Select all

576 of 1500 games completed...
time control: 3000+50
Date: 04/09/17 : 23:27:54

576 game(s) loaded
Rank Name         Rating   Δ     +    -     #     Σ    Σ%     W    L    D   W%    =%   OppR 
---------------------------------------------------------------------------------------------------------
   1 Kohai 1.1     3157   0.0   17   17   383  236.0  61.6  154   65  164  40.2  42.8  3072 
   2 Senpai 1.0    3120  37.3   17   17   385  206.5  53.6  118   90  177  30.6  46.0  3090 
   3 Crafty-25.2   3024  96.0   18   18   384  133.5  34.8   74  191  119  19.3  31.0  3138 
---------------------------------------------------------------------------------------------------------

I did not expect ~35 ELO was this change, but if this 35 point ELO gain holds, it will deserve a 2.0 moniker. CCRL testers should be free to test under the Senpai engine category as the frequency for Kohai updates will be about once a year. (Unless somebody wants to add or help me add "syzygy" code).

[cdani]

Thanks for this!
I don't know if you noticed it, but here there is probably an improvement for Senpai:
http://talkchess.com/forum/viewtopic.ph ... enpai+hash

[MikeB]

Thanks for this!
I don't know if you noticed it, but here there is probably an improvement for Senpai:
http://talkchess.com/forum/viewtopic.ph ... enpai+hash

Was not aware, I can close my current test here - since its pretty decisive - has faded a bit, but still gaining in the plus column:

Code: Select all

1200 of 1500 games completed...
time control: 3000+50
Date: 04/10/17 : 01:18:06

1200 game(s) loaded
Rank Name         Rating   Δ     +    -     #     Σ    Σ%     W    L    D   W%    =%   OppR 
---------------------------------------------------------------------------------------------------------
   1 Kohai 1.1     3146   0.0   12   12   799  476.5  59.6  294  140  365  36.8  45.7  3077 
   2 Senpai 1.0    3127  18.8   12   12   800  441.0  55.1  263  181  356  32.9  44.5  3086 
   3 Crafty-25.2   3027 100.4   12   12   801  282.5  35.3  153  389  259  19.1  32.3  3137 
---------------------------------------------------------------------------------------------------------
1200 game(s) loaded 

[kranium]

Thanks for this!
I don't know if you noticed it, but here there is probably an improvement for Senpai:
http://talkchess.com/forum/viewtopic.ph ... enpai+hash

Was not aware, I can close my current test here - since its pretty decisive - has faded a bit, but still gaining in the plus column:

Code: Select all

1200 of 1500 games completed...
time control: 3000+50
Date: 04/10/17 : 01:18:06

1200 game(s) loaded
Rank Name         Rating   Δ     +    -     #     Σ    Σ%     W    L    D   W%    =%   OppR 
---------------------------------------------------------------------------------------------------------
   1 Kohai 1.1     3146   0.0   12   12   799  476.5  59.6  294  140  365  36.8  45.7  3077 
   2 Senpai 1.0    3127  18.8   12   12   800  441.0  55.1  263  181  356  32.9  44.5  3086 
   3 Crafty-25.2   3027 100.4   12   12   801  282.5  35.3  153  389  259  19.1  32.3  3137 
---------------------------------------------------------------------------------------------------------
1200 game(s) loaded 

Hi Michael-
Glad to see someone trying to improve Senpai, especially someone with your enthusiasm,
but, I think more games need to be played here in order to call the result decisive...

Does Σ% represents LOS? If so, then a 59.6% is still up in the air.
Would be intereting to know what LOS is just examining the games between Kohai and Senpai...
(If you load the games into Ordo, it's called CFS and can be found in the 2nd text file (details.txt?))

At 3000+50, you could probably run 20000 games in an overnight test session, and if Kohai was indeed better than Senpai,
LOS would reach eventually 100%.

Have you tried cutechess (w/ sprt enabled)? automatically ending the test when it reaches something close to 100%.

Just some friendly advice from one who enjoys your efforts!
Good luck with your project,

Norm

[cdani]

Also after the useful advice of Norman Schmidt, to see if the Null move hash stuff works, I recommend running at least 30 second games or more, as 3 second ones don't put any special pressure on hash.
If you can't do it, just send me two versions (windows ones), and I will test them if you want.

[MikeB]

Thanks for this!
I don't know if you noticed it, but here there is probably an improvement for Senpai:
http://talkchess.com/forum/viewtopic.ph ... enpai+hash

Was not aware, I can close my current test here - since its pretty decisive - has faded a bit, but still gaining in the plus column:

Code: Select all

1200 of 1500 games completed...
time control: 3000+50
Date: 04/10/17 : 01:18:06

1200 game(s) loaded
Rank Name         Rating   Δ     +    -     #     Σ    Σ%     W    L    D   W%    =%   OppR 
---------------------------------------------------------------------------------------------------------
   1 Kohai 1.1     3146   0.0   12   12   799  476.5  59.6  294  140  365  36.8  45.7  3077 
   2 Senpai 1.0    3127  18.8   12   12   800  441.0  55.1  263  181  356  32.9  44.5  3086 
   3 Crafty-25.2   3027 100.4   12   12   801  282.5  35.3  153  389  259  19.1  32.3  3137 
---------------------------------------------------------------------------------------------------------
1200 game(s) loaded 

Hi Michael-
Glad to see someone trying to improve Senpai, especially someone with your enthusiasm,
but, I think more games need to be played here in order to call the result decisive...

Does Σ% represents LOS? If so, then a 59.6% is still up in the air.
Would be intereting to know what LOS is just examining the games between Kohai and Senpai...
(If you load the games into Ordo, it's called CFS and can be found in the 2nd text file (details.txt?))

At 3000+50, you could probably run 20000 games in an overnight test session, and if Kohai was indeed better than Senpai,
LOS would reach eventually 100%.

Have you tried cutechess (w/ sprt enabled)? automatically ending the test when it reaches something close to 100%.

Just some friendly advice from one who enjoys your efforts!
Good luck with your project,

Norm

Thanks Norm , appreciate the kind words. Let me know (via PM) if there is anything I can do to help you get a native Fire 5.0 out for the macOS. Would be a great gift to those of us on a macOS. Naturally, I would sign an NDA to help make it happen.

best regards,

Michael