How much it helps your engine?
I've still played only a small set of [15 sec] games (~200) and doesn't see any improve for SmarThink (version w/o TBs 4 score points better at present). I'm probing in root and in all non-quiescence nodes where rule50 == 0.
It's interesting if I made something wrong or not)
Syzygy 5-piece Elo gain
Moderator: Ras
-
Sergei S. Markoff
- Posts: 227
- Joined: Mon Sep 12, 2011 11:27 pm
- Location: Moscow, Russia
Syzygy 5-piece Elo gain
The Force Be With You!
-
Frank Quisinsky
- Posts: 7232
- Joined: Wed Nov 18, 2009 7:16 pm
- Location: Gutweiler, Germany
- Full name: Frank Quisinsky
Re: Syzygy 5-piece Elo gain
Hi Sergei,
nice that you are working on it. I made some experiments for some years (around in the year 1999). At this time of course with a normaly harddisk and Crafty / Capture chess programs.
Crafty:
comparing with 4-pieces to no TBs in game play = +10 Elo
comparing with 5-pieces to no TBs in game play = +10 Elo
Capture:
comparing with 4-pieces to no TBs in game play = +15 Elo
comparing with 5-pieces to no TBs in game play = +20 Elo
And the French Capture programmer wrote me that he is very disappointed that I am using 4-pieces (in the time a chess tournament is still running on my older site) because his program is optimated with 5-pieces. Very rarely that I read that, but he is right. Capture are a chess program available on a Commercial Winboard CD with Gandalf and LambChop, others from Gambit-Soft.
Later I do the same with Fruit, testing official in one of my rating list ...
Fruit:
comparing with 4-pieces to no TBs in game play = 5 Elo
I think that 5-pieces could give 10 Elo in average if we are using in game play longer time controls. How aggressive TBs working is the next question. Capture is very aggressive and using 5-pieces if 15 or more pieces on board, processor have only 30% (again on normaly HD) if 5-pieces are still working.
For eng-eng matches I think it's much better to use 4-pieces and with SSD I think 5-pieces are OK too.
I am waiting of SmarThink with fully TB support. Your engine is very strong in endgames so it's possible that you have an advantage from 5 Elo only. Not important, 5-pieces or more are very interesting for endgame analyzes.
On the other hand ...
Looking in Ktulu!
Rahman told me that Ktulu have extra knowledge in rook endgames and 80% of all rook endgames are optimated with chess knowledge. His Special topic in Ktulu development. Ktulu are also very very strong in other endgames. So Rahman told me ... I don't need TBs.
Again, I think also chess programs, stronger in endgames should have it.
Keep up the good work around SmarThink !!
Best
Frank
For myself in testing engines are 4-pieces important only. More nice if moves in clearly endgames comes directly without to wait of it. To use resign=off is very boring but important for statistics. So I hope that engines have TB Support and I have a bit more fun if I see such games on my PCs.
nice that you are working on it. I made some experiments for some years (around in the year 1999). At this time of course with a normaly harddisk and Crafty / Capture chess programs.
Crafty:
comparing with 4-pieces to no TBs in game play = +10 Elo
comparing with 5-pieces to no TBs in game play = +10 Elo
Capture:
comparing with 4-pieces to no TBs in game play = +15 Elo
comparing with 5-pieces to no TBs in game play = +20 Elo
And the French Capture programmer wrote me that he is very disappointed that I am using 4-pieces (in the time a chess tournament is still running on my older site) because his program is optimated with 5-pieces. Very rarely that I read that, but he is right. Capture are a chess program available on a Commercial Winboard CD with Gandalf and LambChop, others from Gambit-Soft.
Later I do the same with Fruit, testing official in one of my rating list ...
Fruit:
comparing with 4-pieces to no TBs in game play = 5 Elo
I think that 5-pieces could give 10 Elo in average if we are using in game play longer time controls. How aggressive TBs working is the next question. Capture is very aggressive and using 5-pieces if 15 or more pieces on board, processor have only 30% (again on normaly HD) if 5-pieces are still working.
For eng-eng matches I think it's much better to use 4-pieces and with SSD I think 5-pieces are OK too.
I am waiting of SmarThink with fully TB support. Your engine is very strong in endgames so it's possible that you have an advantage from 5 Elo only. Not important, 5-pieces or more are very interesting for endgame analyzes.
On the other hand ...
Looking in Ktulu!
Rahman told me that Ktulu have extra knowledge in rook endgames and 80% of all rook endgames are optimated with chess knowledge. His Special topic in Ktulu development. Ktulu are also very very strong in other endgames. So Rahman told me ... I don't need TBs.
Again, I think also chess programs, stronger in endgames should have it.
Keep up the good work around SmarThink !!
Best
Frank
For myself in testing engines are 4-pieces important only. More nice if moves in clearly endgames comes directly without to wait of it. To use resign=off is very boring but important for statistics. So I hope that engines have TB Support and I have a bit more fun if I see such games on my PCs.