Not logged inRybka Chess Community Forum
- - By -valdimor- (**) [be] Date 2007-02-25 21:46
is there a Toga II 1.3?or are they fake on playchess

and what is the best books i can use for toga that work in chesspartner
Parent - - By peter2804 (**) [gb] Date 2007-02-26 03:31 Edited 2007-02-26 04:17
Go to 'Superchessengine.com  click on Toga look down the page to find Toga 11 1.3x4 . You reminded me when i read your message and i have just downloaded it so i don't know what its like yet. You will need Winrar to download it. I don't use chesspartner but i think the Toga books are on the site as well. (Update.. I have just watched someone with Toga 1.3x4 on Playchess com in a 5 game against Rybka 2.1d3 mp draw in 133 moves. So it looks like a good engine. Regards
Parent - - By Vempele (Silver) [fi] Date 2007-02-26 06:32
I did a quick test tournament with the newest Toga:

-Performance was disappointing, 3.5/21 against opposition about the level of 1.2.1. I didn't have time for more games.
-Bitbases apparently don't work (correctly) if you don't have the 5-man ones (340 MB).

I think turning off Extended History Pruning (=PV history reductions) should help, as it made a modified 1.2.1 significantly weaker in my tests. Note that the whole history pruning scheme is very different from 1.2.1.
Parent - - By Vempele (Silver) [fi] Date 2007-02-26 18:15 Edited 2007-02-26 18:25
Ignore my previous test result. I can't find anything else wrong with it other than not using bitbases, but the improvement with 3-4-5 bitbases is tremendous, and it's scoring just fine. It's too early to say anything else, though.

[Edit] I forgot to mention I turned off Extended History Pruning.
Parent - - By -valdimor- (**) [be] Date 2007-02-27 20:31
the Toga II 1.3 x4 download file on superchessengine is not a real engine file?

than how do you install it
Parent - - By joshua2 (*) [us] Date 2007-02-27 20:35
It is real, its just a uci engine instead of a chessbase .eng file. It depends on the interface you are using. Free is Arena. Best is a probably a chessbase one.
Parent - - By -valdimor- (**) [be] Date 2007-02-27 21:08
never mind, i just found out you need Winrar to open the file

thx
Parent - - By -valdimor- (**) [be] Date 2007-02-28 12:03
i ran a little test

windows XP AMD Athlon 1800+,grandmasters opening book,nalimov 3/4/and some 5

Toga II 1.3X4   +6 =8 -6
Toga II 1.2.1a  +6 =8 -6

20 games
timecontrol 2 minutes + 10 sec/move
Parent - By herakleitos (***) Date 2007-03-04 03:39
Fritz Interface, Windows XP, Athlon 3200+ single processor, 2 Gb RAM, 512 MB hash for each engine, 64MB for Nalimov 3-4-5-6(some) tablebases

G/10

Toga 1.3x4      +4-1=1
Toga 1.2.1a     +1-4=1

Not a rigorous test, certainly, but at least suggestive. The new version also does better against the Rybkas than 1.2.1a does.

The new version seems a bit quicker, too.
Parent - By herakleitos (***) Date 2007-03-04 03:44
Fritz Interface, Windows XP, Athlon 3200+ single processor, 2 Gb RAM, 512 MB hash for each engine, 64MB for Nalimov 3-4-5-6(some) tablebases. Rybka II openings book (including for Toga).

Rybkas-Toga  2007, G/10

                                            1             2              3              4   
1   Rybka 2.3 LK 32-bit          ****       ½½11       ½½½½     1½½½   7.5/12
2   Rybka 2.3 32-bit              ½½00       ****       ½½½1      ½½1½   6.0/12
3   Toga II 1.3x4                  ½½½½    ½½½0       ****       ½1½0    5.5/12
4   Rybka 2.2n2 32-bit           0½½½     ½½0½     ½0½1        ****     5.0/12
Parent - - By herakleitos (***) Date 2007-03-04 13:45
Fritz Interface, Windows XP, Athlon 3200+ single processor, 2 Gb RAM, 512 MB hash for each engine, 64MB for Nalimov 3-4-5-6(some) tablebases.

Toga Match  2007, G/10

Toga II 1.3x4        110½110½110½½½½½½½0½   11.0/20
Toga II 1.2.1a       001½001½001½½½½½½½1½    9.0/20
Parent - - By ernest (****) [fr] Date 2007-03-04 19:06

> 64MB for Nalimov 3-4-5-6(some) tablebases.
>
> Toga Match 2007, G/10
>


What do you mean?
Toga doesn't use tablebases...!  (and what is 64MB for Nalimov 3-4-5-6(some)? should be GigaBytes...)
Maybe you mean bitbases...
Parent - By Vempele (Silver) [fi] Date 2007-03-05 06:08

> (and what is 64MB for Nalimov 3-4-5-6(some)? should be GigaBytes...)


He probably means tablebase cache. Or bitbase cache. Keeping them all in RAM would be expensive...

Hopefully the interface is using the TBs to adjudicate games.
Parent - - By herakleitos (***) Date 2007-03-06 03:25
Toga II most certainly uses endgame tablebases, as any quick trial demonstrates.

For this match I devoted 64 MB to in-memory tablebase storage. This is a small amount, really -- not expensive -- when one has 2 Gb of RAM.
Parent - By Vempele (Silver) [fi] Date 2007-03-06 05:39 Edited 2007-03-06 05:42

> Toga II most certainly uses endgame tablebases, as any quick trial demonstrates.


No, it doesn't. What's its analysis of this position?

8/6k1/5q2/p7/8/6K1/5RP1/8 b - - 0 1


Any engine with the 5-men tbs will instantly show 1...Qxf2+ as the best move with a mate score. Without tablebases it'll be only about 10 pawns.

[Edit] Or, if you have KQPkrp, use this one:

8/6k1/5q2/p7/8/6K1/5RPP/8 b - - 0 1


Again, only a mate in x will suffice.
Parent - - By ernest (****) [fr] Date 2007-03-06 20:14
Sorry, Toga has never used Nalimov tablebases (they are not open source)

However, the latest Toga uses Shawul's bitbases, on disk.
Parent - - By herakleitos (***) Date 2007-03-06 23:05
In my experience, Toga II has used the rook and pawn tablebases (in the Fritz interface) for the rook endings I've looked at. It gives the same quick and accurate response that I get with Rybka or Fritz or Hiarcs. It also uses them on king and pawn endings and bishop endings. I just tried a number of them and I get the immediate response that the other two programs give.

I tried Toga and the tablebases on the two positions above, and, you're right, it doesn't use them. The other programs *do* on these positions, so I in fact have the tablebases for them.

From my trials, Shredder doesn't use the five and six piece endings unless they are installed from the commercial 5- or 9-DVD product, in which case it behaves like Fritz or Rybka in these endings.
Parent - - By Vempele (Silver) [fi] Date 2007-03-07 09:30

>In my experience, Toga II has used the rook and pawn tablebases (in the Fritz interface) for the rook endings I've looked at.


That's what I meant by 'having the GUI adjudicate tablebase positions'. The Fritz GUI doesn't tell the engines anything, it just automatically executes the best moves for both sides until mate/3-fold repetition/50-move rule.
Parent - By herakleitos (***) Date 2007-03-08 18:41
You're completely right about this. My misunderstanding of the GUI behavior. Thanks for the correction.

Bob
Parent - By herakleitos (***) Date 2007-03-06 23:17
Fritz Interface, Windows XP, Athlon 3200+ single processor, 2 Gb RAM, 512 MB hash for each engine, 64MB for Nalimov 3-4-5-6(some) tablebases, Rybka II openings book.

40/10, 40/10, etc.
                       
1   Rybka 2.3.1 32-bit   0½01½½11½1½11111½½1½½101111½½½0111½1½011     28.0/40
2   Toga II 1.3x4          1½10½½00½0½00000½½0½½010000½½½1000½0½100     12.0/40
Parent - - By Dragon Mist (****) [hr] Date 2007-02-28 12:39
I try to download Toga 1.3X from superchessengine.com, and when I come to TogaII page, and click download link, my browser reports no such link or something alike??

Dragon Mist
Parent - - By -valdimor- (**) [be] Date 2007-02-28 13:24
you need winrar
Parent - - By RedAlien (*) [hr] Date 2007-02-28 21:19
Hi,
Is this new Toga version mp or not?
Parent - By Vempele (Silver) [fi] Date 2007-03-01 09:29
It's not.
Parent - - By Dragon Mist (****) [hr] Date 2007-03-01 10:16
Just having Winrar won't help, namely, right click and Save target as ... worked.

Dragon Mist
Parent - - By RedAlien (*) [hr] Date 2007-03-01 13:49
yes it would be weird to put mp engine for free...it would almost be a crime:p
Parent - - By jaeger1975 (**) [pe] Date 2007-03-02 22:57
Hi guys

I have tested the last version of Toga against Fritz 10 and it seems to be stronger than Fritz 10!! Does anybody had similar results?
Parent - By -valdimor- (**) [be] Date 2007-03-03 14:17
i did not test it against fritz but against spike 1.2 turin it was 78% for toga and 22% spike
Parent - - By herakleitos (***) Date 2007-03-06 03:28
In my experience, Toga II scores 55% against Fritz 10 single processor version, with a neutral openings book. My friends tell me that it's very slightly worse against Deep Fritz 10 on a fast machine.
Parent - - By -valdimor- (**) [be] Date 2007-03-21 07:27
hello

what does this mean: Toga Lazy Eval?

i have set it to "false" in the engine setting,i tought it had something to do with closed positions

is this the best setting?
Parent - - By Vempele (Silver) [fi] Date 2007-03-21 09:32 Edited 2007-03-21 09:37
It makes the engine slightly faster. Basically, Toga first does the lazy eval and checks whether the score is within [alpha-margin, beta+margin[. If it isn't, the lazy eval score is returned, else it continues with the eval. It's pretty safe and should be left on IMO.

It's a little different with 1.2.1a, though: it also does LE when doing futility pruning without adjusting the margin (should be [alpha-futilitymargin-lazymargin, alpha-futilitymargin+lazymargin[ or no LE as 1.3x4 does in this case). For some reason this doesn't seem to decrease strength, in fact the opposite.

> i have set it to "false" in the engine setting,i tought it had something to do with closed positions


Toga has very little knowledge of closed positions, it's mostly just a penalty for bishops (for having many own pawns on their colour) and the lack of rooks on open files. It has nothing to do with LE.
Parent - - By -valdimor- (**) [be] Date 2007-03-21 11:20
note that i'm only using it for play against humans not engines

do you still say "true" is better?
Parent - By Vempele (Silver) [fi] Date 2007-03-21 11:33
Yes. There's a tiny tradeoff in accuracy for extra speed. The opponent doesn't matter. LE mainly applies to positions with unbalanced material (compared to alpha), anyway.

[unrelated to your question] There was an error in my last post: should have been [alpha-margin, beta+margin] etc.

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill