Jouni wrote:I think we can consider RobboLito as Rybka3+. It fixes 1) hash use bug 2)
node and depth display 3) bishop promotion and additionally it's lightly stronger and free
Jouni
1. Not true. The hash bug does not affect single core Rybka 3 so this cannot be called a bug in SP R3
What is this bug?
Miguel
2. Not true. Vas has on quite a few occasions mentioned that he simply likes to "real" nodes and depth. In other words he does not like to count nodes which he considers as garbage nodes. And for depth he prefers depth were he has confidence in the accuracy of search for that depth. You can disagree with how he counts this but you will find that no two engine writers use the same counting methods, since it is a preference and not an exact science.
3. Again not true. I am assuming you are refering to the "wrong bishop" knowledge being omitted in Rybka 3. The situation arises when the opposing side with inferior material controls the queening square and the bishop's color is not the same as the queening square and thus it is a dead draw. Robo suffers from this just like R3 does, and I am not sure who came up with the fact that it does not. You can check it out yourself in this position
[d]k7/P7/8/1p2K3/8/P7/5B2/8 w - - 0 107
Actually IMHO in the following position shows that Robo suffers from R3 in a very similar fashion as it does not show a dead draw here. At least not on my computer.
This took almost 100 hours to finish and to prove that Robbo can play long games and knows all the different ways how to cook the "fish".
If more people would test the engine instead making comments without testing it,they would get the idea that this engine is stronger that Rybka 3.
And the strength of the engine is a reason why Vas had to change his mind about releasing Rybka 4.
If Rybka 4 was able to beat this engine we would be all testing new Rybka instead if commenting about Robbo .And this is also "people's project" with so many individuals participating in it and that is what makes it fun.And that project is continuing with MP version coming soon.
I wonder how Vas is going to sell his Rybka 4 "without any warranty" and his internet based Rybka if his engine can not beat engine that is free and easily accessible.Maybe by yelling together with his close followers:clone, clone, clone...
50 games played / Tournament is finished
Name of the tournament: Toga Testing
Site/ Country: XP-2G, United States
Level: Blitz 60/0
Hardware: Intel(R) Pentium(R) 4 CPU 2.00GHz 1994 MHz with 1,024 MB Memory
Operating system: Microsoft Windows XP Professional Service Pack 3 (Build 2600)
solis wrote:This took almost 100 hours to finish and to prove that Robbo can play long games and knows all the different ways how to cook the "fish".
If more people would test the engine instead making comments without testing it,they would get the idea that this engine is stronger that Rybka 3.
And the strength of the engine is a reason why Vas had to change his mind about releasing Rybka 4.
If Rybka 4 was able to beat this engine we would be all testing new Rybka instead if commenting about Robbo .And this is also "people's project" with so many individuals participating in it and that is what makes it fun.And that project is continuing with MP version coming soon.
I wonder how Vas is going to sell his Rybka 4 "without any warranty" and his internet based Rybka if his engine can not beat engine that is free and easily accessible.Maybe by yelling together with his close followers:clone, clone, clone...
50 games played / Tournament is finished
Name of the tournament: Toga Testing
Site/ Country: XP-2G, United States
Level: Blitz 60/0
Hardware: Intel(R) Pentium(R) 4 CPU 2.00GHz 1994 MHz with 1,024 MB Memory
Operating system: Microsoft Windows XP Professional Service Pack 3 (Build 2600)
You are running Rybka 3 in conditions that crippled it, not allowing it to use the remaining 3 cpus of your machine.
Robo is single core engine. Why should Rybka get the advantage of using all 4 cores. Do u consider it fair ? I am sure Mr Popovich just tested how single Robo performs VS single Rybka, and the results speak for themselves
bigo wrote:Everyone is claiming it's actually better than Rybka3 true
Well, its not allowed to link to it. It is updated often and it is quite bugfree.
It is stronger than Rybka 3 on single core without pondering by 70 ELO. Thats the average of my tests, and of others.
Got to agree it is much stronger than Rybka.
The time to move has been fixed in version e4.
Now if the end game would be fixed.
Like the top comm. programs the end
game is weak.
Jouni wrote:I think we can consider RobboLito as Rybka3+. It fixes 1) hash use bug 2)
node and depth display 3) bishop promotion and additionally it's lightly stronger and free
Jouni
1. Not true. The hash bug does not affect single core Rybka 3 so this cannot be called a bug in SP R3
2. Not true. Vas has on quite a few occasions mentioned that he simply likes to "real" nodes and depth. In other words he does not like to count nodes which he considers as garbage nodes. And for depth he prefers depth were he has confidence in the accuracy of search for that depth. You can disagree with how he counts this but you will find that no two engine writers use the same counting methods, since it is a preference and not an exact science.
3. Again not true. I am assuming you are refering to the "wrong bishop" knowledge being omitted in Rybka 3. The situation arises when the opposing side with inferior material controls the queening square and the bishop's color is not the same as the queening square and thus it is a dead draw. Robo suffers from this just like R3 does, and I am not sure who came up with the fact that it does not. You can check it out yourself in this position
[d]k7/P7/8/1p2K3/8/P7/5B2/8 w - - 0 107
Actually IMHO in the following position shows that Robo suffers from R3 in a very similar fashion as it does not show a dead draw here. At least not on my computer.
[d]Q7/1k6/3K4/1p6/8/P7/5B2/8 b - - 0 108
Your diagrams prove nothing because black has a pawn at the b file and
white can force black to move the b pawn when white can play later axb
Here it is a draw because of stalemate but robbolito probably like some other programs with knowledge about blind bishop(for example stockfish and toga) does not evaluate stalemate in the evaluation function and simply evaluate some leaf positions after white plays axb as a win for white.
Stockfish and toga clearly have the knowledge about the blind bishop and your diagram is an exception that has an explanation.
You are running Rybka 3 in conditions that crippled it, not allowing it to use the remaining 3 cpus of your machine.
Miguel
[quote]
Miguel,
these games were played on a single core computer so Rybka could not use 4 cores any anyway.Do you want me to send you the games where Rybka 3 lost the tournaments playing 4 cores against Robbo on single core.Just do the testing and see for yourself.Or go to the site where the Robbo testing is being done and see that Robbo has beat in tournaments all available engines that were also using all available cores.The scores abound but you wont like it. There you will find my reports where Robbo was playing Rybka on 4 cores and that also did not help
And if you are talking about fairness would that be fair to compare the same thing like one core.When MP versions comes what would than be excuse for loosing.And people who have only one core computers are using"crippled' Rybka according to you.So all of them should buy Quads to be able to use Rybka.Just sore loser argument
But to be fair I have analyzed why Rybka lost the match.Here are the findings:
Rybka being postmenopausal had hot flashes so she could not concentrate.
Playing 100 games without the break was causing her a problem.She desperately had to go and pee.
According to her this was old and slow computer.Wanted more cores.
The temperature in the room was not right.Hot flashes as I said.
The room was in the lower floor and she wanted the study upstairs.
There was not enough light in that room even during the day so she missed some good moves and lost.Maybe she needs glasses.
Robbo was singing :I know 1000 ways how to prepare the fish so she was mentally distracted.
The time for the games was to long and it was tiring her .
Some other things I forgot but I would appreciate if Miguel and other Rybka defendants would add more reasons to make this picture more complete.
At the end I would like to thank the authors of Ippolit and RobboLito and especially Norman Schmidt and others from the gang who made testing of these engines such a pleasurable experience.The friendship and camaraderie will be unforgettable.
And the dream goes on !!
Damir wrote:Robo is single core engine. Why should Rybka get the advantage of using all 4 cores. Do u consider it fair ? I am sure Mr Popovich just tested how single Robo performs VS single Rybka, and the results speak for themselves
Rybka is multicore, why should Robbo get the advantage of crippling all the power of Rybka?
You can test whatever you want, but you cannot say "engine A is better than B" if you do cut the potential features of engine B. Authors spend quite a bit of time tuning for SMP for a reason.
The fact is that If you have at home a quad, the engine that will produce better results is Rybka. Today's hardware is multi CPU.
This is fair: You get $1000 to buy hardware, I get $1000 to buy hardware. You install RL, I install rybka 3 and we play a match.
