100 long games Rybka 4 vs Houdini 1.03a

[Steve B]

it is getting a bit tiresome for you to be constantly attacking the moderators
i notice you did not attack us when you complained about a post and we removed it

I usually do not complain to moderation and implying this next to for example Mr. Banks who has more complaints than all other members together is just ridiculous. I did not even complain for the post few lines up when I got asked "from which ass did I crawl" which would be removed by any moderator with a bit of dignity without even complaining.

trying running for moderator on a platform that you will remove all posts dealing with tournament results of commercial engines even before members can answer to it and challenge the results

One thing is providing at least tournament results, TC, play conditions, if not PGN and completely another thing is just repeating like parrot "Rybka 4 is the strongest on LTC, with and without ponder, and on big hardware" without ever offering anything at all but an imaginary respect of the chess community (actually there is respect from Rybka forum members and probably Vas, but nobody with integrity gives a damn about it).
You are obviously pretending not to see the difference and that only tells about you.

the fact that Graham posted at almost the exact same time i did has nothing to do with my reply to you
Grahams opinions are his own and it is common knowledge that we have had strong differences of opinion between us in the past as well as general agreement on other issues

having said that i know of no moderator in the past 13 years here who would consider your interpretation of the charter to be a reasonable one
the point of my reply to you was to explain to you why posts like Ansari's will not be removed so you do not continue to imply there is some secret unknown reason at play here
your attempt to portray me and\or the other moderators as somehow biased in favor of any particular commercial engine is beyond ridiculous and borders on the absurd given the fact that we ran on a platform allowing links and full discussion of Ippo and family engines
you have the reason whether you agree with it ..or choose to believe it or not
Steve

[Steve B]

The Mod team has removed several posts in this thread due to personal attacks
if you are missing a post we are sorry for the inconvenience

[Laskos]

The Mod team has removed several posts in this thread due to personal attacks
if you are missing a post we are sorry for the inconvenience

Fine, I will reiterate, Rybka 4 is so weak compared to some free private and even open source engines because in its progress there were used unqualified testers with no elementary knowledge in statistics, who do not know what LOS is, how the error margins are calculated, and what essential GUI's are there for testing purposes. They are disqualified as testers and as propagandists for a commercial engine like Rybka 4. In fact, I would say even more, most, if not all rating lists presented here, are biased towards commercial engines, being part of the fuss showing how strong they are. They are weak and period.

For everyone, if you want the best engines out there, the most stable, the best for analysis, best bitbases, go only for free engines, they are at least one year ahead of the best commercial engines.

That's fine now?

Kai

[tomgdrums]

The Mod team has removed several posts in this thread due to personal attacks
if you are missing a post we are sorry for the inconvenience

Fine, I will reiterate, Rybka 4 is so weak compared to some free private and even open source engines because in its progress there were used unqualified testers with no elementary knowledge in statistics, who do not know what LOS is, how the error margins are calculated, and what essential GUI's are there for testing purposes. They are disqualified as testers and as propagandists for a commercial engine like Rybka 4. In fact, I would say even more, most, if not all rating lists presented here, are biased towards commercial engines, being part of the fuss showing how strong they are. They are weak and period.

For everyone, if you want the best engines out there, the most stable, the best for analysis, best bitbases, go only for free engines, they are at least one year ahead of the best commercial engines.

That's fine now?

Kai

I think it depends on what you are using the engine for. I have had better luck with commercial engines such as Hiarcs and Shredder...but that is me.

[Albert Silver]

The Mod team has removed several posts in this thread due to personal attacks
if you are missing a post we are sorry for the inconvenience

Fine, I will reiterate, Rybka 4 is so weak compared to some free private and even open source engines because in its progress there were used unqualified testers with no elementary knowledge in statistics, who do not know what LOS is, how the error margins are calculated, and what essential GUI's are there for testing purposes. They are disqualified as testers and as propagandists for a commercial engine like Rybka 4. In fact, I would say even more, most, if not all rating lists presented here, are biased towards commercial engines, being part of the fuss showing how strong they are. They are weak and period.

For everyone, if you want the best engines out there, the most stable, the best for analysis, best bitbases, go only for free engines, they are at least one year ahead of the best commercial engines.

That's fine now?

Kai

What is your statistical basis (as you keep touting) for your comment on the free engines being a year ahead of the best commercial engines?

[Graham Banks]

The Mod team has removed several posts in this thread due to personal attacks
if you are missing a post we are sorry for the inconvenience

Fine, I will reiterate, Rybka 4 is so weak compared to some free private and even open source engines because in its progress there were used unqualified testers with no elementary knowledge in statistics, who do not know what LOS is, how the error margins are calculated, and what essential GUI's are there for testing purposes. They are disqualified as testers and as propagandists for a commercial engine like Rybka 4. In fact, I would say even more, most, if not all rating lists presented here, are biased towards commercial engines, being part of the fuss showing how strong they are. They are weak and period.

For everyone, if you want the best engines out there, the most stable, the best for analysis, best bitbases, go only for free engines, they are at least one year ahead of the best commercial engines.

That's fine now?

Kai

What is your statistical basis (as you keep touting) for your comment on the free engines being a year ahead of the best commercial engines?

Dreams are free also.

[Laskos]

The Mod team has removed several posts in this thread due to personal attacks
if you are missing a post we are sorry for the inconvenience

Fine, I will reiterate, Rybka 4 is so weak compared to some free private and even open source engines because in its progress there were used unqualified testers with no elementary knowledge in statistics, who do not know what LOS is, how the error margins are calculated, and what essential GUI's are there for testing purposes. They are disqualified as testers and as propagandists for a commercial engine like Rybka 4. In fact, I would say even more, most, if not all rating lists presented here, are biased towards commercial engines, being part of the fuss showing how strong they are. They are weak and period.

For everyone, if you want the best engines out there, the most stable, the best for analysis, best bitbases, go only for free engines, they are at least one year ahead of the best commercial engines.

That's fine now?

Kai

What is your statistical basis (as you keep touting) for your comment on the free engines being a year ahead of the best commercial engines?

That, for example:
I picked randomly one of the last IvanHoes and put it against Rybka 4 at 1''+0.1'' time control.

Code: Select all

 1.  IvanHoe B50hCx32         	6402.5/12170	4546-3911-3713
 2.  Rybka 4w32               	5767.5/12170	3911-4546-3713

No losses on time or some other reasans than mate.
LOS = 99.9999999998% for IvanHoe against Rybka 4, there is no question which is stronger.

ELO statistics:

Code: Select all

    Program                            Score        %    Av.Op.  Elo    +   -    Draws

  1 IvanHoe B50hCx32               : 6402.5/12170  52.6   3191   3209    5   5   30.5 %
  2 Rybka 4w32                     : 5767.5/12170  47.4   3209   3191    5   5   30.5 %

The random picked IvanHoe is 18 +/- 5 Elo points stronger than Rybka 4 with 95% confidence, 18 +/- 2.5 Elo points stronger 68% confidence.

And that without Triple Bases, which Rybka does not have, which bring additional 4 +/-3 Elo points for IvanHoe 95% confidence.

All in all, this IvanHoe is ~20 +/-5 Elo points stronger than Rybka 4 with 95% confidence, it has a "Game" mode and an "Analysis" mode, uses endgame bitbases, and is clearly ahead of Rybka 4 by at least a year if we are to judge the progress of Rybka. It is much more useful in Game mode (Playchess for example) and Analysis mode, because it doesn't have infinite time analysis bug as Rybka 4 has (stuck at some depth).

Do you want more? I can continue the test to bring error margins even smaller. Nowhere here such tests were conducted, and probably some folks here (paid or not) will flock to argue that Rybka is better at some outlandish time controls, with all its bugs!

For everyone here, latest free Ippo* engines are the best engines out there for game and analysis, at least a year ahead of the best commercial engines. Period.

Kai

[Laskos]

Dreams are free also.

I really can imagine that you are paid or something, although I don't really believe it. Tester, give me the LOS in a match +300 =500 -270, if you are a tester without Kiril. You use some CPU time on chess, that is all.

Kai

[M ANSARI]

Let me see ... you base your assumptions on a 1" + .1s games ... in 32bit no less ... and you call yourself an expert ???. You have the audacity to make fun of the entire Rybka beta team, and many of the well respected testers, of whom some are best minds in computer chess, and yet you think you are not making a fool of yourself by making your assumptions based on such games. Do you have the file with all those games? I would certainly like to look at them. Have you looked at any of the games? Do you know how long it takes a process based engine to synchronize engine startup compared to a threaded engine ... do you also know how critical for a process based engine it is to make sure that you have your hardware setup correctly. Have you checked to see if your data might somehow be corrupted by Rybka 4 losing many games in the endgame simply because it was running out of time? I certainly saw that a lot in my test matches in super fast time controls, in many cases R4 will lose a totally won endgame simply because it takes too long to initialize and synchronize its processes. Also again, why on earth are you running 32 bit ?? Instead of making a blanket statement that engine X is stronger than engine Y because I know how to calculate LOS .... you should maybe say that engine X vs. engine Y scores Z at a 1" +.1s time control and on a 32bit platform.

Again before trashing an entire group of testers, get your facts straight, get your testing protocol setup right, understand that you are ONLY testing a certain set of control parameters, and that the results of your data is only accurate in that very narrow field of parameters, and that the result is not a blanket equivalent of strength at all time controls.

[Uri Blass]

The Mod team has removed several posts in this thread due to personal attacks
if you are missing a post we are sorry for the inconvenience

Fine, I will reiterate, Rybka 4 is so weak compared to some free private and even open source engines because in its progress there were used unqualified testers with no elementary knowledge in statistics, who do not know what LOS is, how the error margins are calculated, and what essential GUI's are there for testing purposes. They are disqualified as testers and as propagandists for a commercial engine like Rybka 4. In fact, I would say even more, most, if not all rating lists presented here, are biased towards commercial engines, being part of the fuss showing how strong they are. They are weak and period.

For everyone, if you want the best engines out there, the most stable, the best for analysis, best bitbases, go only for free engines, they are at least one year ahead of the best commercial engines.

That's fine now?

Kai

What is your statistical basis (as you keep touting) for your comment on the free engines being a year ahead of the best commercial engines?

That, for example:
I picked randomly one of the last IvanHoes and put it against Rybka 4 at 1''+0.1'' time control.

Code: Select all

 1.  IvanHoe B50hCx32         	6402.5/12170	4546-3911-3713
 2.  Rybka 4w32               	5767.5/12170	3911-4546-3713

No losses on time or some other reasans than mate.
LOS = 99.9999999998% for IvanHoe against Rybka 4, there is no question which is stronger.

ELO statistics:

Code: Select all

    Program                            Score        %    Av.Op.  Elo    +   -    Draws

  1 IvanHoe B50hCx32               : 6402.5/12170  52.6   3191   3209    5   5   30.5 %
  2 Rybka 4w32                     : 5767.5/12170  47.4   3209   3191    5   5   30.5 %

The random picked IvanHoe is 18 +/- 5 Elo points stronger than Rybka 4 with 95% confidence, 18 +/- 2.5 Elo points stronger 68% confidence.

And that without Triple Bases, which Rybka does not have, which bring additional 4 +/-3 Elo points for IvanHoe 95% confidence.

All in all, this IvanHoe is ~20 +/-5 Elo points stronger than Rybka 4 with 95% confidence, it has a "Game" mode and an "Analysis" mode, uses endgame bitbases, and is clearly ahead of Rybka 4 by at least a year if we are to judge the progress of Rybka. It is much more useful in Game mode (Playchess for example) and Analysis mode, because it doesn't have infinite time analysis bug as Rybka 4 has (stuck at some depth).

Do you want more? I can continue the test to bring error margins even smaller. Nowhere here such tests were conducted, and probably some folks here (paid or not) will flock to argue that Rybka is better at some outlandish time controls, with all its bugs!

For everyone here, latest free Ippo* engines are the best engines out there for game and analysis, at least a year ahead of the best commercial engines. Period.

Kai

This is simply not correct.

20 elo is not a year and we usually get more than 20 elo per year in software(if you compare rybka with the best software of 2005 you get clearly more than 100 elo) and Rybka clearly earns more speed from 64 bits relative to IvanHoe.

The reason is that Vasik did not care to optimize rybka for 32 bits so the gap between 32 bits and 64 bits is bigger for rybka.

I believe that
Vas can make Rybka 32 bits 20-30% faster but he does not care about it because most customers who use 32 bits do not care about rating.