where can I find robbolito-rybka at long time control

[Uri Blass]

I remember that somebody posted results at long time control without mentioning the clone name in the rybka forum but it seems that the thread is deleted from the rybka forum.

I remember that rybka was leading the last time that I saw the thread

Uri

[beachknight]

I remember that somebody posted results at long time control without mentioning the clone name in the rybka forum but it seems that the thread is deleted from the rybka forum.

I remember that rybka was leading the last time that I saw the thread

Uri

Hi Uri,

IMHO, best thing you can do is sending a pm to Rybka forum mod.

Best,

PS: I'd like to use this opportunity thanking our mods for
their democratical stance on these issues.

[M ANSARI]

I think Paul (NATIONAL) had some LTC games but stopped the match due to MP version coming out. I did some 60 or 70 games at relatively LTC using 4 Ghz processors. If you are interested PM me and I can send them to you. I had planned to make more games but I got the information that I needed. Basically that RL is simply a slightly faster R3 personality and scores are linearily dependent on time controls. I might be wrong but that is the sense I get from my data.

[beachknight]

Hi,

Like Uri, I'd be also interested in them. May I send you a pm?

Best,

[Uri Blass]

I am only interested in results and not in the games
I hope that rybka can win.
I do not like engine with unknown authors to beat rybka at long time control and I hope that stockfish can beat rybka so we can forget about robbo or iggorit.

I wonder how much time stockfish needs to get better place than rybka3 in CCRL or CEGT.

Uri

[Dann Corbit]

I am only interested in results and not in the games
I hope that rybka can win.
I do not like engine with unknown authors to beat rybka at long time control and I hope that stockfish can beat rybka so we can forget about robbo or iggorit.

I wonder how much time stockfish needs to get better place than rybka3 in CCRL or CEGT.

Uri

Unfortunately, the 40/120 testing seems to have hit a wall, so 40/40 is the slowest regular testing see see today.

Last test was Feb 2009:
http://www.husvankempen.de/nunn/rating120.htm

[NATIONAL12]

I remember that somebody posted results at long time control without mentioning the clone name in the rybka forum but it seems that the thread is deleted from the rybka forum.

I remember that rybka was leading the last time that I saw the thread

Uri

it was me that posted 40/40 match results on Rybka forum

HS Masterbook 10 moves,win 64,1024 hash,ponder off,contempt 15,turbo3 tb's,i7 860 quad.

Rybka3(1cpu)-RobboLito 085g3(1cpu)

+12=29-5

[M ANSARI]

I did testing at 16_1 and 30_1, but these were done on fast hardware of 4Ghz and ponder ON. I am having trouble finding the 30_1 games as I screwed up all my CB software when I was trying to install 4 different CB GUI's so I could do multiple single core matches on my Octa's. It would have been an easy way to get a large collection of LTC games (60_1) very quickly and get a definitive answer. I would have been able to get maybe 200 LTC per day. Unfortunately it did not work and installing more than 1 instance of CB GUI corrupts all the programs and destroys the graphics and files.

Basically RLg3 is around 70 ELO stronger at very fast time controls (1_1)and this is reduced when you put contempt to 0 on R3 to around 50 ELO. The advantage of RL drops dramatically as you go up in time and is only around 20 ELO at 16_1 and actually became negative at 30_1 (all contempt 0). There seems to be a linear correlation between ELO difference and time. It is obvious that this is a R3 clone that somehow is running the executable faster. There are some improvements in that the well documented bugs in R3 are fixed ... but it has the same weaknesses as R3 ... and of course the same strengths. The best I can describe it is that it is another personality of R3. It has way more similarity to R3 default than Rybka Human or Rybka Dynamic have to R3 Default.

[NATIONAL12]

on latest Igorrit 086v6(8cpu) at 4+2 i was getting a lot of time out games 20% until it locked up altogether.taking those games out after 92 games R3 was +57 ELO.

[beachknight]

I did testing at 16_1 and 30_1, but these were done on fast hardware of 4Ghz and ponder ON. I am having trouble finding the 30_1 games as I screwed up all my CB software when I was trying to install 4 different CB GUI's so I could do multiple single core matches on my Octa's. It would have been an easy way to get a large collection of LTC games (60_1) very quickly and get a definitive answer. I would have been able to get maybe 200 LTC per day. Unfortunately it did not work and installing more than 1 instance of CB GUI corrupts all the programs and destroys the graphics and files.

Basically RLg3 is around 70 ELO stronger at very fast time controls (1_1)and this is reduced when you put contempt to 0 on R3 to around 50 ELO. The advantage of RL drops dramatically as you go up in time and is only around 20 ELO at 16_1 and actually became negative at 30_1 (all contempt 0). There seems to be a linear correlation between ELO difference and time. It is obvious that this is a R3 clone that somehow is running the executable faster. There are some improvements in that the well documented bugs in R3 are fixed ... but it has the same weaknesses as R3 ... and of course the same strengths. The best I can describe it is that it is another personality of R3. It has way more similarity to R3 default than Rybka Human or Rybka Dynamic have to R3 Default.

Thanks very much.

Best,