|
< Earlier Kibitzing · PAGE 65 OF 65 ·
Later Kibitzing > |
Dec-18-09
 |
| pulsar: <zanshin> I share <whatthefat>'s opinion. :) |
 |
| Dec-20-09 |
| mrandersson: Guest player strikes again.time control was 4 mins plus 3 seconds in this game. Like the last guest i played this guy/lady was from germany. The german chess clubs must be really strong. [Event "rated untimed match"]
[Site "?"]
[Date "????.??.??"]
[Round "?"]
[White "guest"]
[Black "owensdefence"]
[Result "*"]
[WhiteElo "0"]
[BlackElo "1548"]
[ECO "A40"]
[TimeControl "4"]
1. d4 e6 2. c4 f5 3. Nf3 Nf6 4. g3 b6 5. Bg2 Bb7 6. O-O Bb4 7. Qb3 a5 8. a3
Bd6 9. Nc3 Be7 10. Rd1 Ne4 11. Bf4 Nxc3 12. Qxc3 O-O 13. Qe3 Be4 14. d5 Bf6
15. dxe6 Bxb2 16. Ng5 Bxa1 17. Bxe4 Nc6 18. Bxc6 Bf6 19. exd7 Rb8 20. Ne6 Qe7
21. Bxc7 *
This pgn i hope works now |
|
Dec-29-09
 |
| mysql: How is Rybka's rating calculated? Is it a different rating category from humans? |
|
Jan-02-10
 |
| zanshin: <mysql: How is Rybka's rating calculated? Is it a different rating category from humans?> <mysql> Ratings for engines are evaluated against other engines. Otherwise, the methods are similar as for humans. |
|
Jan-02-10
|
| mysql: Thanks. Have they ever tried rating it against humans? |
|
Jan-06-10
|
| zanshin: <mysql: Thanks. Have they ever tried rating it against humans?> <mysql> Not really, because there have not been enough games between GM and Rybka under tournament conditions. The only way you can get an ELO rating for an engine that would be on the same scale as that for humans would be to have the engine(s) compete regularly in top level GM tournaments. In the meantime, you can look at engine ratings based on contests against other engines: http://www.computerchess.org.uk/ccr... or you can estimate based on limited contests between engines and human GMs: http://www.rybkachess.com/index.php... |
|
Jan-06-10
|
| mysql: <zanshin: <mysql: Thanks. Have they ever tried rating it against humans?> <mysql> Not really, because there have not been enough games between GM and Rybka under tournament conditions. The only way you can get an ELO rating for an engine that would be on the same scale as that for humans would be to have the engine(s) compete regularly in top level GM tournaments. In the meantime, you can look at engine ratings based on contests against other engines: http://www.computerchess.org.uk/ccr... or you can estimate based on limited contests between engines and human GMs: http://www.rybkachess.com/index.php...; Thanks again. By the way, I noticed that Filipino-made engine "TwistedLogic" is number 25 on the list. We need to submit some games of that engine so that it will have a page here in chessgames. |
|
Jan-07-10
|
| zanshin: <mysql> You probably know that <Edsel Apostol> has joined <CG> as User: twistedlogic I wanted to ask him some questions about engine programming, but he hasn't posted in over a month. |
|
Jan-21-10
 |
| notyetagm: http://talkchess.com/forum/viewtopi... PostPost subject: <Re: Firebird 1.0 beta vs Rybka 3, 15+10>
Posted: Sat Jan 16, 2010 10:55 pm
Reply to topic Reply with quote
dear John,
i ve just finished my first test with firebird 1.0. On my Compaq 6710b, T8100 , 2100 Mhz a Blitz 4m+2s Nunn2 testmatch. Both 2 CPU 64mb Hash. An IMO almost unbelievable good result. The best in the Ip/Rob/Ig series sofar. 1 FireBird 1.0 beta w32 +20/=22/-8 <62.00%> 31.0/50 2 Rybka 3 32-bit +8/=22/-20 <38.00%> 19.0/50 Grts Bram |
|
Jan-21-10
|
| mysql: <zanshin: <mysql> You probably know that <Edsel Apostol> has joined <CG> as User: twistedlogic I wanted to ask him some questions about engine programming, but he hasn't posted in over a month.> I want to try engine programming but I'm having a headache just even thinking of reviewing ab pruning and other algorithms/data structures. Any "easy" resources on this (as well as documentation how to interface with UCI)? |
|
Jan-23-10
|
| notyetagm: <notyetagm: http://talkchess.com/forum/viewtopi...
PostPost subject: <Re: Firebird 1.0 beta vs Rybka 3, 15+10> Posted: Sat Jan 16, 2010 10:55 pm Reply to topic Reply with quote
dear John,
i ve just finished my first test with firebird 1.0. On my Compaq 6710b, T8100 , 2100 Mhz a Blitz 4m+2s Nunn2 testmatch. Both 2 CPU 64mb Hash. An IMO almost unbelievable good result. The best in the Ip/Rob/Ig series sofar. 1 FireBird 1.0 beta w32 +20/=22/-8 <62.00%> 31.0/50 2 Rybka 3 32-bit +8/=22/-20 <38.00%> 19.0/50 Grts Bram>
Gee, no comment on my Firebird post? |
|
Jan-23-10
|
| notyetagm: http://talkchess.com/forum/viewtopi...
<Posted:
Fri Jan 22, 2010 6:20 pm
Post subject: Re: <<<Firebird holds his superiority at longer timecontrol>>> Reply to topic Reply with quote
i also see no reason why firebird shouldn't have just the same success against rybka 3 over longer time controls. <<<most of my testing so far has been in 4/2 games (when firebird is just destroying rybka)>>> but i have just started some 25min games and it is 2-1 to firebird with a big plus in the 4th game.> FIREBIRD -> http://www.chesslogik.com/firebird.... |
|
Jan-23-10
|
| notyetagm: <FIREBIRD> is really *awesome*. You <RYBKA> users should give it a try or you'll get owned by us <FIREBIRD> users. :-)
http://www.chesslogik.com/firebird.... |
|
| Jan-29-10 |
| Mate Hunter: <notyetagm>
Thanks for the info.
Open alternatives lead to tougher competition. |
|
| Jan-29-10 |
| zaxcvd: firebird wins at 0+30s (30 s per move or 40 moves/20 min)
22.5 /40 nunnsuite.. Rybka contempt=0
FIRE!!!! |
|
| Jan-29-10 |
| zaxcvd: btw 4 cpu pond off ... so its multicore! |
|
Jan-31-10
|
| notyetagm: <Mate Hunter: <notyetagm> Thanks for the info.
Open alternatives lead to tougher competition.> I do not believe that it is open, but I have paid for *both* Rybka 2 and Rybka 3 so I do not feel guilty using it. |
|
| Feb-04-10 |
| Wood Mover: Gents, a question out of left field.
I take it computing power will tend to dictate the time it takes for Rybka to evaluate a position to a certain ply. I gather the complexity of the chess position is another factor. So looking forward to 10 years time when CPUs will be more powerful (in line with Moore's law etc), how much deeper ply can we expect the evaluations to be if we permit the same processing time? Understand the gist of what I'm saying?
Thanks |
|
| Feb-04-10 |
| Wood Mover: Just expanding on this line of thought. I read on the rybka forum that for each additional ply of rybka analysis, the ELO goes up 75. But for Rybka to achieve that extra ply takes twice as much processing time. Going by Moore's law, CPU processing power doubles every 2 years (correct me if I'm wrong). So by 2010 you could make a crude prediction Rybka's ELO would go up by 375 from CPU processing power advancements alone. thoughts? |
|
Feb-04-10
|
| mysql: Yes, CPU processing power goes up, but nowadays it's due to additional 'cores'. They might have reached a plateau already when it comes to single processor speed. So, I think yes, Rybka's ELO should go up if it could be programmed to utilize all the cores effectively. If not, then it's strength would remain the same no matter how many cores are added. |
|
Feb-04-10
 |
| rogge: Google "winrar". |
|
Feb-07-10
|
| rogge: Classy. Ask for help, get help (Google "winrar"), and delete the question. You're welcome, dude. |
|
| Feb-08-10 |
| truefriends: @rogge: I left a post clarifying the situation in your profile... |
|
Feb-08-10
 |
| imag: <Wood Mover: Just expanding on this line of thought. I read on the rybka forum that for each additional ply of rybka analysis, the ELO goes up 75. But for Rybka to achieve that extra ply takes twice as much processing time.> I don't think that's accurate. For example, there is an enormous difference between analyzing at 1 ply and at 2 plies whereas being able to analyze 36 plies instead of 35 plies does not constitute significant improvement. So fewer and fewer ELO points are gained when achieving extra plies. Maybe now it is 75 points per ply, but for the next ply it will only be 50 points, and then only 30 points for the next and so on. <Going by Moore's law, CPU processing power doubles every 2 years (correct me if I'm wrong).> The number of transistors integrated in a chip doubles every 2 years, it does not necessarily mean that the processing power doubles. Fortunately, chess analyzing software can be quite efficiently parallelized. <So by 2010 you could make a crude prediction Rybka's ELO would go up by 375 from CPU processing power advancements alone. thoughts?> IMHO a 100 ELO improvement would already be a success. |
|
Feb-09-10
 |
| whatthefat: Copying across a post by <Wood Mover> from <RandomVisitor>'s forum: <I found this for lower ply on this link:http://rybkaforum.net/cgi-bin/rybka...
2plyVs1ply +60/ =35/-05 77.5%
3plyVs2ply +59/ =30/-07 77.1%
4plyVs3ply +51/ =47/-10 69.0%
5plyVs4ply +53/ =41/-10 70.7%
6plyVs5ply +53/ =45/-16 66.2%
7plyVs6ply +42/ =55/-09 65.6%
8plyVs7ply +46/ =50/-08 68.3%
9plyVs8ply +44/ =53/-09 66.5%
10plyVs9ply +77/=155/-16 62.29%>
This is an excellent demonstration of the diminishing returns given by extra ply depth. |
|
 |
|
< Earlier Kibitzing · PAGE 65 OF 65 ·
Later Kibitzing > |
