|
< Earlier Kibitzing · PAGE 35 OF 91 ·
Later Kibitzing> |
| Aug-28-08 | | zanshin: <kutz> I will be doing some baseline runs. For now, analysis is locked on position 31 (another nice feature) in the GMT game, our move. Preserve analysis and persistent hash are checked; persistent hash size is 1024; hash size is 1034. After 30 minutes on a quad core (while still doing other tasks):
click for larger view [+0.50] d=20 31.b4 Nxb4 32.Bxe4 fxe4 33.Nd7 Rb7 34.Nc5 Ree7 35.Nxb7 Rxb7 36.g4 Rb6 37.Kg2 Kf6 38.Kg3 Ke7 39.h4 Nd3 40.Rb1 Nb2 41.Rdc1 Kd6 42.f5 gxf5 43.gxf5 Rb8 44.Kh3 Rb3 45.Kg4 Rb4 46.Kg5 (0:29.14) [+0.50] d=19 31.Bxe4 fxe4 32.Rd2 Nb8 33.Rc5 Ra6 34.Rdc2 Kf6 35.Kg2 Rxe5 36.fxe5 Kxe5 37.R5c3 Rb6 38.b3 Kd6 39.Rf2 Rb7 40.Rf6 Ke7 41.Rf1 (0:20.28) [+0.20] d=19 31.Nd7 Rxb2 32.Bxa6 Nf2 33.Ne5 Nxd1 34.Rxd1 Rb4 35.Bc8 Rd6 36.Bd7 Rxa4 37.Bxc6 Ra3 38.Bb5 Rb3 39.Bd7 Kf6 40.Bc6 Ke6 41.Ra1 Rxe3 42.Bb5 Rb6 43.Rxa5 Rc3 44.Bf1 (0:23.09) [+0.00] d=19 31.Rc2 Nb4 32.Bxe4 Nxc2 33.Bxc2 Rxb2 34.Bb1 Rb4 35.Bc2 Rb2 36.Bb1 Rb4 37.Bc2 Rb2 38.Bb1 Rb4 39.Bc2 Rb2 40.Bb1 Rb4 41.Bc2 Rb2 42.Bb1 Rb4 43.Bc2 Rb2 44.Bb1 Rb4 45.Bc2 Rb2 46.Bb1 (0:23.38) [-0.15] d=19 31.g4 Nac5 32.gxf5 gxf5 33.Bxe4 Nxe4 34.Rc2 Rd6 35.Kh1 c5 36.Rg1 Kf6 37.Rg8 Rb7 38.Rf8 (0:23.55) [-0.19] d=19 31.Bxa6 Rxa6 32.Rc2 Rd6 33.Nd3 Rb6 34.Kf1 h6 35.Kg2 Kf6 36.Kf3 g5 37.Rdc1 g4 38.Ke2 Ke7 39.Rb1 Re6 40.Rbc1 h5 41.Ne5 (0:24.25) [-0.19] d=19 31.Bc2 Nac5 32.Bxe4 Nxe4 33.Rc2 Rd6 34.Nd3 h6 35.Kg2 Kf6 36.Kf3 g5 37.Rdc1 g4 38.Ke2 Ke7 39.Rb1 Re6 40.Rbc1 h5 41.Ne5 (0:25.01) [-0.19] d=19 31.Bb1 Nac5 32.Bxe4 Nxe4 33.Rc2 Rd6 34.Nd3 h6 35.Kg2 Kf6 36.Kf3 g5 37.Rdc1 g4 38.Ke2 Ke7 39.Rb1 Re6 40.Rbc1 h5 41.Ne5 (0:25.01) [-0.19] d=19 31.Kf1 Nac5 32.Bxe4 Nxe4 33.Rc2 Rd6 34.Nd3 h6 35.Kg2 Kf6 36.Kf3 g5 37.Rdc1 g4 38.Ke2 Ke7 39.Rb1 Re6 40.Rbc1 h5 41.Ne5 (0:26.21) [-0.20] d=19 31.Be2 Nac5 32.Rc2 h5 33.Ra1 Nb3 34.Ra3 Ned2 35.Kf2 Nb1 36.Ra2 Rd6 37.Nf3 c5 38.Bf1 Rb7 39.Bg2 Rb8 40.Kf1 Rbb6 (0:25.39) |
|
| Aug-28-08 | | DukeAlba: <kutz>
I have a question. What do temporary forum assistants do? |
|
Aug-28-08
 | | kutztown46: <DukeAlba>
Good question. You would assist the forum host most likely in one of two ways: 1) clerical function - copying relevant posts from previous forums into the current forum to provide continuity. Also copy relevant posts from main page to your forum when author of post forgets to do so. 2) chess function - if the move or line assigned to your forum is not being analyzed by anyone else, someone should perform minimal analysis, perhaps by using an engine and sliding forward several moves and posting the engine's assessment of the line at that point. It would be between you and the forum host exactly how the duties are divided. |
|
| Aug-28-08 | | zanshin: Same parameters, analysis after 60 minutes:
[+0.50] d=21 31.b4 Nxb4 32.Bxe4 fxe4 33.Nd7 Rb7 34.Nc5 Ree7 35.Nxb7 Rxb7 36.g4 Rb6 37.Kg2 Kf6 38.Kg3 Ke7 39.h4 Nd3 40.Rb1 Nb2 41.Rdc1 Kd6 42.f5 gxf5 43.gxf5 Rb8 44.Kh3 Rb3 45.Kg4 Rb4 46.Kg5 (0:45.34) [+0.50] d=21 31.Bxe4 fxe4 32.Rd2 Nb8 33.Rc5 Ra6 34.Rdc2 Kf6 35.Kg2 Rxe5 36.fxe5 Kxe5 37.R5c3 Rb6 38.b3 Kd6 39.Rf2 Ke7 40.Rc5 Rxb3 41.Re2 h5 42.Rxa5 Ke6 (0:49.45) [+0.26] d=21 31.Nd7 Rxb2 32.Bxa6 Nf2 33.Ne5 Nxd1 34.Rxd1 Rb4 35.Ra1 Kf6 36.Kf2 Rb2 37.Be2 Rxe5 38.fxe5 Kxe5 39.Ke1 c5 40.Rc1 c4 41.Bf3 Rb3 42.Kd2 Ke6 43.Rc2 Ra3 44.Ke2 (0:56.31) [+0.00] d=20 31.Rc2 Nb4 32.Bxe4 Nxc2 33.Bxc2 Rxb2 34.Bb1 Rb4 35.Bc2 Rb2 36.Bb1 Rb4 37.Bc2 Rb2 38.Bb1 Rb4 39.Bc2 Rb2 40.Bb1 Rb4 41.Bc2 Rb2 42.Bb1 Rb4 43.Bc2 Rb2 44.Bb1 Rb4 45.Bc2 Rb2 46.Bb1 (0:33.15) [+0.00] d=20 31.g4 Nac5 32.gxf5 gxf5 33.Bxe4 Nxe4 34.Rc2 Rd6 35.Kh1 c5 36.Rg1 Kf6 37.Rg8 Rb7 38.Rf8 Ke6 39.Re8 Re7 40.Ra8 Rb6 41.Rxa5 Reb7 42.Ra8 Rxb2 43.Rxb2 Rxb2 44.a5 Ra2 45.a6 Ra1 46.Kg2 (0:37.18) [-0.19] d=20 31.Bxa6 Rxa6 32.Rc2 Rd6 33.Nd3 Rb6 34.Kf1 h6 35.Kg2 Kf6 36.Kf3 g5 37.Rdc1 g4 38.Ke2 Ke7 39.Rb1 Re6 40.Rbc1 h5 41.Ne5 (0:38.13) [-0.20] d=20 31.Be2 Nac5 32.Rc2 h5 33.Ra1 Nb3 34.Ra3 Ned2 35.Kf2 Nb1 36.Ra2 Rd6 37.Nf3 c5 38.Bf1 Rb7 39.Bg2 Rb8 40.Kf1 Rbb6 (0:42.45) [-0.24] d=20 31.Bc2 Nac5 32.Bxe4 Nxe4 33.Rc2 Rd6 34.Nd3 h6 35.Kg2 Kf6 36.Kf3 g5 37.Rdc1 g4 38.Ke2 Ke7 39.Rb1 Re6 40.Rbc1 h5 41.Rd1 Rd6 (0:40.19) [-0.24] d=20 31.Bb1 Nac5 32.Bxe4 Nxe4 33.Rc2 Rd6 34.Nd3 h6 35.Kg2 Kf6 36.Kf3 g5 37.Rdc1 g4 38.Ke2 Ke7 39.Rb1 Re6 40.Rbc1 h5 41.Rd1 Rd6 (0:41.26) [-0.24] d=20 31.Kf1 Nac5 32.Bxe4 Nxe4 33.Rc2 Rd6 34.Nd3 h6 35.Kg2 Kf6 36.Kf3 g5 37.Rdc1 g4 38.Ke2 Ke7 39.Rb1 Re6 40.Rbc1 h5 41.Rd1 Rd6 (0:42.45) |
|
Aug-28-08
| | kutztown46: <z>
Well, if it takes an hour to get to 21-ply on move 31, it will take much longer than that at move 15 or 20! Can you run the product with a previous version of Rybka, and if so, will you get deeper faster? BTW, I'm sure by now you have figured out the meaning of the times in the parentheses. |
|
| Aug-28-08 | | zanshin: <Can you run the product with a previous version of Rybka, and if so, will you get deeper faster?> <kutz> I'm pretty sure I can get deeper plies faster with Rybka 2.3 mp. However, Vas cautions not to try to compare ply-levels between engines. He maintains that R3 analysis is better quality. I did think about just using the R3 engine with Aquarium, but now feel that it would be a waste. There is something a learning curve for Aquarium, but I'm confident it's worth it. |
|
| Aug-28-08 | | zanshin: <kutz> I tried another experiment running the Rybka 3 engine with the Fritz GUI. This was a failure. R3 still ran at the same rate and the lines were all truncated. I did not even get the number of PV lines I requested. Also, you don't get the amount of feedback between R3 and Fritz as when you use the GUIs with their native engines. |
|
| Aug-30-08 | | zanshin: <kutz> I'm fairly confident I know how to use Aquarium in the sliding forward method we have developed. However, the use of Interactive Deep Analysis (IDeA) is intriguing. IDeA produces the analysis trees we covet, so I thought that deep runs in IDeA mode would make forum assignments so much easier. And they still might. The problem is that the output produced is difficult to post on the forums. It is a color-coded tree with line evaluations for each branch or sub-branch. It's not easy for me to interpret, though I suspect I'll get better with time. Another limitation of IDeA is that ply depths are not given. They call this a bug in the Rybka forums - I think it is more of a TODO. In respones to a question I posted on the Rybka forums, you'll want to remove the color-coding when you copy the lines. Go to the Display options, and in PGN copy options, select the 2nd button. To give you an idea of sample IDeA output, here's a quick 2 minute run (dual core) on today's chess puzzle: 
click for larger view1... Rxf3 2. Qxf3 (2. Rxh4 Qxh4 (2... Rg3+ 3. Kxg3 -9.00 (DA))3. Kxf3 Qh6 -9.00 (DA)) 2... Qd2+ 3. Kg1 (3. Kf1 Qxb2 -9.00 (DA)) 3... Nd4 4. Bxd4 Bf2+ -9.00 (DA) * And here's similar 2-minute output using conventional infinite analysis we using in sliding forward. [-4.60] d=9 1...Rxf3 (0:00.00)
[-0.52] d=9 1...b4 2.Rh2 Rhf7 3.Rch1 Rxf3 4.Rxh4 Rf2 5.Qxf2 Rxf2 6.Kxf2 Qd2 7.Kg3 Qxb2 8.axb4 Nxb4 9.g5 Kf8 10.Rh7 (0:00.00) [-0.22] d=9 1...Rf8 2.Rxc5 Rxf3 3.Qxf3 Qd2 4.Kf1 Qxb2 5.Rxc6 Qxb1 6.Ke2 Qb2 7.Kd1 Qxb3 8.Ke2 Qb2 9.Kd1 Rf7 10.Rc8 Kg7 11.Qe2 (0:00.00) [-0.18] d=9 1...Rf6 2.Rc2 Rff7 3.Ba2 Qf4 4.Rh3 Nd4 (0:00.00) [-0.15] d=9 1...Nd4 2.Bxd4 cxd4 3.Rc8 Rf8 4.Qc2 Rhf7 5.Rxf8 Rxf8 6.Qe2 Kg7 7.b4 (0:00.00) [-0.15] d=9 1...Rff7 2.Rc2 Nd4 3.Bxd4 cxd4 4.Rc8 Rf8 5.Qc2 Rhf7 6.Rxf8 Rxf8 7.Qe2 Kg7 8.b4 (0:00.00) [-0.14] d=9 1...Qg5 2.Rxc5 Rxf3 3.Qxf3 Qd2 4.Kf1 Qxb2 5.Rxc6 Qxb1 6.Ke2 Qb2 7.Kd1 Qxb3 8.Ke2 Qb2 9.Kd1 Rf7 10.Rc8 Kg7 11.Qe2 (0:00.00) Both solved the puzzle - no surprise, but the IDeA output is harder to interpret. |
|
| Aug-31-08 | | zanshin: <kutz> I've been running some baseline tests to try to reassure you about Rybka 3. I used the following position from one of Wesley So's games as a test position with Rybka 2.3 mp, Deep Rybka 3 and Fruit 2.2. Obviously, they all found the winning move <Rxf5!>. In one minute, Fruit got to 14 plies, Rybka 2.3 to 17 and Rybka 3 only got to 10. This confirms what we already know, Rybka 3 moves down ply levels more slowly than the other engines. 
click for larger viewBut the question is which produces better analysis? I used engine matches as a measure (10 games, 5 minutes each per game). Rybka 3 beat Rybka 2.3 by 5.4 to 4.5. As expected, Rybka 3 beat Fruit quite easily 8.5 to 1.5. Despite your reservations, I think it would be a good idea to use Rybka 3 predominantly in the game. You may want to do engine tests between Rybka 3 and Fritz 10, but it might be unfair unless you have the mp version of Fritz. |
|
| Sep-01-08 | | zanshin: <kutz> Since we have settled on the engines for forum prep, this is just an fyi. I ran an engine match between Fritz 10 and Rybka 3 using the standard Fritz blitz set up. The result: <Rybka 3 won 7 matches, drew 2 and lost 1>. Fritz 10 would be reporting node rates at over 2000 kN/s, compared to Rybka which rarely went over 100. The results of this unscientific study suggest that the quality of Rybka analysis is better, despite the low node rate and ply levels reported. If you want to try this yourself, make sure your tablebase path is cleared or you will not be able to load the Rybka 3 engine - again the well know bug that chessbase either cannot fix or refuses to fix. The other consideration is that I am using Deep Rybka vs the single-core version of Fritz. The bottom line for me is that I am comfortable using Deep Rybka 3 for forum preparation. |
|
Sep-02-08
| | kutztown46: <Pairings of Temporary Forum Hosts and Assistants> At this point, there have been no additional volunteers for Forum Assistants. I anticipate that we will need to kick off the forums sometime in the next few weeks, so now is a good time to pair up the forum hosts and assistants. Since there are only four assistants, I am giving the four new forum hosts one assistant each. The four experienced forum hosts will not have an assistant, unless we get more volunteers. The forum hosts will be:
<dalbertz> (no assistant) <Red October> (no assistant) <Karpova> (no assistant) <mckmck> (*) (no assistant) <achieve> assisted by <hms123> <ajile> assisted by <truefriends> <tweet77> assisted by <Hootey> <classF> assisted by <crawfb5> * <mckmck> has not been heard from in three weeks, so I am not certain of him hosting a forum. Does anyone know of a way to contact him to find out what's going on? There have been isolated suggestions that we should kick off the temporary forums. My preference is to hold off a little longer, but when there is an obvious consensus among the team, then I will assign forums. |
|
| Sep-02-08 | | capafan: <kutztown>Should <mckmck> for any reason not be able to host his forum, I would be willing to host a forum in his absence. Leave me a message on my forum if you need my assistance. You, as always, continue to perform a very valuable service to the team, and I appreciate your efforts. |
|
Sep-02-08
| | kutztown46: <DaringSpeculator: Okay, I'll volunteer.> Thank you! You can assist <Red October>. |
|
| Sep-02-08 | | ajile: Hi kutztown46
If possible I would like to be assigned variations in my forum that are the most interesting to me. Of course these lines would have to be one of the main lines we all agree are viable for Black. So for example I currently like the sequence 1.e4 c5 2.Nc3 Nc6 3.Nf3 e5 which is a known and viable line for Black. If possible I would like this line assigned to my forum. If it doesn't work out for this to happen that's fine too. |
|
Sep-02-08
| | kutztown46: <ajile>
Thanks for stopping by. Let me first say that I prefer to wait a while longer before I begin assigning the temporary forums. I would like to see us get closer to the end of the opening book. Probably, at some earlier point, a consensus will emerge that the forums need to be started. At that point, I will yield to the will of the team. But I don't want to hide behind that. I need to make sure that you understand what is coming your way as a forum host. It will not be possible for me to assign you the most interesting variation with any degree of regularity. I try to ensure that every forum host gets their chance to host discussion of the perceived main line. You are one of eight forum hosts. At least 25% of the time you will not have anything assigned to you (my pledge is to never have more than six forums at a time) and even when you have a move or line assigned to you, the majority of the time it will not be the move or line which is generating the most discussion. And once in a while (hopefully not too often) you will be assigned to host a move or line which will generate no interest at all. I try to make sure that every playable move has a forum, but that does not mean that the team will give strong consideration to every playable move. I'm being completely honest with you. I was happy when you volunteered to host a forum and I hope you will stick with it. However, if you prefer to be one of the "roving analysts" instead of hosting a forum, I will not be upset with you. I would not want you to be unhappy in your role. |
|
| Sep-02-08 | | twinlark: <kutztown46>
Way to go! I hope you're getting enough sleep! |
|
Sep-02-08
| | kutztown46: <twinlark>
We have not even gotten to the fun part yet! Just preparation so far. |
|
| Sep-02-08 | | twinlark: It just looks like you're already feverishly busy!
Good work on bringing in so many helpers. In fact, to give credit where it's due - amazing work. Have you thought of advertising for forum assistants in The Kibitzer's Café ? |
|
Sep-02-08
| | kutztown46: <twinlark>
Good idea! I never go there, so I never considered it. I'll get on that. Thanks for the idea. |
|
| Sep-03-08 | | ajile: It's no biggie. Once we reach past the next few moves it won't matter what I get assigned. |
|
| Sep-04-08 | | tweet77: Hi <kutzdown46>,
A quick suggestion if we get to forum assignments on the early side, it might be a good idea to formally assign <spare> hosts and assistants to active fora. Hopefully this will ensure best use of our limited supply of forum assistants. Just a thought. |
|
Sep-04-08
| | kutztown46: <tweet77>
Thanks for the suggestion. I would rather leave that voluntary. It could be a long game, and I want the forum hosts to have periodic breaks. But if they want to pitch in at another forum while they are "on break", I won't try to stop them! :) |
|
| Sep-04-08 | | kieranchez: Hi <kutzdown46>
I volunteer to be an assistant at least till my membership runs out, if needed.
Post reply on my forum.
Kieranchez |
|
| Sep-05-08 | | zanshin: <kutz> My Rybka 3 will not be very useful at this point because I did not get the Rybka opening book sold separately by Chessbase. I understand it has good coverage of the Sicilian, but I figured that there are more than enough databases here, including yours. I will attempt an experiment over the weekend using persistent hash tables to see if we can speed up ply searches. |
|
Sep-05-08
| | kutztown46: <z>
I may have some extra time tomorrow if Hanna wipes out my plans. If I do, I'll fire up Aquarium again and try that application priority tip you sent me. BTW, I can't print that article you sent me a link to. When I try, the browser becomes unresponsive. |
|
 |
|
|
< Earlier Kibitzing · PAGE 35 OF 91 ·
Later Kibitzing> |
Kibitzing
A majority vote for 1-0 means we resign. 