Not logged inRybka Chess Community Forum
Forum RybkaChess Help Search Register Login
Up Topic The Rybka Lounge / Correspondence Chess / Deleted Vs. Vytron (DPA Vs. IA) 1/2 - 1/2
1 2 3 4 5 6 Previous Next  
Parent - - By Banned for Life (Gold) Date 2009-07-08 18:58
I am rather clueless about what the best settings would be. I will use the following:

- New persistent hash (8 GB which is serious overkill) so I don't take advantage of past experience
- Rybka MultiPV threshold set to 25 cp to allow a larger branching factor in DPA
- Rybka read and write thresholds for PH set to 10 (need to verify this works)
- Rybka hash table size of 2GB
- Rybka TB settings TBD
- DPA total time of 1800 seconds
- DPA branching factor of 5, 4, 3, 2
- DPA depth of 6 plies (this doesn't include the root)

I will post the tree for each move for your viewing enjoyment.
Parent - - By Uly (Gold) [mx] Date 2009-07-08 20:08

> - DPA depth of 6 plies


Wow, I'd be much more impressed if you draw, I think one tactical shoot will be enough because DPA will waste too much time revising useless moves at low depth. And I don't think PH will be used (if depth 10 will not be reached, it's 4 plies short of writing) - besides the root.
Parent - - By Banned for Life (Gold) Date 2009-07-08 20:46
You are confusing DPA "Length of Variation" this with IDeA depth (with my help). I have no clue how deep Rybka will get on the positions. The CB GUI will decide how much time to spend on each position by pre-calculating the number of positions and splitting up the 1800 seconds accordingly. DPA length of variation defines how many plies are evaluated in the tree. In my case it will be seven (6+1). So if the tree is fully populated (i.e. there are always enough moves within 25 cp), The tree will look like this:

W0a, B01a, W02a, B03a, W04, B05, W06
W0a, B01a, W02a, B03b, W04, B05, W06
W0a, B01a, W02b, B03a, W04, B05, W06
W0a, B01a, W02b, B03b, W04, B05, W06
W0a, B01a, W02c, B03a, W04, B05, W06
W0a, B01a, W02c, B03b, W04, B05, W06
W0a, B01b, W02a, B03a, W04, B05, W06
W0a, B01b, W02a, B03b, W04, B05, W06
W0a, B01b, W02b, B03a, W04, B05, W06
W0a, B01b, W02b, B03b, W04, B05, W06
W0a, B01b, W02c, B03a, W04, B05, W06
W0a, B01b, W02c, B03b, W04, B05, W06
W0a, B01c, W02a, B03a, W04, B05, W06
W0a, B01c, W02a, B03b, W04, B05, W06
W0a, B01c, W02b, B03a, W04, B05, W06
W0a, B01c, W02b, B03b, W04, B05, W06
W0a, B01c, W02c, B03a, W04, B05, W06
W0a, B01c, W02c, B03b, W04, B05, W06
W0a, B01d, W02a, B03a, W04, B05, W06
W0a, B01d, W02a, B03b, W04, B05, W06
W0a, B01d, W02b, B03a, W04, B05, W06
W0a, B01d, W02b, B03b, W04, B05, W06
W0a, B01d, W02c, B03a, W04, B05, W06
W0a, B01d, W02c, B03b, W04, B05, W06

W0b, B01a, W02a, B03a, W04, B05, W06
W0b, B01a, W02a, B03b, W04, B05, W06
W0b, B01a, W02b, B03a, W04, B05, W06
W0b, B01a, W02b, B03b, W04, B05, W06
W0b, B01a, W02c, B03a, W04, B05, W06
W0b, B01a, W02c, B03b, W04, B05, W06
W0b, B01b, W02a, B03a, W04, B05, W06
W0b, B01b, W02a, B03b, W04, B05, W06
W0b, B01b, W02b, B03a, W04, B05, W06
W0b, B01b, W02b, B03b, W04, B05, W06
W0b, B01b, W02c, B03a, W04, B05, W06
W0b, B01b, W02c, B03b, W04, B05, W06
W0b, B01c, W02a, B03a, W04, B05, W06
W0b, B01c, W02a, B03b, W04, B05, W06
W0b, B01c, W02b, B03a, W04, B05, W06
W0b, B01c, W02b, B03b, W04, B05, W06
W0b, B01c, W02c, B03a, W04, B05, W06
W0b, B01c, W02c, B03b, W04, B05, W06
W0b, B01d, W02a, B03a, W04, B05, W06
W0b, B01d, W02a, B03b, W04, B05, W06
W0b, B01d, W02b, B03a, W04, B05, W06
W0b, B01d, W02b, B03b, W04, B05, W06
W0b, B01d, W02c, B03a, W04, B05, W06
W0b, B01d, W02c, B03b, W04, B05, W06

I'll leave out the other 3 root moves, but you can see that if I use branching of 5, 4, 3, and 2 and Length of Variation = 6, I end up with 360 positions to evaluate. 1,800 / 360 = 5 seconds per position.

If instead, I use the same branching factor but Length of Variation = 4, I end up with 120 positions to evaluate. In this case I end up with 1,800 / 120 = 15 seconds per position. I like 15 seconds per position more than 5, so I may go with LoV = 4.
Parent - By Uly (Gold) [mx] Date 2009-07-08 20:50

>You are confusing DPA "Length of Variation" this with IDeA depth (with my help).


Oops, I see.
Parent - - By titanium cranium (***) [us] Date 2009-07-08 21:38
We're finally getting somewhere the average user can understand. If you're an expert in DPA, but are clueless about the best settings, imagine how the rest of us feel. With the above settings in mind, what would they be if you were a correspondence player with days to analyze a position?
Parent - By Banned for Life (Gold) Date 2009-07-08 22:17
With days to analyze a position, I like to spend 5-10 minutes on each position (rather than the 15 seconds that will be allocated here). Other than that, the settings are very similar.
Parent - - By titanium cranium (***) [us] Date 2009-07-09 12:22
"Rybka MultiPV threshold set to 25 cp to allow a larger branching factor in DPA."

Hmm, I follow all of your computer-ease except for "MultiPV threshold set to 25 cp." There's nothing in R3 materials or Help section on that term. Is it the number of lines of R3 evaluation output?
Parent - - By Banned for Life (Gold) Date 2009-07-09 16:23
Rybka allows you to set MultiPVcp which gives the evaluation difference (in centipawns) between the best move and the worst move that will be considered. By using this feature, you can set the the number of PVs very high, and Rybka will only consider the reasonable moves (i.e. the ones that are reasonably close to the best move in eval).

As with all chess products, documentation is somewhere between poor and nonexistent.
Parent - - By titanium cranium (***) [us] Date 2009-07-09 17:46
I have spent many, many hours in the Engine Parameters window and there is nothing there (or in Help) that looks even remotely like "MultiPV_cp."  Therefore, I have assumed that is has some other name, correct?!
Parent - - By Banned for Life (Gold) Date 2009-07-09 17:49
In the Chessbase GUI, if you open the Engine Parameters window you can set MultiPV_cp. I am not aware of any other name for this UCI option.
Parent - - By titanium cranium (***) [us] Date 2009-07-09 17:54 Edited 2009-07-09 18:05
Not in mine. Nor is there anything at chessbase.com.

My GUI Version is Fritz 11.
Parent - - By Banned for Life (Gold) Date 2009-07-09 18:03
It is a Rybka engine parameter so you will not find anything about it at the Chessbase website. In the F8 GUI, it is accessed by Engine -> Change Main Engine -> Engine Parameters. In the F8 GUI it is on the top right, under Persistent Hash Merge File.
Parent - - By titanium cranium (***) [us] Date 2009-07-09 18:11 Edited 2009-07-09 18:18
I have no idea what the F8 GUI is, but my Persistent Hash Merge File says C:\...\*.rph.

The next button down is Persistent Hash Do Merge, below that is Contempt Play, below that Contempt Analyze.

I am either using a different GUI or there's an error in mine. I've owned Deep Rybka for 6 months, and have never been able to locate the mysterious Multipv_cp function, but it sounds important, and sounds like something I would use.

I must be using a different version of Deep Rybka3. Mine has a July 2008 date.
Parent - - By Banned for Life (Gold) Date 2009-07-09 18:18
Are you using Aquarium? What I gave you referred to a pre-F11 Chessbase GUI. I can't open Aquarium right now because both of my machines are running other things.
Parent - - By titanium cranium (***) [us] Date 2009-07-09 18:20
No, Deep Rybka 3, CD version, purchased early this year.

I have heard about this mysterious Multipv_cp function from the outset, and would really like to use it, but haven't made any progress in six months. Suggestions?
Parent - - By keoki010 (Bronze) [us] Date 2009-07-09 18:28
In Aquarium click on Engines mode.  In engines click on Rybka3 to highlight it.  On the bottom of the page, could be elseware on your display, there will be engine personality  scroll down to ponder.  2 lines under it is multipv_cp which has a defaut value of 32768.  Have fun....:lol:
Parent - - By titanium cranium (***) [us] Date 2009-07-09 18:33
Got it, piece of cake. This Engine personality window is not in my R3 GUI.

Now, what settings do i use for Multipv and Multipv_cp?
Parent - By keoki010 (Bronze) [us] Date 2009-07-09 21:56
I leave mine at the defaults.  If you click on analysis you can set the amount of pv's there.
Parent - - By Dadi Jonsson (Silver) [is] Date 2009-07-09 19:35

> As with all chess products, documentation is somewhere between poor and nonexistent.


It's confusing to give advice when you are not sure about the software being used, but in case it's Aquarium, I think this point is covered fairly well in the FAQ. See the answer to the question "What does MultiPV_cp do and what is the recommended value for this parameter?" in the Infinite Analysis section.
Parent - By titanium cranium (***) [us] Date 2009-07-09 20:13
Thanks, all this time I've been looking for multipv in Deep Rybka, but it's not there.
Parent - - By Vempele (Silver) [fi] Date 2009-07-09 16:28

>There's nothing in R3 materials or Help section on that term.


http://rybkachess.com/index.php?auswahl=Engine+parameters

rybkachess.com->Rybka 3 user area->Engine parameters.

How did you try to find it?
Parent - - By titanium cranium (***) [us] Date 2009-07-09 17:30 Edited 2009-07-09 17:44
I have spent many, many hours in the Engine Parameters window and there is nothing there (or in Help) that looks even remotely like "MultiPV_cp."  Therefore, I have assumed that is has some other name, correct?!
Parent - - By Banned for Life (Gold) Date 2009-07-09 18:29
I still don't know what GUI you are using. How the UCI options are listed will depend on this.
Parent - By titanium cranium (***) [us] Date 2009-07-09 21:38 Edited 2009-07-09 22:26
Neither do I, let's figure it out. And, how can I then change GUI's?

Ah ha! Fritz 11. I also have Arena, and Aquarium. I know, I'm a computer chess hacker, but it's not terminal.
Parent - By Banned for Life (Gold) Date 2009-07-08 22:19
5. Nf3

My last book move...
Parent - - By Uly (Gold) [mx] Date 2009-07-09 08:53
Ok, let's get this on!

5...Bd6

1. b3 d5 2. Bb2 Nf6 3. e3 Bf5 4. f4 e6 5. Nf3 Bd6
Parent - - By Nelson Hernandez (Silver) [us] Date 2009-07-09 11:49
This could be fun.  50.4% position with only 21.7% draws...very open-ended board position.
Parent - By Banned for Life (Gold) Date 2009-07-09 17:07
Yes, this opening can be very decisive if white goes for a kingside pawn storm or attempts to weaken blacks queenside pawn structure. I like to use this against opponents who are good centaurs but highly reliant on engine evals (i.e. not good OTB players). Unfortunately, Rybka won't go for either motif by itself, and will instead most likely back out with some anti-thematic move...
Parent - By Banned for Life (Gold) Date 2009-07-09 16:30
While I'm running this position, I'll put up the results from the last position below (run to test settings). I discovered that the Chessbase total time is wildly inaccurate, so I will give you the actual time taken along with the run (trying to keep it close to, but less than 30 minutes). Of course I will make sure that PH does not include any information that shouldn't be included, and hash will always be cleared before each run.

[Date "2009.07.09"]
[White "Deleted, DPA."]
[Black "Vytron, IA."]
[Result "*"]
[ECO "A01"]
[PlyCount "14"]

{2048MB, ATS-PENRYN} 1. b3 d5 2. Bb2 Nf6 3. e3 Bf5 4. f4 e6 5. Nf3 Be7 (5...
Bd6 6. Bd3 (6. Nh4 Be4 (6... Nc6 7. Nxf5 (7. Bb5 Bg4 $11 {[%eval -20,15]}) 7...
exf5 $11 {[%eval -4,15]}) (6... O-O 7. Be2 (7. Nxf5 exf5 $11 {[%eval -7,15]})
7... Bc5 $11 {[%eval -4,15]}) 7. d3 (7. Be2 h6 $11 {[%eval -13,15]}) 7... Bg6
$11 {[%eval -6,15]}) (6. Nc3 O-O (6... c6 7. Bd3 (7. Ne2 O-O $11 {
[%eval -10,15]}) 7... Bxd3 $11 {[%eval -8,15]}) (6... Nbd7 7. Bd3 (7. Nb5 Be7
$11 {[%eval -10,15]}) 7... Bxd3 $11 {[%eval -7,15]}) 7. Bd3 (7. Be2 a6 $11 {
[%eval -18,15]}) 7... Bxd3 $11 {[%eval -8,15]}) (6. Be2 h6 (6... Nc6 7. O-O (7.
Nc3 a6 $11 {[%eval -12,15]}) 7... h6 $11 {[%eval -5,15]}) (6... O-O 7. Nh4 (7.
O-O h6 $11 {[%eval -16,15]}) 7... Nc6 $11 {[%eval -5,15]}) 7. Nc3 (7. O-O O-O
$11 {[%eval -16,15]}) 7... a6 $11 {[%eval -12,15]}) 6... Bxd3 (6... Bg6 7. Qe2
(7. O-O Bxd3 $11 {[%eval 1,15]}) 7... Nc6 $11 {[%eval 5,15]}) (6... O-O 7. Bxf5
(7. O-O Bxd3 $11 {[%eval -6,15]}) 7... exf5 $11 {[%eval 6,15]}) 7. cxd3 O-O $11
{[%eval -5,15]}) (5... Nbd7 6. Bd3 (6. Nc3 c6 (6... Bd6 7. Bd3 (7. Nb5 Be7 $11
{[%eval -10,15]}) 7... Bxd3 $11 {[%eval -7,15]}) (6... Bb4 7. Bd3 (7. Ne2 Bg6
$11 {[%eval -8,15]}) 7... Bxd3 $11 {[%eval -4,15]}) 7. Bd3 (7. h3 Bd6 $11 {
[%eval -8,15]}) 7... Bxd3 $11 {[%eval -7,15]}) (6. Nh4 g6 (6... Be7 7. Be2 (7.
Nxf5 exf5 $11 {[%eval -9,15]}) 7... Bg4 $11 {[%eval -6,15]}) (6... Bg4 7. Be2 (
7. Nf3 c6 $11 {[%eval -17,15]}) 7... Bxe2 $11 {[%eval -6,15]}) 7. Be2 (7. c4
Bg7 $11 {[%eval -7,15]}) 7... Bg7 $11 {[%eval -7,15]}) (6. Be2 h6 (6... Bc5 7.
Bd3 (7. Nc3 c6 $11 {[%eval -9,15]}) 7... O-O $11 {[%eval -8,15]}) (6... c6 7.
O-O (7. Bd3 Bxd3 $11 {[%eval -11,15]}) 7... h6 $11 {[%eval -9,15]}) 7. Nc3 (7.
d3 Bc5 $11 {[%eval -17,15]}) 7... c6 $11 {[%eval -15,15]}) 6... Bxd3 (6... Be4
7. Bxe4 (7. O-O Be7 $11 {[%eval -2,15]}) 7... dxe4 $11 {[%eval 3,15]}) (6...
Bg6 7. Bxg6 (7. O-O Bxd3 $11 {[%eval -2,15]}) 7... hxg6 $11 {[%eval -1,15]}) 7.
cxd3 Bd6 $11 {[%eval -5,15]}) (5... c6 6. Bd3 (6. Nh4 Be4 (6... g6 7. Be2 (7.
Nf3 Nbd7 $11 {[%eval -15,15]}) 7... Bg7 $11 {[%eval -9,15]}) (6... Bg4 7. Be2 (
7. Nf3 Nbd7 $11 {[%eval -16,15]}) 7... Bxe2 $11 {[%eval -4,15]}) 7. Nc3 (7. Be2
h6 $11 {[%eval -14,15]}) 7... Be7 $11 {[%eval -10,15]}) (6. Nc3 Nbd7 (6... Bd6
7. Bd3 (7. Ne2 Nbd7 $11 {[%eval -9,15]}) 7... Bxd3 $11 {[%eval -7,15]}) (6...
Be7 7. Bd3 (7. a3 O-O $11 {[%eval -8,15]}) 7... Bxd3 $11 {[%eval -4,15]}) 7.
Bd3 (7. h3 Bg6 $11 {[%eval -11,15]}) 7... Bxd3 $11 {[%eval -7,15]}) (6. Be2 h6
(6... Bc5 7. O-O (7. Bd3 O-O $11 {[%eval -6,15]}) 7... O-O $11 {[%eval -6,15]})
(6... Nbd7 7. O-O (7. Nh4 Bc5 $11 {[%eval -12,15]}) 7... h6 $11 {[%eval -10,15]
}) 7. O-O (7. Bd3 Bxd3 $11 {[%eval -11,15]}) 7... Nbd7 $11 {[%eval -10,15]})
6... Bxd3 (6... Bg6 7. Bxg6 (7. O-O Bxd3 $11 {[%eval -3,15]}) 7... hxg6 $11 {
[%eval 0,15]}) (6... Be7 7. Bxf5 (7. O-O Bxd3 $11 {[%eval -5,15]}) 7... exf5
$11 {[%eval 3,15]}) 7. cxd3 h6 $11 {[%eval -5,15]}) (5... h6 6. Nc3 (6. Bd3
Bxd3 (6... Be7 7. Bxf5 (7. O-O Bxd3 $11 {[%eval -8,15]}) 7... exf5 $11 {
[%eval 1,15]}) (6... Nc6 7. Bxf5 (7. O-O Bxd3 $11 {[%eval 0,15]}) 7... exf5 $11
{[%eval 1,15]}) 7. cxd3 Be7 $11 {[%eval -8,15]}) (6. Ne5 Nbd7 (6... Be7 7. Bd3
(7. Nc3 O-O $11 {[%eval -19,15]}) 7... Bxd3 $11 {[%eval -12,15]}) (6... c6 7.
Bd3 (7. a4 Nbd7 $11 {[%eval -21,15]}) 7... Bxd3 $11 {[%eval -12,15]}) 7. Bd3 (
7. Bb5 Be7 $11 {[%eval -25,15]}) 7... Ne4 $11 {[%eval -18,15]}) (6. Nd4 Bg6 (
6... Bh7 7. Bd3 (7. Nc3 g5 $11 {[%eval -22,15]}) 7... Bxd3 $11 {[%eval -12,15]}
) (6... Be4 7. Nc3 (7. Nf3 Be7 $11 {[%eval -11,15]}) 7... Bh7 $11 {
[%eval -7,15]}) 7. Nc3 (7. Bd3 Bxd3 $11 {[%eval -12,15]}) 7... Bd6 $11 {
[%eval -12,15]}) 6... Be7 (6... c6 7. h3 (7. Bd3 Bxd3 $11 {[%eval -7,15]}) 7...
Bh7 $11 {[%eval -5,15]}) (6... Bb4 7. Bd3 (7. Ne2 O-O $11 {[%eval -4,15]}) 7...
Bxd3 $11 {[%eval -3,15]}) 7. Bd3 (7. Bb5+ Nbd7 $11 {[%eval -8,15]}) 7... Bxd3
$11 {[%eval -8,15]}) 6. Bd3 (6. Nc3 O-O (6... c5 7. Bb5+ (7. Ne5 h6 $11 {
[%eval -10,15]}) 7... Nbd7 $11 {[%eval -7,15]}) (6... h6 7. Bd3 (7. Bb5+ c6 $11
{[%eval -8,15]}) 7... Bxd3 $11 {[%eval -4,15]}) 7. Bd3 (7. Be2 a6 $11 {
[%eval -17,15]}) 7... Bxd3 $11 {[%eval -12,15]}) (6. Bb5+ Nbd7 (6... c6 7. Bd3
(7. Be2 O-O $11 {[%eval -18,15]}) 7... Bxd3 $11 {[%eval -12,15]}) (6... Nfd7 7.
Nc3 (7. O-O O-O $11 {[%eval -10,15]}) 7... a6 $11 {[%eval -10,15]}) 7. O-O (7.
Nc3 O-O $11 {[%eval -14,15]}) 7... O-O $11 {[%eval -14,15]}) (6. Nh4 Nc6 (6...
O-O 7. Nxf5 (7. Be2 Be4 $11 {[%eval -16,15]}) 7... exf5 $11 {[%eval -15,15]}) (
6... Be4 7. Nc3 (7. Nf3 O-O $11 {[%eval -11,15]}) 7... O-O $11 {[%eval -11,15]}
) 7. Bb5 (7. Nxf5 exf5 $11 {[%eval -19,15]}) 7... O-O $11 {[%eval -16,15]})
6... Bxd3 (6... O-O 7. Bxf5 (7. Nc3 Bxd3 $11 {[%eval -12,15]}) 7... exf5 $11 {
[%eval -5,15]}) (6... c5 7. Bxf5 (7. O-O Bxd3 $11 {[%eval -12,15]}) 7... exf5
$11 {[%eval -5,15]}) 7. cxd3 O-O $11 {[%eval -12,15]} *
Parent - - By Banned for Life (Gold) Date 2009-07-09 17:41
This move took me 53 minutes, so please scale your time appropriately (Use of total time of 30 minutes resulted in 4 minutes and 35 seconds of analysis). Future moves should be faster, due to warm PH and maybe opening book. The move itself is not as bad as I feared (I thought it would go with the anti-thematic Bd3).

6. Nh4

1. b3 d5 2. Bb2 Nf6 3. e3 Bf5 4. f4 e6 5. Nf3 Bd6 6. Nh4

[White "New game"]
[Black "Rybka 3"]
[Result "*"]
[PlyCount "15"]

{2048MB, vytron.ctg, ATS-PENRYN} 1. b3 d5 2. Bb2 Nf6 3. e3 Bf5 4. f4 e6 5. Nf3
Bd6 6. Nh4 (6. Nc3 c6 (6... Nbd7 7. d3 (7. h3 O-O (7... c6 8. g4 $11 {
[%eval -1,15]}) 8. g4 $11 {[%eval -10,15]}) (7. Bd3 Bxd3 (7... c6 8. Bxf5 $11 {
[%eval 0,15]}) 8. cxd3 $11 {[%eval -10,15]}) 7... O-O (7... Bg6 8. Qd2 $11 {
[%eval -5,15]}) 8. Qd2 $11 {[%eval -5,15]}) (6... c5 7. Nh4 (7. Bb5+ Nc6 (7...
Nbd7 8. Ne2 $11 {[%eval -2,15]}) 8. O-O $11 {[%eval -6,15]}) (7. Bd3 Bxd3 (7...
Nc6 8. Bxf5 $11 {[%eval 0,15]}) 8. cxd3 $11 {[%eval -9,15]}) 7... Bg4 (7... O-O
8. Nxf5 $11 {[%eval 5,15]}) 8. Be2 $11 {[%eval -4,15]}) (6... O-O 7. Bd3 (7.
Nb5 Be7 (7... Nc6 8. Nxd6 $11 {[%eval 1,15]}) 8. Bd3 $11 {[%eval -21,15]}) (7.
Be2 a6 (7... c6 8. Bd3 $11 {[%eval -12,15]}) 8. O-O $11 {[%eval -13,15]}) 7...
Bxd3 (7... c6 8. Bxf5 $11 {[%eval 3,15]}) 8. cxd3 $11 {[%eval -6,15]}) 7. h3 (
7. Bd3 Bxd3 (7... Nbd7 8. Bxf5 $11 {[%eval 0,15]}) 8. cxd3 $11 {[%eval -10,15]}
) (7. Nh4 O-O (7... Nbd7 8. Nxf5 $11 {[%eval -12,15]}) 8. Nxf5 $11 {
[%eval -12,15]}) 7... O-O (7... Bg6 8. Bd3 $11 {[%eval -5,15]}) 8. d3 $11 {
[%eval -9,15]}) (6. Bd3 Bxd3 (6... Bg6 7. Qe2 (7. O-O Bxd3 (7... c5 8. Bxf6 $11
{[%eval 5,15]}) 8. cxd3 $11 {[%eval 0,15]}) (7. Nc3 a6 (7... Bxd3 8. cxd3 $11 {
[%eval -1,15]}) 8. O-O $11 {[%eval -3,15]}) 7... Nc6 (7... Bxd3 8. Qxd3 $11 {
[%eval 8,15]}) 8. O-O $11 {[%eval 4,15]}) (6... O-O 7. Bxf5 (7. O-O Bxd3 (7...
Be4 8. Bxe4 $11 {[%eval 4,15]}) 8. cxd3 $11 {[%eval -7,15]}) (7. Nc3 Bxd3 (7...
c6 8. Bxf5 $11 {[%eval 3,15]}) 8. cxd3 $11 {[%eval -6,15]}) 7... exf5 8. O-O
$11 {[%eval 6,15]}) (6... Be4 7. Qe2 (7. Bxe4 dxe4 (7... Nxe4 8. c4 $11 {
[%eval 15,15]}) 8. Ne5 $11 {[%eval 3,15]}) (7. O-O Bxd3 (7... O-O 8. Bxe4 $11 {
[%eval 4,15]}) 8. cxd3 $11 {[%eval 0,15]}) 7... O-O (7... Bxd3 8. Qxd3 $11 {
[%eval 8,15]}) 8. Bxe4 $11 {[%eval 3,15]}) 7. cxd3 O-O (7... h6 8. O-O $11 {
[%eval -7,15]}) 8. O-O $11 {[%eval -7,15]}) (6. Bb5+ c6 (6... Nbd7 7. Bd3 (7.
O-O O-O (7... h6 8. Bxd7+ $11 {[%eval -11,15]}) 8. Bd3 $11 {[%eval -12,15]}) (
7. Bxd7+ Nxd7 (7... Qxd7 8. d3 $11 {[%eval -6,15]}) 8. O-O $11 {[%eval -17,15]}
) 7... Bxd3 (7... Bg4 8. Nc3 $11 {[%eval -11,15]}) 8. cxd3 $11 {[%eval -12,15]}
) (6... Nfd7 7. O-O (7. Bd3 Bxd3 (7... O-O 8. Bxf5 $11 {[%eval 14,15]}) 8. cxd3
$11 {[%eval -6,15]}) (7. d4 Nc6 (7... c6 8. Bd3 $11 {[%eval -2,15]}) 8. Bd3 $11
{[%eval -6,15]}) 7... O-O (7... a6 8. Be2 $11 {[%eval 0,15]}) 8. d4 $11 {
[%eval -3,15]}) (6... Nc6 7. d3 (7. Qe2 Bg6 (7... O-O 8. Bxc6 $11 {
[%eval 11,15]}) 8. Bxc6+ $11 {[%eval 10,15]}) (7. Nd4 O-O (7... a6 8. Bxc6+ $14
{[%eval 52,15]}) 8. Nxc6 $11 {[%eval 14,15]}) 7... O-O (7... a5 8. Nd4 $14 {
[%eval 29,15]}) 8. Bxc6 $11 {[%eval 15,15]}) 7. Bd3 (7. Be2 Nbd7 (7... h6 8. c4
$11 {[%eval -19,15]}) 8. Nh4 $11 {[%eval -21,15]}) 7... Bxd3 (7... Bg6 8. O-O
$11 {[%eval -8,15]}) 8. cxd3 $11 {[%eval -12,15]}) (6. Be2 h6 (6... Be7 7. Nh4
(7. O-O h6 (7... O-O 8. Nh4 $11 {[%eval 0,15]}) 8. Nc3 $11 {[%eval -2,15]}) (7.
Nc3 O-O (7... h6 8. O-O $11 {[%eval -2,15]}) 8. O-O $11 {[%eval -2,15]}) 7...
O-O (7... Nc6 8. Nxf5 $11 {[%eval 0,15]}) 8. Nxf5 $11 {[%eval 0,15]}) (6... a6
7. Nh4 (7. O-O h6 (7... Nc6 8. Nh4 $11 {[%eval 0,15]}) 8. Bd3 $11 {
[%eval -11,15]}) (7. c4 h6 (7... Nc6 8. Nh4 $11 {[%eval -4,15]}) 8. O-O $11 {
[%eval -10,15]}) 7... Bc5 (7... Be4 8. d3 $11 {[%eval 0,15]}) 8. O-O $11 {
[%eval -2,15]}) (6... Nc6 7. O-O (7. Nh4 d4 (7... Bc5 8. O-O $11 {[%eval -7,15]
}) 8. Nxf5 $15 {[%eval -28,15]}) (7. Bb5 O-O (7... Bg6 8. O-O $11 {[%eval 0,15]
}) 8. Bxc6 $11 {[%eval 0,15]}) 7... O-O (7... a6 8. Nh4 $11 {[%eval 0,15]}) 8.
Nh4 $11 {[%eval 0,15]}) 7. O-O (7. Nc3 a6 (7... c6 8. d3 $11 {[%eval -9,15]})
8. O-O $11 {[%eval -14,15]}) (7. c4 O-O (7... c6 8. O-O $11 {[%eval -11,15]})
8. O-O $11 {[%eval -11,15]}) 7... Nbd7 (7... O-O 8. c4 $11 {[%eval -11,15]}) 8.
d4 $11 {[%eval -11,15]}) 6... Be4 (6... O-O 7. Be2 (7. Nxf5 exf5 8. Bd3 $11 {
[%eval -5,15]}) (7. c4 Be4 (7... Ne4 8. g3 $11 {[%eval -2,15]}) 8. Be2 $11 {
[%eval -24,15]}) 7... Nc6 (7... Be4 8. d3 $11 {[%eval 1,15]}) 8. O-O $11 {
[%eval 1,15]}) (6... Nbd7 7. Nxf5 (7. Be2 O-O (7... Be4 8. d3 $11 {[%eval 3,15]
}) 8. O-O $11 {[%eval 3,15]}) (7. c4 Be4 (7... Bg4 8. Be2 $11 {[%eval 0,15]})
8. Nc3 $11 {[%eval -10,15]}) 7... exf5 8. Be2 $11 {[%eval 3,15]}) (6... Bg4 7.
Be2 (7. Bb5+ Nbd7 (7... Nc6 8. Bxc6+ $11 {[%eval -17,15]}) 8. Be2 $11 {
[%eval -19,15]}) (7. Nf3 O-O (7... h6 8. Be2 $11 {[%eval -14,15]}) 8. Nc3 $11 {
[%eval -19,15]}) 7... Bxe2 (7... h5 8. Nf3 $11 {[%eval 3,15]}) 8. Qxe2 $11 {
[%eval 0,15]}) 7. Nc3 (7. d3 Bf5 (7... Bg6 8. Be2 $11 {[%eval -5,15]}) 8. Be2
$11 {[%eval -5,15]}) (7. Be2 h6 (7... Nc6 8. d3 $11 {[%eval -5,15]}) 8. O-O $11
{[%eval -11,15]}) 7... Qe7 (7... Be7 8. Be2 $11 {[%eval 3,15]}) 8. Be2 $11 {
[%eval -2,15]} *
Parent - By Banned for Life (Gold) Date 2009-07-09 18:33
Does anyone know if there is a way to capture and display the DPA information here, as it is laid out in the CB GUI? I think the PGN has all the same info, but its almost impossible to read it...
Parent - - By Uly (Gold) [mx] Date 2009-07-09 20:51

>The move itself is not as bad as I feared (I thought it would go with the anti-thematic Bd3).


Me too, does it play Nh4 if set to LESS than 30 minutes? I'd like to scale to that, I don't want to sit for 1 hour 21 minutes just because it overstep your time limit and mine. I'm in my 64bit partition so I can't use all my 32bit software while I'm waiting.

Can we please restart move 6 in where you play a 30 minutes or less move? If you don't accept I understand, and I'd sit for 82 minutes, but this out of book move took you almost twice the allowed time.

I'm also willing to only use 46 minutes this move if you take 7 minutes the next (so total time of both your moves goes back to one hour) :P
Parent - - By Banned for Life (Gold) Date 2009-07-09 21:09
The problem is that I don't have any way to precisely control how much time DPA will take. Of course this move should take the longest, since there is no history to rely on.

As you mention, there are several options here:

1) I can make a third run, to get back under 30 minutes. In this case, all moves will be done in under 30 minutes and you will have to scale your time down accordingly.
2) I can do future runs with the same settings. Since I am using PH filled in during this move, the future runs at the same settings, will be much faster (almost certainly less than 30 minutes).
3) You can bank the time to allow you to run all moves at a given depth, or using any other strategy to make best use of your time.

I would prefer options 2 or 3 to get a consistent quality of play from DPA. Making up for lack of history by using shallow depths is a pet peeve of mine. Also, with DPA, as you decrease depth, efficiency (in terms of total node count) drops like a stone as you end up spending more time switching between positions than analyzing.
Parent - - By Uly (Gold) [mx] Date 2009-07-09 21:14
Ok, I'll got with 3, no problem.
Parent - - By Banned for Life (Gold) Date 2009-07-09 21:32
OK, I will await your move. I did a test using IA on the position before white's move 6, and it seems that R3 will never play anything other than Bd3 at lower and higher depths, and Nc3 at some intermediate depths. So it is clear that the style of play will certainly be influenced by the use of DPA. But here we hope to find out more about the quality of play. Note that all positions in the DPA that I ran were evaluated to depth 15. I assume you will reach depth 20 or 21 in the same time period.

My prediction going in would be that DPA may outperform IA in the opening and in the endgame, but I'm not sure it will survive the middle game.
Parent - - By Uly (Gold) [mx] Date 2009-07-09 22:16
I reached depth 21 at 33:18, the irony is that I figured out I should warm up the hash as well! So, sitting until iteration 22 is done with seems like a must even if I don't bank much time.

How are we going to deal with obvious recaptures, though? Is it beneficial for DPA to run full strength on them while IA wouldn't benefit as much? Do I play them ASAP and bank the extra time? (does the time I don't use get banked?) Or do we ignore time controls on them and just play immediately? Or something else?
Parent - By Banned for Life (Gold) Date 2009-07-09 22:22
Actually, because of Rybka's problems with +/- 5.09, obvious recaptures are problematic for DPA much more often than for IA. It is very easy to imagine a situation where Rybka will sit at the root, where its large branching factor captures positions that are won or lost.

I suggest we bank the time in cases where there is clearly only one reasonable move.
Parent - By Banned for Life (Gold) Date 2009-07-09 22:22
Actually, because of Rybka's problems with +/- 5.09, obvious recaptures are problematic for DPA much more often than for IA. It is very easy to imagine a situation where Rybka will sit at the root, where its large branching factor captures positions that are won or lost.

I suggest we bank the time in cases where there is clearly only one reasonable move.
Parent - - By Banned for Life (Gold) Date 2009-07-09 22:23
Look! A double post! I didn't think this was possible here!
Parent - By Vempele (Silver) [fi] Date 2009-07-09 22:26
Look! A quadruple post!

> I didn't think this was possible here!


As long as the posts are more than 5 seconds apart.
Parent - By Vempele (Silver) [fi] Date 2009-07-09 22:26
Look! A quadruple post!

> I didn't think this was possible here!


As long as the posts are more than 5 seconds apart.
Parent - - By Banned for Life (Gold) Date 2009-07-09 22:25
I'm glad you're spending extra time on the first move out of book. I'm always unhappy when the first move out of book is analyzed to a lower depth than succeeding moves.
Parent - By Uly (Gold) [mx] Date 2009-07-09 22:34
Yup, though apparently the move will be the same than at depth 21, at least it won't be Nc6!
Parent - By Nelson Hernandez (Silver) [us] Date 2009-07-09 22:50
6.Nh4 kicked me out book too.  What a weird choice--I had 11 alternative moves in book!  Interesting; this has a definite logic to it.  I can already envision an all-out attack on that side.
Parent - - By titanium cranium (***) [us] Date 2009-07-09 22:30 Edited 2009-07-09 22:35
Which is better in the end game? My experience is R3 struggles badly in pure endgame. Or, is there another program that's clearly better in the endgame?
Parent - By Banned for Life (Gold) Date 2009-07-09 23:18
Rybka is fine in most endgames. In some endgames, I prefer engines that allow turning off null-move pruning. If EGTBs are important, Shredder is probably the best choice.
Parent - By Uly (Gold) [mx] Date 2009-07-09 20:55
Or you could let me use the extra 36 minutes I'd get in this move sparingly on future moves as I see fit, I'd be happy with that.
Parent - - By Uly (Gold) [mx] Date 2009-07-09 22:49
Time banked before move: 35:32

6...O-O

1. b3 d5 2. Bb2 Nf6 3. e3 Bf5 4. f4 e6 5. Nf3 Bd6 6. Nh4 O-O

rn1q1rk1/ppp2ppp/3bpn2/3p1b2/5P1N/1P2P3/PBPP2PP/RN1QKB1R w KQ -


Engine: Rybka 3 (64 MB)
by Vasik Rajlich,Larry Kaufman
12.01   0:01   -0.09   6...Bg4 7.Be2 Bxe2 8.Qxe2 O-O 9.O-O Nbd7 10.Nf3 h6 11.Qb5 Nb6 (152.793) 120
13.01   0:02   -0.06   6...Bg4 7.Be2 Bxe2 8.Qxe2 O-O 9.O-O Nbd7 10.Nf3 c6 11.d3 Qb6 12.Nbd2 (291.486) 118
14.01   0:06   -0.04   6...Bg4 7.Be2 Bxe2 8.Qxe2 O-O 9.O-O Nbd7 10.c4 c6 11.cxd5 Nxd5 12.Nf3 Qb6 13.Kh1 (781.595) 115
14.02   0:12   -0.07   6...O-O 7.Nxf5 exf5 8.Bd3 g6 9.O-O Nbd7 10.c4 Nc5 11.Be2 dxc4 12.Bxc4 c6 13.Qf3 (1.617.601) 127
15.01   0:22   -0.09   6...O-O 7.Nxf5 exf5 8.Bd3 g6 9.O-O Nbd7 10.c4 Nc5 11.Be2 dxc4 12.Bxc4 c6 13.b4 Ncd7 (2.861.409) 130
16.01   0:55   -0.07   6...O-O 7.Nxf5 exf5 8.c4 Qe7 9.Be2 dxc4 10.bxc4 c6 11.O-O Na6 12.Qc2 Ne4 13.Qc1 Nac5 14.Nc3 (7.060.822) 129
17.01   1:51   -0.02   6...O-O 7.Be2 Nc6 8.O-O Bc5 9.d4 Bb4 10.Nxf5 exf5 11.a3 Bd6 12.Nc3 Qe7 13.Qd2 Rae8 (14.157.162) 129
17.03   3:42   -0.03   6...Be4 7.d3 Bg6 8.Be2 Nc6 9.O-O Qd7 10.Nf3 O-O 11.Ne5 Qe7 12.Nd2 Bc5 (28.556.646) 131
18.01   5:48   -0.02   6...Be4 7.d3 Bf5 8.Be2 h6 9.O-O Bh7 10.Nf3 O-O 11.Ne5 Qe7 12.Nd2 Nbd7 13.Ndf3 (44.987.894) 132
19.01  10:38   -0.04   6...Be4 7.d3 Bf5 8.Be2 h6 9.O-O Bh7 10.Nf3 O-O 11.Ne5 Bxe5 12.fxe5 Nfd7 13.Nd2 Qg5 14.Rf3 (85.295.047) 136
19.24  12:05   -0.05   6...Nc6 7.Nxf5 exf5 8.Bb5 O-O 9.O-O Qe7 10.a3 Ne4 11.Nc3 Nxc3 12.Bxc3 Bxa3 (95.086.085) 134
20.01  13:32   -0.04   6...Nc6 7.Nxf5 exf5 8.Bb5 O-O 9.O-O Qe7 10.a3 Ne4 11.Nc3 Nxc3 12.Bxc3 Bxa3 (108.946.618) 137
20.03  28:39   -0.08   6...O-O 7.Be2 Be4 8.d3 Bf5 9.O-O h6 10.a3 Bh7 11.Nf3 Nc6 12.Qd2 Bc5 13.Bxf6 Qxf6 14.d4 Be7 15.Bb5 (231.168.048) 137
21.01  33:18   -0.07   6...O-O 7.Be2 Be4 8.d3 Bf5 9.O-O h6 10.a3 Bh7 11.Nf3 Nc6 12.Qd2 Bc5 13.Bxf6 Qxf6 14.d4 Be7 15.Bb5 (274.319.375) 140
22.01  70:00   -0.08   6...O-O 7.Nxf5 exf5 8.Bd3 (597.472.634) 145
best move: O-O time: 79:33.578 min  n/s: 142.528  nodes: 664.415.436

Time banked after move: 2:19 :yell:
Parent - - By Nelson Hernandez (Silver) [us] Date 2009-07-09 23:13
If he moves Be2 I'll be back in book via transposition.
Parent - By Banned for Life (Gold) Date 2009-07-09 23:58
If I were playing white here, I would trade for the bishop, castle queenside, and mount an all out kingside attack. I would love to have this type of position against the unaided cluster.
Up Topic The Rybka Lounge / Correspondence Chess / Deleted Vs. Vytron (DPA Vs. IA) 1/2 - 1/2
1 2 3 4 5 6 Previous Next  

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill