Not logged inRybka Chess Community Forum
Up Topic Rybka Support & Discussion / Aquarium / Pachman - Hromadka, Prague Ch 1944
- - By Dadi Jonsson (Silver) Date 2009-02-03 23:48 Edited 2009-02-04 00:05
I was asking for more IDeA test positions, so I decided to let IDeA have a look at a couple of positions that have recently been discussed on other forums. This is the first position and I'll start a new topic for the other one (probably later this week).

8/8/6k1/2p1p3/2P1P1K1/3N4/8/b7 w - - 0 1


This position has puzzled chess players and engines for many years. It has been discussed on various forums (at least 2 or 3 times on CCC since 2002) and in several books and articles. Pachman (see his Complete Chess Strategy trilogy) seized the opportunity to gain an immediate material advantage by playing 1.Nxc5, after which the game can no longer be won. This is also the move chosen by most or all chess engines.

Instead of 1.Nxc5 Pachman recommended 1.Ne1 with the goal of making progress on the kingside by manoeuvring his knight. In that case Black's c5-pawn would actually shield White's c-pawn from attacks by the Black king. Graham Burgess in his 1997 book The Mammoth Book of Chess agreed that 1.Nxc5 was bad, adding that "This obvious, materialistic move throws away the win!". He also recommended Pachman's "winning" plan.

I think that Robin Smith in his Modern Chess Analysis (2004?) was the first to show that the "winning plan" was actually not winning at all. After White succeeds in forcing the black king to back away towards the center Robin's plan was to wait for the appropriate moment and then rush towards a5 with his king in order to attack White's c4 with Kb4. This is an unexpected defense, but actually seems to hold. Karsten Müller discussed this "amazing surprise" in one of his Endgame Corner columns on ChessCafe.

I ran IDeA on this position. It liked 1.Nxc5, so after a while I excluded that move from the analysis (colored it red). It came up with lots of different lines, all leading to equality. The most interesting result of the analysis was that Black seems to have a much simpler defense than the one found by Robin Smith. Here is the start of the main variation from Karsten's article:

1.Ne1 Bd4 2.Nf3 Bc3 3.Nh4+ Kf6 4.Kh5

8/8/5k2/2p1p2K/2P1P2N/2b5/8/8 b - -


Here Karsten continues with 4...Bd4 5.Nf5 Bc3 6.Ne3 Bd4 7.Ng4+ Ke6 8.Kg6 and White has succeeded in pushing the black king back and will sooner or later win Black's e-pawn.

It seems, however, that Black can achieve the draw in a simpler and more convincing way by playing 4...Bd2! instead of 4...Bd4:

8/8/5k2/2p1p2K/2P1P2N/8/3b4/8 w - -


This move, which was found by IDeA, stops any attempts by White to make progress on the kingside and Black will easily hold the draw. He doesn't have to worry even if the knight succeeds in breaking through (e.g. Nh5-f5-d6) as he will be harmless without the support of the king.

After 4...Bd2 Black can prevent White from following the same plan as in Karsten's variation above: 5.Nf5 Bf4

8/8/5k2/2p1pN1K/2P1Pb2/8/8/8 w - -


Black has no way to transfer his knight to g4 (with the purpose of driving the black king away). The black king's only role is to prevent the white king from reaching g6 and f5 (the bishop covers the other key squares). He has plenty of squares to do that (Kf6, Kf7, Kg7, Kh7 covers g6 and similar for f5), so there is no danger of ending up in a zugzwang.

The main challenge when analyzing a completely drawn position like this one is to come up with a main variation that makes sense to a human player and helps him understand why the position is drawn. I think that the defense shown here is easier to understand than the one found by Robin Smith.
Parent - - By Roland Rösler (****) Date 2009-02-04 03:11
Very fine! You always speak about IDeA, but not about the engine(s) you used. Please, give us a hint!

Okay, you suggest IDeA can help in endgame analysis. In your example we have an 8 man TBs and after Nxc5 only 7 man. Let an trustworthy! engine run with 6 man TBS on fast SSD (< 0.1ms) on octa (quad?) and you have the result, that Nxc5 don´t win (result in an hour or much quicker)! A normal user with some time see it too! :-) So you have to prove, position is drawn (next best move). And in your line 4. ... Bd2! isn´t the only move (Bb2, Be1)!

Show me IDeA works in endgame analysis (and I will buy Aquarium :-)). Here are the test positions (Bronstein give up with black in move 57). Last results (Schach 7/2008; Oliver Zierke vs. Rybka/Shirov)): White wins after move 57 (in move 76. Ba3)! Try it from the end (move 62)!

8/1p3k1p/np2p3/n2p1p2/P2PP3/2KB1PB1/7P/8 b - - 0 41


2n5/1p2n3/1p4k1/3p2Bp/P2P3P/2K2B2/8/8 b - - 0 57


8/8/2pn4/Pp3k1p/3P3P/2K5/8/2B5 b - - 0 62


the whole game (incl. Botvinniks line from 57-62)

[Event "World Championship 19th"]
[Site "Moscow"]
[Date "1951.05.08"]
[Round "23"]
[White "Botvinnik, Mikhail"]
[Black "Bronstein, David I"]
[Result "1-0"]
[ECO "D71"]
[PlyCount "123"]
[EventDate "1951.03.16"]
[EventType "match"]
[EventRounds "24"]
[EventCountry "URS"]
[Source "ChessBase"]

1. d4 Nf6 2. c4 g6 3. g3 c6 4. Bg2 d5 5. cxd5 cxd5 6. Nc3 Bg7 7. Nh3 Bxh3 8.
Bxh3 Nc6 9. Bg2 e6 10. e3 O-O 11. Bd2 Rc8 12. O-O Nd7 13. Ne2 Qb6 14. Bc3 Rfd8
15. Nf4 Nf6 16. Qb3 Ne4 17. Qxb6 axb6 18. Be1 Na5 19. Nd3 Bf8 20. f3 Nd6 21.
Bf2 Bh6 22. Rac1 Nac4 23. Rfe1 Na5 24. Kf1 Bg7 25. g4 Nc6 26. b3 Nb5 27. Ke2
Bf8 28. a4 Nc7 29. Bg3 Na6 30. Bf1 f6 31. Red1 Na5 32. Rxc8 Rxc8 33. Rc1 Rxc1
34. Nxc1 Ba3 35. Kd1 Bxc1 36. Kxc1 Nxb3+ 37. Kc2 Na5 38. Kc3 Kf7 39. e4 f5 40.
gxf5 gxf5 41. Bd3 Kg6 42. Bd6 Nc6 43. Bb1 Kf6 44. Bg3 fxe4 45. fxe4 h6 46. Bf4
h5 47. exd5 exd5 48. h4 Nab8 49. Bg5+ Kf7 50. Bf5 Na7 51. Bf4 Nbc6 52. Bd3 Nc8
53. Be2 Kg6 54. Bd3+ Kf6 55. Be2 Kg6 56. Bf3 N6e7 57. Bg5 Nc6 58. Bxd5 Nd6 59.
Bf3 Kf5 60. Bc1 b5 61. Bxc6 bxc6 62. a5 1-0
Parent - - By buffos (Silver) Date 2009-02-04 06:45
The whole tone of your answer is not polite (to say the least).

> Show me IDeA works in endgame analysis (and I will buy Aquarium :-)).


you dont need to.

Analyse for example Bronstein - Simagin 1961 ch rus (position after d5)

Its not about the eval. After 3 days on IDEA, the wealth of variations IDEA gave is amazing. You actually FEEL what is going on. What works and what not. Various fine details you can never see in IA and much more.
Parent - - By Roland Rösler (****) Date 2009-02-04 17:12
The whole tone of your answer is not polite (to say the least).

This wasn´t intended.
Parent - By buffos (Silver) Date 2009-02-04 17:14
no hard feelings :)
Parent - By Rowlando (***) Date 2009-02-04 09:56 Edited 2009-02-04 10:14
I'm fairly certain that IDeA is capable of finding a win in this endgame, given enough time.  In just a few minutes on ancient hardware, I was able to clearly see that IDeA was making forward progress at a steady rate. 

I didn't run IDeA long enough to see where this goes but it's at least promising and shows that IDeA is capable of finding a way through complex endgame positions. 

42.Bb1 dxe4 43.fxe4 Kg5 44.exf5 exf5 45.d5 Nc5 46.Bc7 f4 47.Bxh7 f3 48.Bd3 Nxd3 49.Kxd3 Nc4 50.Bg3 Nb2+ 51.Ke3 (51.Ke4 Nxa4 52.d6 b5 53.Kxf3 Kf6 54.Be5+ Ke6 55.h4 Nc5 56.Bc3 Nd7 (56...Kxd6) 57.Kg4)51...Nxa4 52.Kxf3 b5 53.d6 etc. etc.  If she eventually reaches the point where it's obvious that the line is drawn, she'll circle back and try another route -- it's just a matter of time before the correct path will be found.  If you don't buy Aquarium, download the demo so you can test the position.
Parent - By Dadi Jonsson (Silver) Date 2009-02-04 18:25

> the engine(s) you used.


Rybka, of course. Is there another engine? ;)

> Okay, you suggest IDeA can help in endgame analysis.


Yes.

> and you have the result, that Nxc5 don´t win


This was already known after the game was played (1944). Most of the analysis since then has been concerned with finding a way to win without capturing the pawn immediately.

> And in your line 4. ... Bd2! isn´t the only move (Bb2, Be1)!


Yes, there are many ways to draw. The main point of my post was to show an easier (and more easily understood) way of drawing this position. I don't think this method has been shown before. That was the only reason I attached a "!" to 4...Bd2.

> Show me IDeA works in endgame analysis (and I will buy Aquarium :))


:)

> Here are the test positions (Bronstein give up with black in move 57). Last results (Schach 7/2008; Oliver Zierke vs. Rybka/Shirov)): White wins after move 57 (in move 76. Ba3)! Try it from the end (move 62)!


You always seem to come up with more interesting (and more difficult :)) positions than most others. Just for fun I started IDeA from the final postition (move 62) on my slow laptop (30kn/s) and only gave it 10 seconds/position. I forgot about it and didn't check until a few hours later, but then the evaluation for the best moves was +5.50 (it started below +3). I only had time to check a couple of variations, but they looked convincing. So it seems that 10 sec/pos is plenty of time on a faster computer to "solve" this position even without touching the interactive features of IDeA. Of course the other two positions are harder. I'll run some tests on a more powerful computer and post the results.

I plan to build a single tree, starting with the final position and then move backwards in a few steps up to the first position you posted.
Parent - - By Dadi Jonsson (Silver) Date 2009-02-05 09:22

> 57. Bg5 Nc6 58. Bxd5 Nd6 59. Bf3 Kf5 60. Bc1 b5 61. Bxc6 bxc6 62. a5 1-0


I started the IDeA analysis from scratch on a faster computer and as was to be expected it also showed an evaluation of 5+ when starting from the position after 62.a5. Then I moved to the position after 57.Bg5 and it found the variation highlighted above (from 57...Nc6 to 62.a5), which is known from previous analysis (Oliver Zierke, Shirov and probably numerous other people). However, this is currently not the best line. I haven't used the interactive features of IDeA so this is all automatic analysis. I'll be busy until the weekend, but I'll post some results after that.
Parent - - By Roland Rösler (****) Date 2009-02-06 05:33
However, this is currently not the best line.

Best news! I have the same problem. :-)

I´m interested in the following two answers (shown by IDeA)

1. Bronstein was lost, when he resigns (move 57. ...)?
2. Move 42. Bd6 (?) (sealed move) was a mistake, which give away the win; but 42. Bb1 (!) was a win? Is 42. Bb1 really a win?

Garry Kasparov commented in My Great Predecessors the following moves from Bronstein as bad: 43. ... Kf6? (better Na7!=); 44. ... fxe4? (better h6!=).

PS: Many thanks for your efforts! I´m very curious.
PS2: If you become angry about the only engine :-), please keep cool! There are some others! :-)
PS3: I was very shocked, when I read in Garry´s book, what a monster Botvinnik really was (how he treated by example his second S. Flohr in this game).
Parent - - By Dadi Jonsson (Silver) Date 2009-02-06 11:06

> 1. Bronstein was lost, when he resigns (move 57. ...)?
> 2. Move 42. Bd6 (?) (sealed move) was a mistake, which give away the win; but 42. Bb1 (!) was a win? Is 42. Bb1 really a win?


I'm currently analyzing something else, but I will get to those questions when I can continue my analysis. Don't expect anything until after the weekend though.

> Garry Kasparov commented in My Great Predecessors the following moves from Bronstein as bad: 43. ... Kf6? (better Na7!=); 44. ... fxe4? (better h6!=).


I'll let you know what IDeA says about that too. I have some doubts about the quality of Kasparov's analysis in general after I wrote this article: Large-scale Analysis with Rybka Aquarium.

> PS: Many thanks for your efforts! I´m very curious.


:) I'm also looking forward to seeing what comes out of this.

> PS2: If you become angry about the only engine :-), please keep cool! There are some others! :-)


Time to Google, I guess ;)

> PS3: I was very shocked, when I read in Garry´s book, what a monster Botvinnik really was (how he treated by example his second S. Flohr in this game).


I haven't read Kasparov's book, but I heard a version of that story and I wasn't sure what to make of it.
Parent - - By CSullivan (**) Date 2009-02-12 21:32
Hi Dadi,
I'm sure your Googling has found these links already, but just in case....

This one is Karsten Müller's column regarding Shirov's find in Botvinnik-Bronstein...
http://www.chesscafe.com/text/mueller91.pdf

This one has an addendum regarding a find I made that Bronstein still had a draw as late as move 52...
http://www.chesscafe.com/text/mueller93.pdf

I do most of my analysis work with Zappa Mexico II (since it propagates in reverse better than Rybka3 and its endgame analysis seems much better) but Rybka3 analyzes in the forward direction a little better (but again, Zappa often is better is sharp, tactical situations).  I've just purchased Aquarium and I hope to begin increasing my use of it over time.  It has a lot of potential.
Parent - - By buffos (Silver) Date 2009-02-13 06:18
Zappa Mexico , will not work with IDEA , since it does not implement the "ignore moves" command in the UCI protocol (all other UCI engines do)
Parent - - By oddg (**) Date 2009-02-13 10:33

> Zappa Mexico , will not work with IDEA


No, but isn't his method much better?
This is how I have analysed since CA started to include Chess Tiger and later Shredder. You get much deeper (more excact) in a shorter time, and in an addition you learn (remember) what you analyze.
Rybka isn't any good at this, I didn't know Zappa was good to do this but I use Shredder and Gandalf and they both is pretty good.
Parent - By buffos (Silver) Date 2009-02-13 10:40

> No, but isn't his method much better?


It depends on what you are trying to do. If you are trying to build a metro-tunel of variations (like IDEA, you have to move forward)

> You get much deeper (more excact) in a shorter time


This is indeed true. Actaully since you are a CA user, you know that in Background analysis, you can retro analyse variations (which actually is an application of that rule).

By the way, it would be nice to create an IDEA-optimized engine, that would report various values (besides evals) and would be able to read write to the tree at once.
Parent - - By Dadi Jonsson (Silver) Date 2009-02-15 23:45
Finally, the results of the IDeA analysis of the Botvinnik-Bronstein game.

> 1. Bronstein was lost, when he resigns (move 57. ...)?


It seems so. The IDeA evaluation after 57.Bg5 is +2.63. I still haven't had much time to look at the variations though.

> 2. Move 42. Bd6 (?) (sealed move) was a mistake, which give away the win; but 42. Bb1 (!) was a win? Is 42. Bb1 really a win?


Yes, 42.Bd6 (+0.37) seems to give away the win, according to IDeA. Instead 42.Bb1 scores +1.91. From what I have seen it's almost impossible for Black to hold the draw after that. He can defend for a long time though and the win isn't as easy as some annotators have thought (Charles Sullivan has shown one way to win with a variation where Black defends well).

> Garry Kasparov commented in My Great Predecessors the following moves from Bronstein as bad: 43. ... Kf6? (better Na7!=); 44. ... fxe4? (better h6!=).


According to IDeA, Bronstein's 43...Kf6 (+0.44) is only slightly worse than 43...Na7 (+0.37). It doesn't seem to give the draw away.

Bronstein's 44...fxe4 (+0.44) is the best move according to IDeA and Kasparov's 44...h6 has a slightly worse evaluation (+0.60).

Here are the moves with the IDeA evaluations. If the evaluation changes, it means that a suboptimal move has been played (according to IDeA). When that happens I show the best move (again according to IDeA!) in parenthesis with its evaluation.

8/1p5p/np2p1k1/n2p1p2/P2PP3/2KB1PB1/7P/8 w - -


42.Bd6 +0.37 (42.Bb1 +1.91) 42...Nc6 +0.37 43.Bb1 +0.37

8/1p5p/npnBp1k1/3p1p2/P2PP3/2K2P2/7P/1B6 b - -


43...Kf6 +0.44 (43...Na7 +0.37) 44.Bg3 +0.44 44...fxe4 +0.44 45.fxe4 +0.44 45...h6 +0.72 (45...h5 +0.44) 46.Bf4 +0.48 (46.exd5 +0.72) 46...h5 +0.48 47.exd5 +0.48 47...exd5 +0.48 48.h4 +0.48

8/1p6/npn2k2/3p3p/P2P1B1P/2K5/8/1B6 b - -


48...Nab8 +0.64 (48...Ne7 +0.48) 49.Bg5+ +0.64 49...Kf7 +0.64 50.Bf5 +0.37 (50.Bc2 +0.64) 50...Na7 +0.37 51.Bf4 +0.37 51...Nbc6 +0.37 52.Bd3 +0.37

8/np3k2/1pn5/3p3p/P2P1B1P/2KB4/8/8 b - -


52...Nc8 +2.38 (52...Ne7 +0.37) 53.Be2 +2.38 53...Kg6 +2.38

2n5/1p6/1pn3k1/3p3p/P2P1B1P/2K5/4B3/8 w - -


54.Bd3+ +0.85 (54.Bf3 +2.38) 54...Kf6 +1.02 (54...Kg7 +0.85) 55.Be2 +1.02

2n5/1p6/1pn2k2/3p3p/P2P1B1P/2K5/4B3/8 b - -


55...Kg6 +2.38 (55...N8e7 56.Bg5+ Kg7 57.Bxe7 Nxe7 58.Bxh5 Nf5 59.Bf3 Nxh4 60.Bxd5 Nf5 61.Bxb7 Nd6 +1.02) 56.Bf3 +2.38 56...N6e7 +2.63 57.Bg5 +2.63

Bronstein resigned here. The following moves are analysis that you showed, plus some additional moves from Karsten's column (see the reference in Charles Sullivan's post in this topic).

57...Nc6 +2.88 58.Bxd5 +2.88 58...Nd6 +3.75 (58...Kf5 +2.88)

8/1p6/1pnn2k1/3B2Bp/P2P3P/2K5/8/8 w - -


59.Bf3 +2.13 (59.Bg2 +3.75) 59...Kf5 +2.13 60.Bc1 +2.13 60...b5 +3.75 (60...Ne4+ +2.13) 61.Bxc6 +3.75 61...bxc6 +3.75 62.a5 +3.75 62...Ke6 +3.75 63.Bf4 +3.75 63...Kd5 +4.50 64.a6 +3.75 64...Nc8 +3.75 65.Bg3 +3.75 65...Ke6 +3.75

Most of the analysis produced by IDeA confirms the findings of Charles Sullivan (and refutes some older analysis, including Kasparov's). It would be interesting if someone had a look at the improvement suggested by IDeA on Black's 55th move: 55...N8e7 instead of Bronstein's 55...Kg6. If the analysis is correct it seems to me that Black may have had a draw there. If that is the case, then Botvinnik's repetition of moves was a mistake that could have cost him the title.

If you have any questions about the IDeA analysis or want to see some of the variations just let me know. I should have more time tomorrow :)
Parent - - By ernest (****) Date 2009-02-15 23:56
Can you tell us something about how you obtained these results:
* What IDeA time and/or depth did you use
* How much interactivity did you put in it
* What was the total duration of this IDeA analysis
* What would you improve if you had to do it again (not knowing the results you obtained) :-)
Parent - - By Dadi Jonsson (Silver) Date 2009-02-16 11:40

> * What IDeA time and/or depth did you use


I used Rybka 3 at 30 sec/position (30s AND d=1; v=5). I have tested this setting on many positions. Partially because I think it is a fairly good setting (for my configuration), but also because I am interested in seeing if there are practical drawbacks for using a time-based -- rather than depth-based --settings for Rybka in IDeA.

I think that 30s is rather "conservative" and I could probably get good results by choosing a much faster setting. However, this suits me fine since I want IDeA to do its magic with as little interference from me as possible, and I get a tree with a relatively high proportion of interesting positions. I usually run IDeA overnight and sometimes for several days and I don't want to spend more than 30 minutes/day examining the results and using the interactive IDeA features. Finally, I would like to point out that a few months ago I thought that 30s was too fast for "serious" analysis and 2-5 minutes were necessary to get fairly reliable results. However, after closely examining the analysis of many positions, I saw that a greater number of analyzed positions makes up for the deeper analysis of each position.

I guess I'm pointing out that the question "What are the best settings for IDeA?" is not only a technical question and it doesn't have a simple answer. Besides depending on the position (as Eastendboy and others have pointed out) and the power of your computer it also depends on how you want to use IDeA. I also think that strong chessplayers need different settings from weak players.

> * How much interactivity did you put in it


I try to spend no more than 30 mins. on the interactive features for each day that I run IDeA on a particular position. How much time I actually spend is not always entirely clear, however :) These days, for instance, I often work at the computer where IDeA is running or I have a Remote Desktop Connection to it. I sometimes switch to Aquarium, color some moves and switch back to whatever I was doing. Then I may check back in an hour, change or remove the coloring, add alternatives etc. At other times I take one session in the morning and another one in the evening. Sometimes I may leave it running for a couple of days without touching anything, but of course it is more fun when you have some time to work on the analysis.

Sometimes I take longer sessions. This happened yesterday, for instance, when I saw that Black possibly had a draw in the Botvinnik-Bronstein game after 55..N8e7. I had deliberately skipped over the analysis of the move repetition as I thought it wasn't interesting. When I was about to post the results I noticed that this didn't look "right." After all, this is one of the most famous games in the history of chess and the game had been declared lost long before this move by all annotators, including Kasparov. I thought that IDeA simply hadn't analyzed the position deep enough so I spent an hour experimenting with move coloring and alternatives. Of course it is possible that the draw doesn't hold, but the point is that when I run into something unexpected or interesting, I take the opportunity to have some fun :)

>* What was the total duration of this IDeA analysis


My original plan was to run IDeA for a couple of days. Then I got too busy with other things, so I couldn't post the results. I just left IDeA running and in the end I had about 10 days worth of analysis(!) Note that I analyzed all positions in the game from move 42 to 57 and then up to move 65 in the analysis shown by Roland Rösler above and Karsten Müller on ChessCafe.

This game is surprisingly complex and has been analyzed by so thoroughly that afterwards I'm quite happy that I ended up doing such extensive analysis.

> * What would you improve if you had to do it again (not knowing the results you obtained) :-)


I don't know how much of an improvement it would be, but here is what I would like to try:

1) Based on what I found out using my slow laptop at 10 seconds/move it seems that it would be interesting to rerun the analysis using much shorter time/pos.

2) Do the same analysis using Shredder or some other engine that is supposedly strong in the endgame (and can be used in IDeA).

Since I hadn't planned to run the analysis for this long, I wasn't very organized in selecting the order in which I analyzed the positions. Basically, I just selected the next position which I found interesting. Perhaps it would be better to start at the final position and the work your way back to the first position as I did when I analyzed Fischer-Geller, Skopje 1967.
Parent - By ernest (****) Date 2009-02-16 16:46
Thanks, Dadi. Very informative!
Parent - - By CSullivan (**) Date 2009-02-16 02:28
Dadi,
It will be sensational news if Bronstein had a draw as late as move 55!  After 55...N8e7 56.Bg5+ I had examined 56...Kg6 (and eventually convinced myself that Botvinnik still had a narrow win), but not 56...Kg7.  Now I will be spending the next few days looking at that!  Please let us know the setting you used to examine the game.
Thanks for your very interesting post.
Parent - - By Dadi Jonsson (Silver) Date 2009-02-16 11:56

> It will be sensational news if Bronstein had a draw as late as move 55!


I agree, but let's analyze some more. This "simple" endgame is surprisingly complex and there may still be a win for White.

> I had examined 56...Kg6 (and eventually convinced myself that Botvinnik still had a narrow win), but not 56...Kg7.


56...Kg6 is of course the natural reaction to 56.Bg5+, but 56...Kg7 is like a move from a study.

> Now I will be spending the next few days looking at that!


I guess I'm forced to do that too :)

> Thanks for your very interesting post.


Thanks for your interesting analysis of this game :)
Parent - - By CSullivan (**) Date 2009-02-16 13:41
Hi Dadi,
After looking at Botvinnik's notes -- he did not comment upon move 55, but he did make one at move 56 -- and looking at my previous analysis, I'm pretty sure that White needs to keep both bishops on the board (else he is in danger of ending up with a bishop of the wrong color and/or not having enough mating material), so White's 57.Bxe7 must not be optimal.  Also, preventing Black's pawn move to b5-square is a key.  Right now I'm looking at 57.Bf3 (looks good) and 57.Bxh5 (not as good).  Perhaps White needs to keep both bishops and pawns on both wings to win this.  And, of course, Zugzwang -- the bugaboo for chess programs! -- plays a very large role in this drama.  I'll keep you posted.
Parent - - By Dadi Jonsson (Silver) Date 2009-02-16 14:25
Sounds logical. 56.Bf3 and 56.Bxh5 also need to be analyzed, but so far I haven't found anything there. A pretty variation came up after 56.Bc7:

8/1pB1n3/1pn2k2/3p3p/P2P3P/2K5/4B3/8 b - -


56....Nf5 57.Bxh5

8/1pB5/1pn2k2/3p1n1B/P2P3P/2K5/8/8 b - -


57...b5 58.axb5 Ncxd4 59.Bd8+ Ke5 60.b6

3B4/1p6/1P6/3pkn1B/3n3P/2K5/8/8 b - -


60...Ne6 61.Bc7+ Kf6 62.Bf3

8/1pB5/1P2nk2/3p1n2/7P/2K2B2/8/8 b - -


62...Nfd4 63.Bxd5

8/1pB5/1P2nk2/3B4/3n3P/2K5/8/8 b - -


63...Nxc7! 64.Kxd4 (64.bxc7 Nb5+)

8/1pn5/1P3k2/3B4/3K3P/8/8/8 b - -


64...Na8!

n7/1p6/1P3k2/3B4/3K3P/8/8/8 w - -


Moving the knight into the corner secures the draw. He will take the b-pawn on the next move and the bishop is of the wrong color to assist the h-pawn :)
Parent - - By CSullivan (**) Date 2009-02-18 16:17
Yes, a very pretty variation.  Before we let this variation go, here is what I think I have found...
After 55...N8e7 56.Bg5+ Kg7 I believe that 57.Bf3 wins (at least, I have not found a draw for Black).  Here is a variation with a witty finish: 57...Nf5 (Obviously not the toughest defense, but all moves fail -- so let's choose one that allows a finish that we can understand!) 58.Bxd5 Ncxd4 59.Bd8 Ne2+ 60.Kb3 Nd6 61.Bxb6 Nf4 62.Bf3 Kf6 63.a5 Nd3 64.Bxh5 Kf5 65.Bd1 Ne5 66.h5 Ndf7 67.Bc5 Nd7 68.Bd4 Nde5 69.Bxe5 Nxe5 70.Bc2+ Kg5 71.Be4 Nd7
8/1p1n4/8/P5kP/4B3/1K6/8/8 w - - 4 72

72.Bxb7! Nc5+ 73.Kb4 Nxb7 74.a6
8/1n6/P7/6kP/1K6/8/8/8 b - - 0 74

and the pawn queens.
[Dadi:  After 55...N8e7 56.Bg5+ Kg7 57.Bf3 have you found a drawing line?]
Parent - By CSullivan (**) Date 2009-02-19 23:56
Correction: Upon further analysis, after 55...N8e7 56.Bg5+ Kg7 57.Bf3 Nf5 58.Bxd5 Ncxd4 59.Bd8 Ne2+, the strongest move is 60.Kb4 (rather than 60.Kb3 as in my post just above).  For example: 60...Nf4 61.Bxb7 Nd6 62.Bc6 Nd3+ 63.Kc3 Ne5 64.Bd5 and Black is lost.  The finish might be 64...Nc8 65.Kd4 Ng6 66.Be4 Nh8 67.Kd5 Nf7 68.Bc7 Kf6 69.Kc6 Ne5+ 70.Kb7 Ne7 71.Kxb6 etc.
Parent - - By Dadi Jonsson (Silver) Date 2009-02-22 10:10

> Dadi:  After 55...N8e7 56.Bg5+ Kg7 57.Bf3 have you found a drawing line?


The best line I have found so far starts with 57...Kg6. I am not sure how White should continue, but here is an example:

57...Kg6 58.Bf4 Kf5 59.Bc7 Ke6 60.Bxh5 Nf5 61.Bg4 Ncxd4 62.h5 Kf6 63.Bxb6 Nc6 64.Bf3 Ke6

Is this a draw? I'm not sure, and maybe this variation can be improved earlier on.
Parent - - By CSullivan (**) Date 2009-02-23 03:56
Dadi,
After 55...N8e7 56.Bg5+ Kg7 57.Bf3 Kg6 58.Bf4 Kf5 59.Bc7 Ke6 60.Bxh5 Nf5 61.Bg4 Ncxd4, I think White should vary with 62.Bxb6 Ke5 63.h5 Ne6 64.a5 Kf6! 65.Bf3 Ne7 66.Ke2 Kf7 67.Kd3 Kf6 68.Ke3 Kf5 (+1.05) but this is drawish.
But the real improvement is 58.Bh1.  I have examined every reasonable try (I think) for Black in this position and have found that White always wins; one of the tries for Black is 58...Nf5 59.Bxd5 and then 59...Nd6 60.Bg2 Kf5 61.d5 Ne5 62.Kd4 Nef7 63.Bh3+ Kg6 64.Bf4 b5 65.a5 b4 66.Kc5 b3 67.Bc1 Kf6 68.Bb2+ Ke7 69.Kb4 NBe4 70.Kxb3 Kd6 and Zappa rates this winning (+3.37) for White.
Parent - - By Roland Rösler (****) Date 2009-02-23 15:51
Can you show the win after 59. ... Nfxd4?
Here is my line: 60. Be4 Nf5 61. Be3 Kf6 62. Bxb6 Ke6 63. a5 Nfe7 64. Kc4 Kd6 65. Kb5 Nd5 66. Bxd5 Kxd5 67. Bc7 Nd4 68. Kb6 Ne6 69. Kxb7 Nc5
8/1KB5/8/P1nk3p/7P/8/8/8 w - - 0 70


I think it´s a draw!
Parent - - By CSullivan (**) Date 2009-02-23 20:34
Hi Roland,
Indeed, after 55...N8e7 56.Bg5+ Kg7 57.Bf3 Kg6 58.Bh1 Nf5 59.Bxd5 Nfxd4, your line with 60.Be4 looks pretty drawish.  I had in mind 60.Be3 with 3 main variations:
(1) 60...Ne2+ 61.Kc4 b5+ (if 61...Ne5+ Zappa gives 62.Kb4 as +4.06: 62...b5 63.axb5 Kf6 64.Bxb7 etc.) 62.axb5 Ne5+ 63.Kb4 b6 64.Bxb6 and wins
(2) 60...Nf5 61.Bxb6 Nfe7 62.Bg2 Kf6 63.Bc5 Nd8 64.Kb4 Nec6+ (if 64...Nc8 65.Bf3 Kg6 66.Kb5 Kf6 67.Bxh5 wins) 65.Kb5 Ne5 66.Bd4 Ndf7 67.Bxb7 Ke7 68.a5 wins
(3) 60...b5 61.Bxd4 bxa4 62.Bxc6 bxc6 and Zappa is already calling this a mate-in-30 (approximately)

(I hope I have copied the moves correctly!)

If you need clarification on one of these lines, please ask.  I have given only the main variations.  I have many, many more.
Parent - By Dadi Jonsson (Silver) Date 2009-02-23 20:48
This is also what I have: 60.Be3 etc. +4.50.
Parent - By Roland Rösler (****) Date 2009-02-24 00:35
Many thanks for 60. Le3!! and line (2).

Now I feel pretty fine in the line 55...N8e7 56.Bg5+ Kg7 57.Bf3 Kg6. I hope it was my last hole in this line! :-)
Some days ago I thought I´m ready with this line but then decided to have a deeper look to 59. Nxfd4.
Parent - - By Dadi Jonsson (Silver) Date 2009-02-23 20:44

> But the real improvement is 58.Bh1.


Very nice! This is an innocent looking, but poisonous little move! I think it will prove hard, and most likely impossible, to find a way to save Black after this. IDeA's evaluation is +3.25.

IDeA is a beast with a thousand arms and for every arm you cut off it grows ten new ones :) Here is the next challenge in this variation:

55...N8e7 56.Bg5+ Ke6!?

8/1p2n3/1pn1k3/3p2Bp/P2P3P/2K5/4B3/8 w - -


How should White continue? 57.Bxe7 (does it work any better here when the king is further away from the h-pawn?), 57.Bxh5 (the most natural looking move) and the old faithful 57.Bf3 come into consideration. Here is one variation with IDeA evaluation:

55... N8e7 56. Bg5+ Ke6 57. Bxe7 Nxe7 58. Bxh5 b5 59. Bg4+ Kf7 60. axb5 b6 {+1.46} In this endgame an evaluation of +1.5 doesn't necessarily mean that the position is won.

8/4nk2/1p6/1P1p4/3P2BP/2K5/8/8 w - -

Here it seems that White will not be able to make any progress.

I'll post more variations later.
Parent - - By CSullivan (**) Date 2009-02-23 20:57
I like 56...Ke6 (and I'll spend a little time seeing if I find a potential win for White), but really! isn't one good draw (55...Kf5!) enough! :-)
Parent - By Roland Rösler (****) Date 2009-02-24 01:29
..., but really! isn't one good draw (55...Kf5!) enough! :-)

I don´t think so! If you can show there are two moves (55. N8e7 and 55. Kf5; or even more?) for a draw in move 55 for black, you punish the further analysts even more! :-) And you have a very good market place for IDeA! :-)
But you must be sure! Quantity of the (maybe) drawing moves can´t replace the quality of the lines.

PS: Today I think the weaknesses of Rybka in endgame play aren´t bugs but features for IDeA! :-)
Parent - By Roland Rösler (****) Date 2009-02-24 00:41
IDeA is a beast with a thousand arms and for every arm you cut off it grows ten new ones :-)

Let´s call her the real Hydra! :-)
Parent - - By Roland Rösler (****) Date 2009-02-16 17:02
Many thanks for the results. Great stuff! Now I need some time for good questions. :-)

But some questions ex ante:

1. What´s the machine, you do your analysis? (Quad or Octo?; ~kn/s? in these positions)
2. What´s the best move for black (+2.63) after 57. Bg5? (57. ... b5?)
3. Do you have any doubts analyzing these positions with R3 knowing of the wrong bishop bug and mp egtb bug? (I have bellyache about it! :-))

PS: I´m rather sure in your line 57. Bf3 (not 57. Bxe7?) is a win. Okay, we will see ...
Parent - - By Dadi Jonsson (Silver) Date 2009-02-17 22:13 Edited 2009-02-17 22:53

> 1. What´s the machine, you do your analysis? (Quad or Octo?; ~kn/s? in these positions)


Octal. I haven't run infinite analysis on this game, but let me do that now.... On move 38 I get 230kn/s.

> 2. What´s the best move for black (+2.63) after 57. Bg5? (57. ... b5?)


57...Nf5 has a slightly better evaluation than 57...Nc6. The difference is too small to be significant (IMO).

> 3. Do you have any doubts analyzing these positions with R3 knowing of the wrong bishop bug and mp egtb bug? (I have bellyache about it! :-))


No, I don't see that as a practical problem as long as you keep it in mind. Nevertheless, it would be interesting to repeat the analysis with another engine to see if I'm right :)

> PS: I´m rather sure in your line 57. Bf3 (not 57. Bxe7?) is a win. Okay, we will see ...


I agree that it is a better move and I wouldn't be surprised if it leads to a win, but as you say: we will see :)

After looking over the IDeA variations I see that a chessplayer who analyzes this endgame deeply will learn a lot. I've already posted one interesting variation. In another variation I saw an endgame with two bishops vs. a knight and yet another variation ended with a queen vs. two knights, which is even more interesting.

Here is a different and more aggressive approach to this endgame by Black. Do you think that White can win? I only give one line and let you suggest the improvements :)

2n5/1p6/1pn2k2/3p3p/P2P1B1P/2K5/4B3/8 b - -


Now, instead of the rather defensive 55...N8e7 (or Bronstein's 55...Kg6?!) it's interesting to see what happens if Black gives the h-pawn and even the b6-pawn too, for a more aggressive position: 55...Kf5 56.Bc7 Ke4 This is the idea. White can't prevent the black king from reaching the center of the board. Black gets the d4-pawn in exchange for h5, but will the king be too far away to stop White's passed h-pawn? 57.Bxh5 Nxd4 58.Bg6+ Nf5

2n5/1pB5/1p4B1/3p1n2/P3k2P/2K5/8/8 w - -


An uncomfortable pin, but there is nothing else that Black can do. Notice that now the black king is stuck in the center of the board and because of the bishop on c7 he can't approach the h-pawn. Naturally White pushes the pawn. 59.h5 d4+ Although Black would like to have a fast route towards the h-pawn, there are also advantages to having the king in the center! 60.Kd2 Nce7 61.Bh7 White wants to preserve the bishop pair. 61...Kd5 Finally the king can start its journey towards the h-pawn, but only after briefly visiting the queenside :) 62.Kd3 b5 As Charles Sullivan pointed out White wants to prevent this move, but can he do that in this variation? 63.axb5 Ke6 64.Bf4 Kf7 65.Bxf5 Another small victory for Black. Did White have a better move? 65...Nxf5 66.Be5 Ke6 67.Bxd4

8/1p6/4k3/1P3n1P/3B4/3K4/8/8 b - -


And this position is a draw... or is it? White's two pawns are on the opposite wings so having the bishop is a nice bonus on top of being a pawn up. One thing is sure: If you are a chessplayer you will learn something about endgame play if you study this position.
Parent - - By CSullivan (**) Date 2009-02-18 16:41
Dadi,
You have outdone yourself this time!  You gave us a teaser with 55...N8e7 56.Bg5+ Kg7, but now 55...Kf5 looks like the real deal.  After 56.Bc7 Ke4 57.Bh5 Nd4 58.Bg6 Nf5 59.h5 d4+ 60.Kd2 Nce7 61.Bh7 Kd5 62.Kd3 it looks as if 62...Ke6 is the best move, but then White maintains good winning chances.  [Does it look like a win to you?]  But 62...b5!! is absolutely brilliant.  I've spent some time looking at your continuation of 63.axb5 Ke6 64.Bf4 Kf7 65.Bxf5 Nxf5 66.Be5 Ke6 67.Bd4 but cannot find an improvement for White.  And, as you point out, White's worst fears have occurred -- he no longer has The Two Bishops and Black has managed to play ...b5!
In the final position
8/1p6/4k3/1P3n1P/3B4/3K4/8/8 b - - 0 67

you're teasing us again!  Zappa Mexico II spent about 3 hours before discovering its top choice leads to a Black defeat!  Only then did it find the drawing method. 

All in all, truly one of the great discoveries!  Congratulations, and I hope you will be publishing an article about this ASAP!
Parent - - By Dadi Jonsson (Silver) Date 2009-02-22 10:17
Thanks.

> 55...Kf5 looks like the real deal.


I think it's quite possible, although as Roland points out other moves besides 56.Bc7 need to b analyzed.

> Zappa Mexico II spent about 3 hours before discovering its top choice leads to a Black defeat!  Only then did it find the drawing method. 


Yes, it is a draw! The only move that draws in the position you show in your post is 67...Kf7!! Black must play both pieces to the eighth rank in order to save the game (Nc8 and Kg8-h7). A very nice defence.

> I hope you will be publishing an article about this ASAP!


Yes, probably, unless someone comes up with an improvement for White.
Parent - - By CSullivan (**) Date 2009-03-14 15:06
Hi Dadi,
I've emailed Karsten Müller in the last few days and mentioned, in passing, that you have found a draw for Black at move 55.  He is very interested!
Parent - - By Dadi Jonsson (Silver) Date 2009-03-16 22:02
I wouldn't mind if he had a look at what I've got, although I prefer "possibly found a draw" ;) Next I'll have a closer look at the final position. It looks like there are some interesting possibilities there.
Parent - By CSullivan (**) Date 2009-03-17 03:40
I just sent Karsten a link to this topic. 
Parent - - By Roland Rösler (****) Date 2009-02-19 04:48
He Dadi, I see you have a very nice time in Iceland (although I read yesterday a very nice article from Einar Mar Gudmundsson in German newspaper (SZ); not all habitants of Iceland are so happy these days).
I´m working on your last line (55. ... N8e7 56. Bg5 Kg7). I was sure that after 57. Bf3 Kg6 (?) is losing, but have a hard time with 57. ... Kf8 (58. Bh5 isn´t good!). I have to work on it!
Now you come up with a new puzzle! Maybe your last position is a draw after 67. ... Kf7 (with 5-man and R3 I get ~+4.85 for Kd5 or Ke7). But I don´t like your first white move (56. Bc7). There must be something better (okay, I can´t find it yet). But it is to show, that white can´t defend all three pawns with some manouver (bishops at most on the rank 4 or 5) and comes up with a good position.

Last remark:
With octal, 6 man and IDeA you have a nice time. Every day you look for 30 minutes on the tree (created by IDeA and R3 with 20 billion nodes per day; 20 bN/d :-)) and give us some puzzles. Okay, this isn´t bad but unfair! :-). Please give us the best line you have found in move 55 yet!
Parent - By Dadi Jonsson (Silver) Date 2009-02-22 11:32 Edited 2009-02-22 12:43

> He Dadi, I see you have a very nice time in Iceland (although I read yesterday a very nice article from Einar Mar Gudmundsson in German newspaper (SZ); not all habitants of Iceland are so happy these days).


You are right. However, I am one of the 'lucky' ones who were not directly hurt by the collapse. Incidentally, my investments are handled by a bank that was founded by a GM (a former top-100 player, IIRC) a few years ago after he stopped playing. While the banks of the world were collapsing he managed to turn a very healthy profit last year. He clearly analyzed the situation deeper than many others did. Notice that I succeeded in turning this into a chess-related discussion :)

> I´m working on your last line (55. ... N8e7 56. Bg5 Kg7). I was sure that after 57. Bf3 Kg6 (?) is losing, but have a hard time with 57. ... Kf8 (58. Bh5 isn´t good!). I have to work on it!


What was your continuation after 55. ... N8e7 56. Bg5 Kg7 57. Bf3 Kg6? I have very little analysis after 57...Kf8, but 58.Bh6+ is currently at the top of the list (but only with 8 positions).

> Now you come up with a new puzzle! Maybe your last position is a draw after 67. ... Kf7


Yes, that is the (only) drawing move :)

> But I don´t like your first white move (56. Bc7). There must be something better (okay, I can´t find it yet). But it is to show, that white can´t defend all three pawns with some manouver (bishops at most on the rank 4 or 5) and comes up with a good position.


Yes, other moves definitely need to be analyzed. Here are my suggestions:

1) If the bishop leaves the h2-b8 diagonal, Black plays Nd6 and b5, e.g. 56.Bg5 Nd6 57.Bxh5 b5 etc. I think that Black will equalize in all cases.

2) 56.Bg3. This move gives Black a tempo when he plays his knight to f5: 56.Bg3 Ke4 57.Bxh5 Nxd4 58.Bd1 Ne7 59.h5 Nef5 60.Bf2 Nc6 61.Bxb6 d4+ 62.Kd2 d3. I haven't found a win for White here.

3) 56.Bh2. In this case I think it may be best to play 56...Kg6 instead of 56...Ke4. Here is one variation (I haven't really looked it over, so it may be rubbish :)): 57.Bf3 b5 (57...N6e7 also comes into consideration and is probably better) 58.axb5 N6e7 59.Bf4 b6 60.Bg5 Ng8 61.Bxd5 Nf6. 62.Bc6 Kf5. How can White win here?

Both 2) and 3) need more analysis.

> Okay, this isn´t bad but unfair! :-).


Maybe, but I give all my analysis to you. Take a look at my analysis of Fischer-Geller, Skopje 1967. Every single position I analyzed is available for download for you. Has anyone else given you such detailed analysis :) That's one of the advantages of IDeA.

Btw. you asked if I think that IDeA is any good for endgame analysis. Now I ask you the same question :)

> Please give us the best line you have found in move 55 yet!


See above.
Parent - By Zadig (**) Date 2009-02-12 21:35
6-man TBS on SSD ?

Wow, you've got a puresilicon Nitro 1 TB SSD prototype? How did you find it?

:-)
Parent - - By Rowlando (***) Date 2009-02-04 08:07 Edited 2009-02-04 08:12
Fascinating position! 

If possible, can you attach/upload your IDeA tree?  I have some ideas that I'd like to test and using your tree as a starting point may save me some time....
Parent - By Dadi Jonsson (Silver) Date 2009-02-05 09:26 Edited 2009-02-05 10:29
The IDeA analysis tree for Pachman-Hromadka.
Attachment: PachmanHromadka-30s.rar (7k)
Parent - - By Mark (****) Date 2009-02-06 02:56
So, is it pretty certain that this position is a draw?  It's amazing how complex this position is with only 8 pieces!
Parent - By Dadi Jonsson (Silver) Date 2009-02-06 10:50

> So, is it pretty certain that this position is a draw?


Yes.

> It's amazing how complex this position is with only 8 pieces!


True. Chess is a very complex game.
Up Topic Rybka Support & Discussion / Aquarium / Pachman - Hromadka, Prague Ch 1944

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill