Not logged inRybka Chess Community Forum
Up Topic The Rybka Lounge / Test Positions / My latest (beta 5) hard set : EPD + PGN
- - By Vinvin (***) [be] Date 2016-06-25 21:25
The set is updated : some bad positions removed + some added.

Now : 153 best moves + 2 moves to avoid + 6 eval to check
3 categories separated by a blank line in the EPD.
The games in the PGN are not in the same order as in the EPD yet.

EPD here : http://home.scarlet.be/vincentlejeune/chess/hard2016beta5.epd
PGN here : http://home.scarlet.be/vincentlejeune/chess/hard2016pgnbeta5.pgn
Parent - - By user923005 (****) [us] Date 2016-07-19 02:15
This archive:
https://www.dropbox.com/s/zoskszeny3fy12f/vljhard.7z?dl=0
Contains my current solution set for those problems.

A couple are actually draws.

For the following set, I have no convincing mechanical solutions (as produced by engines alone):
'1b6/2p5/1kB5/1PN5/8/3K4/8/q7 w - -',
'8/8/5k2/8/2B4p/5RbP/5pP1/7K b - -',
'7q/b1p5/1p1Npkb1/pPP2ppP/P1P5/3B2P1/5P1R/K3R3 w - -',
'1nrqr3/1b4bk/pp3pp1/3pPp1p/P1pP1P1P/2P1BN2/1PB2QP1/R4RK1 w - -',
'r4r1k/1p4qp/b1pNRbp1/2P5/3p1P1P/PQ4P1/3R2BK/8 w - -',
'4bbk1/3nq2r/pr2p3/Np1p1p1p/1PpP1Pp1/2P1P1P1/PQ1NB1PR/2K4R b - -',
'r2qnrk1/p4pp1/1p2b2p/2p1Pp2/2P5/PnPB1QN1/6PP/R1B2RK1 w - -',
'1r6/4k3/r2p2p1/2pR1p1p/2P1pP1P/pPK1P1P1/P7/1B6 b - -'

The archive also contains a spreadsheet for the above set of open problems with analysis by a number of strong engines.
The headings for the spreadsheet are:
ace, apv, aacd, abm, apm, acd, ce, pv, bm, pm, c0,c1,c2,c3,c4,acs
of which acd, ce, pv, bm, pm, c0,c1,c2,c3,c4 and acs all have the standard EPD meanings.
and :
ace = alternative ce as calculated by a strong chess engine
apv = alternative pv as calculated by a strong chess engine
aacd = alternative acd as calculated by a strong chess engine
abm = alternative bm as calculated by a strong chess engine
apm = alternative pm as calculated by a strong chess engine

It seems clear that several of the above positions have fortress draws not included in the original solutions, but I am not sure that the original solution is not supposed to be a win.
All of the positions have been examined by current Stockfish and Asmfish, by Houdini 4 Tactical and Houdini 4 and by Komodo 10 for at least an hour per position (sometimes more).
Parent - - By Vinvin (***) [be] Date 2016-07-21 00:01
I begin to have doubts about this one :

[Event "?"]
[Site "?"]
[White "W"]
[Black "B"]
[Result "*"]
[FEN "r2qnrk1/p4pp1/1p2b2p/2p1Pp2/2P5/PnPB1QN1/6PP/R1B2RK1 w - - 0 1"]
[Comment "bm Bxh6; c0 \"Domangala-Leszek, Super Puchar cr 2003\";"]

1. Bxh6 Nxa1 2. Bxf5 Nb3 3. Bxe6 gxh6 4. Ne4 Qh4 5. Bxf7+ *

Bxf5 seems stronger than Bxh6

For others positions, the key still correct :-)
Parent - By Vinvin (***) [be] Date 2016-07-24 16:24
I removed this position from the set.
Parent - - By Dadi Jonsson (Silver) [is] Date 2016-11-06 10:11
I used some of these positions as a test for a new Aquarium 2016 feature (Tree Generation Strategies). It is a part of the IDeA toolset and an excellent tool for checking existing analysis from books, magazines, test sets and of course your own analysis. This method was, for instance, used when upgrading CT-ART to version 6. One of the results is that now there is often more than one key move allowed (and scored accordingly) if there are several good moves in a position.

I like this test set because it contains many difficult positions that have been thoroughly analyzed by a good analyst and many people have contributed to the analysis over a long period of time. Another sign of quality of this test set is that it shows the mainline and sometimes quite detailed analysis of the positions. This is actually what made it such an interesting target for this new IDeA feature.

Now that I look at the results again after a couple of months, I think they might be of interest to others. I could give more detailed information about the analysis of each position, but that will have to wait.

One of the questions that the results of my analysis raises is what criteria are used for including a position in a test set? It seems to me that normally such test sets consist of positions where there is one and only one winning move, but that is definitely not the case for this test set. Perhaps those who compile such test sets should specify their criteria.
Parent - - By Dadi Jonsson (Silver) [is] Date 2016-11-06 10:12
The first position in the PGN file is clearly winning and there is only one winning move. However, the second position already seems dubious to me.

2r2rk1/2q2p1p/p1p1nPpQ/3p4/4p1B1/8/PPP3PP/R4R1K w - - 0 1


There are several moves that are clearly winning and besides the proposed solution (1.Rf5), a second move (1.a4) also leads to checkmate, but in a few more moves. The following screen shot shows the top 3 moves.
Parent - By Vinvin (***) [be] Date 2016-11-25 13:05
Rf5 is clearly the sortest mate. Some engines have trouble to find mates.
Parent - - By Dadi Jonsson (Silver) [is] Date 2016-11-06 10:13
r1b3kr/pp1n2Bp/2pb2q1/3p3N/3P4/2P2Q2/P1P3PP/4RRK1 w - -


The third position also has several clearly winning moves. In addition to the intended solution (1.Re5), another move also leads to a checkmate (1.Bh6). The third move scores +31.03.
Parent - By Vinvin (***) [be] Date 2016-11-25 13:05
Re5 is clearly the sortest mate. Some engines have trouble to find mates.
Parent - - By Dadi Jonsson (Silver) [is] Date 2016-11-06 10:13
1qrrbbk1/1p1nnppp/p3p3/4P3/2P5/1PN1N3/PB2Q1PP/1B2RR1K w - -


In the fourth position the given solution scores highest in my analysis (1.Bxh7+; 7.00), but other moves are also winning, such as 1.Ne4 (+3.17).
Parent - By Vinvin (***) [be] Date 2016-11-25 13:06
No doubt Bxh7+ is clearly better than Ne4.
Parent - By Dadi Jonsson (Silver) [is] Date 2016-11-06 10:14
1qrrbbk1/1p1nnppp/p3p3/4P3/2P5/1PN1N3/PB2Q1PP/1B2RR1K w - -


In the fifth position, the score of the given solution (1.Nf5; 2.75) and a second move (1.Ne7+; 2.07) are too close to call one a success and the other one a failure. This applies to many other positions in this test set as already seen in these first few positions (and a few more that I also analyzed). In general the move given as the solution scored highest (although there were exception). I hope these few examples give potential users of this test set some insight into the nature of the positions and the given solutions.
Up Topic The Rybka Lounge / Test Positions / My latest (beta 5) hard set : EPD + PGN

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill