wsogpr wrote:Hi Robert,
I'm afraid I do not know how to cut from Fritz 11 GUI and post my analysis here (despite spending an age trying ). So I have had to write out everything long hand (I was lazy enough to find a position to use as an example on chessok opening explorer so hopefully the FEN is correct).
An example of the kind of position that I mean is:
r1bqkb1r/pp1n2pp/2n1pp2/3pP3/3P1N2/3B4/PP1N1PPP/R1BQK2R b KQkq - 1 9
PGN: 1. e4 e6 2. d4 d5 3. Nd2 Nf6 4. e5 Nfd7 5. Bd3 c5 6. c3 Nc6 7. Ne2 cxd4 8. cxd4 f6 9. Nf4 * {[*]
Hugebase database of human games shows that the most popular moves here for humans are 9..Nd4 (663 games played with a success rate of 52%) while Q...e7 is significantly less popular (102 games with a success rate of 40%). Thus it is fair to say that in human games 9...Nd4 is somewhat better.
I analysed this position with 2 cores for 1 minute with the following engines.
In this position default Rybka 4 dislikes the exchange sac associated with playing 9..Nd4 and instead choses the less natural move Qe7 (at depth 16 assessing it as +0.31).
I also analysed this position with Rybka 4 with the weights changed as per the recommendation by Larry Kaufman to create a more human like evaluation (modifications to Rybka 4 default weights as follows +0.24cpawns for B/N, +0.12 cp for Rook - I have assumed that this may have been to achieve approximately the piece values that I have quoted previously (which came from Kaufman) - though Kaufman used several positions to determine what the most appropriate modifications to the default values should be - I don't think that Vas told him what they were.) This Rybka 4 human chose the much more popular and more successful 9...Nd4 (+0.18 at depth 16). It assesses 9...Qe7 as +0.34(at depth 16)
I also used Houdini 1.5 to analyse this position. It again chose 9..Qe7 (+0.2 depth 18).9...Nd4 is evaluated as +0.29(depth 20)
It is clear from the above that the modifications to the piece values make the analyis more useful for a human over the board player.
If anyone would like to educate me as to how to cut analysis from Fritz 11 GUI and set up board positions I would appreciate it!
Regards,
Sean
To paste the FEN notation, insert the [d] before the fen, therefore, "[d] r1bqkb1r/pp1n2pp/2n1pp2/3pP3/3P1N2/3B4/PP1N1PPP/R1BQK2R b KQkq - 1 9"
[d]r1bqkb1r/pp1n2pp/2n1pp2/3pP3/3P1N2/3B4/PP1N1PPP/R1BQK2R b KQkq - 1 9
To paste the analysis in Fritz GUI, right-click the analysis window and select Clip Analysis. And paste your analysis in the place you want by hitting CTRL-V.
New game
r1bqkb1r/pp1n2pp/2n1pp2/3pP3/3P1N2/3B4/PP1N1PPP/R1BQK2R b KQkq - 0 1
It is interesting to see how the games progress, and how the evals may change as the moves are played. This reminds us not to always take those numbers as the absolute truth (which I must admit I do at times!).
I tried pasting into Arena and it did not work (version 2.5). Maybe all the extra info confuses it. If you paste it to a text file and then try to open the file, it might work better. Maybe a program in the middle could convert for you.
Sean, I have several thousand unique games in this position. I don't give counts of how many I have, but suffice to say it is multiples of what you cited. In order of popularity and playability in my book:
and ten other moves with less than 30 games played, none of which are worth discussing. I should mention that overall this position is unusual in that its aggregate draw-rate is 13% lower than the whole book. You seldom see that with a position that has been played this often.
This position illustrates the weakness of conventional opening books, even ones this large. 9...Qe7 10.Nf3 fxe5 11.dxe5 puts black in a 25.5% position. 9...fxe5 10.Nxe6 Qb6 11.O-O leaves black in a 26.9% position struggling to find a draw. Such results aren't determinative but they aren't where you want to be.
So clearly the theory line of 9...Nxd4 10.Qh5+ Ke7 11.exf6+ Nxf6 12.Ng6+ hxg6 13.Qh8 Kf7 14.O-O e5 15.Nb3 Nxb3 16.axb3 (Logdahl-Rylander, 1999) is to be preferred. From there black has to choose between the fighting 16...Bf5 or the more drawish 16...Be6. I haven't done any interative analysis; that's another approach where others can add their two cents.
Thanks for that Laurence !
My education is nearly complete...now if only I could get all my papers complete by the deadline next week!...perhaps a little less chess chat reading and discussion from me for a bit!
Hi Nelson,
Thanks for that! Just out of interest is your book based on human games?
I think that it may well be that in computer play the outcome is a little different. This position looks fine to me for black after the exchange sac. My main point is that there is a small difference between the value of pieces in human games and computer games. This is something that an over the board player needs to watch out for when analysing games with a computer. For this reason Houdini 1.5 is not my primary analysis engine. It would only take a minor modification to make it more useful, however.
And yes the value of the pieces changes depending on the position. In over the board games I get some idea of what the material balance is based on a quick tally using this article by Larry Kaufman: http://danheisman.home.comcast.net/~dan ... alance.htm
I have added further refinements to this (word document explaining Toga code + "rules of thumb" by Lev albert +"the value of the pieces" by Soltis + other stuff). Kaufman analysed an interesting game (I think on an ICC interview) in which he took advantage of his knowledge of some of the changes caused by exchanges to average piece values, to win a game in the World Senior Chess Championship (that was the year that he won the title - I suppose I should be referring to him as a GM ).
). So I have had to write out everything long hand (I was lazy enough to find a position to use as an example on chessok opening explorer so hopefully the FEN is correct).
).