Archived from groups: rec.games.chess.computer (
More info?)
Here is the position with White to move in DePalma (1860) vs. Fernandez
(2139):
W: Ka2, Qb3, pc5, pg5
B: pa5, Qc2, pf7, pg6 and Black has just moved his Kh7-g7, which loses by
force.
My Shredder 7 has been analyzing this position for 4 minutes now and gives
the evaluation as -0.44.
Now move 63. Qb2+ and see what Herr Shredder says. Mine initially jumps to
a -2 evaluation, but after about 20 seconds says White is winning. If B goes
over to the Q side via f8 W captures on a5 and defends c5. The threat of
queening that pawn wins the game. If black tries ...f5 at any point (which
my opponent did after moving his king) he loses just as badly.
Players as bad as me hardly get to play interesting endings like that (or
play them well). I'm proud of that game because I noticed the f7 pawn in the
way and that I'd be able to outflank my opponent in all variations (or
simply queen).
Angelo
<Chaos@chaos.org> wrote in message
news:d05ph15dom89me8sdll638164jchiq9m42@4ax.com...
> On Sun, 4 Sep 2005 23:56:43 -0400, "Angelo DePalma"
> <angelodpnospam@nospam.gmail.com> using recycled electrons muttered
> something
> about:
>>I'm fascinated that you're interested in human-like computer play.
>
> Actually, it was not until I was watching this engine/settings play that I
> became interested in seeing how like a human this one plays. Here are a
> couple
> of games fro one of the matches:
>
> [Site "Riverdale"]
> [Date "2005.09.04"]
> [Round "2.1"]
> [White "CMX Dayffd 1.0"]
> [Black "Junior 6.0"]
> [Result "1-0"]
> [ECO "C48"]
> [BlackElo "2451"]
>
> 1. e4 e5 2. Nf3 Nf6 3. Nc3 Nc6 4. Bb5 Nd4 5. O-O Nxb5 6. Nxb5 c6 7. Nc3 d6
> 8.
> d4 Qc7 9. Bg5 Be7 10. h3 O-O 11. Qd2 Be6 12. Rad1 Rfd8 13. d5 Bc8 14. Rfe1
> h6
> 15. Bxf6 Bxf6 16. b3 Be7 17. Ne2 Bd7 18. c4 a5 19. Ng3 a4 20. b4 cxd5 21.
> cxd5
> Qb6 22. Nf5 Bxf5 23. exf5 Rdc8 24. Rc1 Qb5 25. a3 Rxc1 26. Rxc1 Bd8?
>
> {Better is 26...Bf6!}
>
> 27. Qd1 Bb6 28. f6! g6 29. Nh2 h5 30. Nf1 Bd8 31. Qf3 Rb8 32. Nd2 Bb6 33.
> Rc4
> Bd4 34. Rc7 Bb6 35. Rc2 Ba7 36. Rc3 Re8 37. Rc4 Bd4 38. Qe4 Rd8 39. Rc2
> Rf8 40.
> Nf3 Bb6 41. Ng5 Qd7 42. Qc4 Ra8 43. Ne4 Rf8 44. Rc1 Ra8 45. b5 Qf5 46. Qe2
> Rc8
> 47. Rxc8+ Qxc8 48. Qd2 Bc5
>
> {Giving up the knight to allow the king to go to h7 without being
> immediately
> forced back by Ng5+ etc., but this sacrifice does not effect the outcome
> of the
> game.}
>
> 49. Nxc5 Kh7 50. Ne4 Qc7 51. Ng5+ Kg8 52. Ne6 fxe6 53. dxe6 Kh7 54. b6 Qc8
> 55.
> e7 1-0
>
> [Event "05CMXD1.0vJunior 6"]
> [Site "Riverdale"]
> [Date "2005.09.04"]
> [Round "6.1"]
> [White "CMX Dayffd 1.0"]
> [Black "Junior 6.0"]
> [Result "1-0"]
> [ECO "A09"]
> [BlackElo "2451"]
>
> 1. Nf3 d5 2. c4 d4 3. g3 g6 4. e3 c5 5. exd4 cxd4 6. d3 Nc6 7. Bg2 Qa5+ 8.
> Nbd2
> Nf6 9. O-O Bg7 10. Nb3 Qb6 11. Re1 Bg4 12. h3 Be6 13. Ne5 Rc8 14. Nxc6
> bxc6 15.
> a4 c5 16. a5 Qa6 17. Qe2 O-O 18. Bf4 Rfe8 19. Be5 Bh6 20. g4 Nd7 21. Bg3
> Bg7
> 22. Qf3 Rf8 23. Bf4 Kh8 24. Re2 Rg8 25. Qb7 Qxb7 26. Bxb7 Rcd8 27. Bc6
> Rge8 28.
> Rxe6 !!
>
> {Junior expected 28.Bxd7 Rxd7 when it is still in the game, although
> white should still win on the queenside.}
>
> 28... fxe6 29. Bc7 e5 30. Bxd8 Rxd8 31. Bxd7 Rxd7 32. Nxc5 Rd8 33. b4 Rc8
> 34.
> Ne4 Kg8 35. b5 Rb8 36. Rb1 Kf7 37. c5 Ke6 38. b6 Kd5 39. bxa7 Ra8 40. Rb7
> Kc6
> 41. a6 Bf6 42. g5 Bg7 43. Nd2 e4 44. Nxe4 Be5 45. Rxe7 Bf4 46. Rxh7 Kb5
> 47. Rb7+
> Kxa6 48. c6 Ka5 49. Nc5 Bd6 50. Ne6 Ka6 51. Rd7 Be5 52. Rd5 Rc8 53. Rxe5
> 1-0
>
> The exclams, queries, etc., are mine. I think it would be difficult to
> tell that
> engines were playing the above games and not humans.
>
>>I am
>>also, but not from the point of view of making the program stronger. I
>>would
>>like to set an engine up so it plays human-like at the 2100-2200 level. I
>>think it would be a good opponent for me (I'm around 1900).
>
> That can be done, although it is best done with engines that have
> changable
> parameters, such as values for center control, piece/pawns values,
> mobility,
> combinations, etc., rather than in Fritz or cousins "handicap" modes and
> the
> like. There the results tend to be to extreme, with the engine making a
> childish
> blunder by poitching a piece or two, then playing like a GM, and so on. It
> requires a finer touch. Other than that, there are numerous free engines
> that
> are simply "weaker" than the other engines. For example, BamBam may give
> you a
> good game, or Doctor? 3.0,. And a number of both uci and wb engines.
>
>>I'll send you those files as zipped PGN tomorrow or Tuesday. I'd like to
>>wait for the results of the G/60s.
>
> Ok.
>
>>BTW, as these engines become capable of beating a higher and higher
>>percentile of GMs, I believe human analysis is becoming irrelevant.
>
> I disagree with you on this. It is true that chess is 99% or so tactics.
> It is
> also true that these engines seldom make tactical errors, at least within
> the
> first move or two. But for an overall plan, and recognition, you still
> need a
> human to advise or suggest what would be a goal to shoot at. For instance,
> 3 - 4
> months ago I was playing with a weak chess player, and during one game,
> late in
> the middle game, I mentioned that she should take care of her bishop and
> not let
> me exchange it since her one passed pawn queening square was the same
> color as
> her bishop. Now, any reasonably competent player would know that all other
> things being equal, that bishop gains in significance as the pawn moves
> closer
> to that square, where a computer engine cannot tell you that at some point
> in
> the undefined future a piece's value may be more than it's current value.
>
> It is true, also, that as you say, "There was a time when a GM or strong
> IM
> could poke fun at chess engines," but that time is certainly over, and has
> been.
>
>> I have a position from one of my games, where I beat a much-higher rated
>>player in a study-like ending. I saw the long series of moves that won the
>>game by force, but to this day my Shredder cannot.
>
> Ummmmmm....how long have you let Shredder work on it? If the entire line
> was
> forced, Shredder may be able to find the line given enough time. Do you
> have the
> game score handy?
>
>>Still, I would trade any 2600 player's (and probably any 2800 player's)
>>chess knowledge and strategy -- all the stock "plans" for playing the
>>canonical positions -- for the sheer calculating power of one of these top
>>engines. They will beat any Nimzoindian expert (or Kalashnikhov expert, or
>>Benoni expert, etc.) at their own game time after time. It is becoming
>>difficult, even silly, to argue with that kind of success.
>>
>>What do you think?
>
> That is a hard question! Think I'll pass - I would prefer to combine both.
> I'll
> tell you one thing that has surprised me the most, I think. If you look at
> the
> analysis of old tournaments by the old masters, for example St.
> Petersburg,
> 1909, or Nottingham, 1936, and so on - the ones that have good game
> annotations,
> it has surprised me how often some of the top engines, such as Fritz 6, 7,
> 8,
> and others, offer the same tactical suggestions as the old masters did. I
> don't
> know if it is that the engines offer the same lines that the masters did,
> or
> vice versa.
>
>
> No problem is so formidable that you can't just walk away from it.
The variety
Facebook
Twitter
Instagram