Hi MR
The engine might not play some of Tal's sacs because some were unsound.
Defences were found in the analylsis room (sometimes weeks later).
A better idea is to go through a book of combinations, sometimes these things
come up with quicker (uglier) wins or missed defences.
Especially if the book is printed before 1980.
If the combination is sound and was played by a human then the modern
computer will find it.
Hi Schach.
I agree with everything you are saying, again think my use of the term
'heavy tactical position' may have been misleading.
Not a simple two move trick which should be mastered by everyone
type of position.
But something like a 6 move combo with tricks and traps in it's many branches.
Cannot think off hand of any sound and forced OTB combination that has been
played by a human winning a piece or mating an opponent that any modern
computer would miss after 30-40 seconds.
Most get them within 10 secs.
Can show 1,000's that have been missed by humans after 10 minutes thought.
(me and most likely you included, we all miss them, these things do not.)
It fails on the intuition sac where long term pressure is guaranteed, this
is soley a human field. The computer might not find the coming mate in 23 moves
time but the human has not seen it either.
But if it is a complex tactcial postion and there is a clockwork type forced win
in there then the computer will thread it's way through and find it.
If it is that complicated that it cannot see it after 30 seconds then it's a
safe bet the human won't see it either.
So on you side Schack but look out for this one.
"EVERYTHING has to be spelled out for computers."
Had this argument all over the net with some pretty serious people,
used something like this once. "EVERYTHING has to be spelled out for computers, "
Apparently these things do have the ability to learn.
If it goes down an opening line put in by humans and does not like what it 'sees'
and 'sees' a better line, it can bump up the good line for future use.
Of course that leaning ability was also put by humans. 🙂
Re blitz.
I thought I was being Mr Clever by only playing 4/0 games thinking the time
was too short for the lad to enter the move into his computer and wait for the reply.
I told months later than you can set up your box so you do not need to enter
the move. It picks up the move and replies without any help.
All the human player does is select it's opponent.
That explained quite a few things in quite a few games I had been playing.
Put me off blitz for ages that one.
Originally posted by greenpawn34
If the combination is sound and was played by a human then the modern
computer will find it.
Cannot think off hand of any sound and forced OTB combination that has been
played by a human winning a piece or mating an opponent that any modern
computer would miss after 30-40 seconds.
Most get them within 10 secs.
Check out Bxh7 by David Rudel, I think you reviewed it; there are examples of boxes missing forced wins even today.
Originally posted by Mad RookBy "personality" I didn't mean a chess engine tweaked to emphasize highly aggressive tactical play, or positional play, or anything of the sort.
Well, I guess I'm not much of a believer in various engine personalities like Kramnik, Anand, Fischer, etc. (Mainly because I don't think I could tell the difference, and I suspect it's more of a gimmick anyway.)
Anywho, I *am* having fun playing with the Pro Deo Tal personality. It's pretty weak, around 2000 elo. I edited my own version and got it to pl ...[text shortened]... he time, I'd go through some of Tal's games to see how many of those the engine will play.
Greenpawn34, regarding computers that "learn", you're just talking about a sort routine where analyzed move sequences are assigned relative values depending on how much material they win, and so forth. All of that is algorithmically spelled out in programming -- every single bit of it -- as is the later search in a given position that brings up the best line found so far.
Originally posted by greenpawn34Ten minutes isn't a long time for move analysis, in standard length chess with traditional time controls. In a highly analytic (i.e., tactical) position it's inadequate, even for very strong players (e.g., grandmasters).
There are 1,000s of instances of sound and forced OTB combinations winning a piece or mating an opponent that have been missed by humans after 10 minutes thought -- including me and most likely you. (paraphrased)
The broader point, however, is that very short time controls are more conducive to poor play by humans, including overlooking combinations, even though things may still be overlooked while playing standard length chess. It takes an exceptionally strong player to play well AND fast, so it's very hard for weak players to get strong by playing speed chess. (Obviously, anyone with an expert rating (or close to it, as is the case with greenpawn34) is not a weak player, so it can be assumed that these remarks are directly primarily elsewhere.)
I wouldn't dream of using myself as a standard. I aspire to be a good player, but I don't consider myself to be one, yet.
Not only do modern chess computers miss things, greenpawn34, but they also commonly disagree with one another. Another poster made reference to a book by David Rudel. He also has some comments, here, comparing various engines (do a term search for "Rudel" and skip down to the question "What computer program or programs do you use for analysis?"
http://www.kenilworthchessclub.org/kenilworthian/2009_12_01_archive.html
Hi Schach.
I was just quoting what I often get hit with when I have a pop at these things.
"Apparently these things do have the ability to learn.
If it goes down an opening line put in by humans and does not like what it 'sees'
and 'sees' a better line, it can bump up the good line for future use. "
I don't know all the technical stuff that enables them to do it.
(if they can ). So I tread away from that one as I cannot match them
when they start going on about how the thing actually works.
I just know it's not chess thinking as humans do it.
"All of that is algorithmically spelled out in programming..."
I like that. I have no idea what it means, but I like it.
Next time I'm having a row with one of the box boys I'll give you a shout.
-----------------------
Hi Greenerpawn.
Just noticed your post, thought I was last to post on here. our nicks are too close.
Good point.
David Rudel gives this game in Bxh7+
Zeynalli - Enkhnar World under 10 ch U10 2007
1.d4 d5 2.e3 e6 3.Bd3 Nf6 4.f4 c5 5.c3 Nc6 6.Nf3 Bd6 7.0-0 Bd7 8.Ne5 0-0 9.Nd2
c4 10.Bc2 Rc8 11.Ndf3 b5 12.b3 Ne8 13.bxc4 bxc4 14.Ng5 f5 15.Nxc6 Rxc6 16.Nf3
Nf6 17.Ne5 Rc8 18.a4 Ne4 19.Bxe4 fxe4 20.Ba3 Bxa3 21.Rxa3 Rb8 22.Qc2 Qe7
23.Raa1 Rb3 24.Rfb1 Rfb8 25.a5 Ba4 26.Rxb3 Bxb3 27.Qb2 Qd6 28.Rb1 ½-½
in this this position after Black has just played 12....Ne8
He states that Rybka after 10 minutes failed to find 13. Bxh7+
But if you are going to quote me, please quote me in full. Not wee snippets.
But if it is a complex tactcial postion and there is a clockwork type forced win
in there then the computer will thread it's way through and find it.
If it is that complicated that it cannot see it after 30 seconds then it's a
safe bet the human won't see it either.
This is far from clockwork, there is a lot going on here.
and the human (then 10 years old!) failed to find it.
This position does huge credit to David Rudel and just shows the effort
and skill that went into this book.
Effort in the fact he did not restrict his examples to GM or well known games
and he did not plug in box to find him the wins.
His winning line was worked out over the board and he spotted the shot was
there based on the principles and ideas he lays down in the book. (this is a skill).
The question is -
is 13.Bxh7+ playable.
No at first glance because Black has Nf6.
13.Bxh7+ KxB 14.Ng5+ Kg8 15.Qh5 Nf6
Now David points out that 16.Qh4 with the Rook lift Rf3-Rh3 is deadly.
It's not all forced Black can sac-backs etc. David does not give a complete
exact winning line because there are too many ways to go.
Black can go for the standard g6-Nh5 defence v Rf3-h3 attacks.
David shows other sound attacking ideas against this plan (g6 then Qh6).
But from the first diagram...
...with Bxh7+ and the Rf3-Rh3 idea as it's back drop then White is winning.
Rybka could not figure all the fanatastic lines from here.
Look how many captures there are in the position. ...Nxe4 and Be7
hitting the g5 Knight seem to thwart the attack (it does not).
The human OTB could not see it (perhaps he felt it but lacked the confidence)
How many writers would have left the box to it. No win. OK.
The bookstalls are full of such writers.
David trusted his intuition (and knows what he talking about regarding this sac).
Rolled up his sleeves and dug out the win.
Quoting me again.
It [the computer] fails on the intuition sac where long term pressure is
guaranteed, this is soley a human field. The computer might not find the
coming mate in 23 moves time but the human has not seen it either.
A perfect example of human intuition over cold number crunching.
I guess that David used RB3 - RB4 is out now.
Does it like Bxh7+ anybody who has RB4 want to try it.
A very good book that and I'm glad you remined me of it as I left
a whole scad of puzzles undone.
Also I have to get in touch with David as the cover is best I've seen on
a chess book and I would like the original art work without the book title on it.
So thank you that greenerpawn.
Original review.
http://www.chessedinburgh.co.uk/chandlerarticle.php?ChandID=376