PDA

View Full Version : Challenge your computer engine! position 1.



qpawn
28-10-2006, 02:39 PM
3rk2r/pb2bppp/1qp1p2n/3pP3/1P3P2/N1PBBQ2/P3K1PP/R6R b k - 0 15

This is a position from my last correspondence game. Now it’s finished I can post it. A few months ago I promised to put up a position that would kerflummox computers. Well, here it is. Or, to be more accurate, here THEY are; I have more positions to send the smoke coming out of your engine.

In an offbeat line of the French [2. Nf3] I have declined a pawn grab to move my opponent’s white king into exposure. I hit upon an ambitious plan that is not strictly sound since without black castling g7 is exposed to a Qg3 at some point with gain of tempo. After 15 moves white has played Be3 and I must respond :

Do I…
Block by …d4? [the move I chose and Bill Jordan quite likes]
Block by…c5 ? [the move that my Fritz 8 likes. I do not like this move on positional grounds; the b file could open up with pressure from the white rooks.]
Evade by…Qc7? [passive but possible.]

To my knowledge these are the 3 sane candidates. So, all you computer nuts out there: do you take the challenge ? Will your engine find the truth of the position? Or does it just churn out claptrap? Ha ah ha.

Only 3 choices. Looks like a piece of cake huh? I can assure you that it isn’t. In my opinion this sort of “open-ended” decision is the sort that computers are clueless in making. Indeed, if anyone wants to play anti-computer chess it should involve giving the computer choices like this.

I can’t play crossboard chess of any type for charity or anything else. All I can do is correspondence chess. I like to write down my ideas. My days of playing any type of crossboard chess are over. I need a few days to make each move :lol: Mind you, in the 19th century the oldmasters used to take that long in crossboard - where's my time machine? :)

Kevin Bonham
28-10-2006, 04:00 PM
Looks to me like Fritz8 is correct and your opponent missed the correct reply to ...d4. 15....c5 and if 16.bxc5 Qb2+ 17.Nc2 d4! 18.c6 Ba6 19.Bxa6 Qxc2+ 20.Bd2 dxc3 21.Qxc3 Rxd2+ 22.Qxd2 Qxc6 (every white move in this sequence appears forced) -+ as Black has way too much play for the exchange or else 16.Bxc5 Bxc5 17.bxc5 Qxc5= and who cares about the open b-file since Black will soon castle and has a half-open c-file to hassle a pawn down.

Fritz6 (yes I still haven't bothered to upgrade it, but I might do someday soon) initially likes your 15...d4 with the main line being 16.Nc4 c5 17.Nxb6 Bxf3+ 18.Kxf3 axb6 but at this point it sours on the position as it realises that after 19.cxd4 cxd4 20.Bd2 White has an extremely strong position - two bishops, outside passed pawn, weak opposing pawns.

If 15...Qc7 16.Qf2! is sneaky and strong.

What happened in the game?

qpawn
28-10-2006, 05:00 PM
I don't know how to do all that pgn upload stuff so here's the entire game:

[Event "CCLA 4/2 tourney "]
[Site "Melbourne"]
[Date "2006.10.28"]
[Round "?"]
[White "Cesar Laguerta"]
[Black "Andrew Thornton"]
[Result "0-1"]
[WhiteElo "1290"]
[BlackElo "1470"]
[PlyCount "58"]
[TimeControl correspondence chess]

{28MB, Fritz8.ctg, DEFAULT}

1. e4 e6 2. Nf3 d5 3. e5 c5 4. d4 cxd4 5. Nxd4 Nc6 6. f4 Bc5 7. c3 Qb6 8. Nxc6 Bf2+ 9. Kd2 bxc6 10. Qf3 Bb7 11. Na3 Rd8!? 12. Bd3 Nh6 13. Ke2 Bc5 14. b4 Be7 15. Be3 d4 [point of last diagram with decision] 16. Nc4 Qc7 17. cxd4 c5 18. Be4 cxd4 19. Rhc1 Bxe4 20. Qxe4 Qd7 21. Bd2 O-O 22. h3 Nf5 23. Kd3? Ne3!? 24. a3 Nd5 25. g3 Qb5 26. Ra2?? Rc8 27. Rac2 Rfd8 28. Qe2 Nc3 29. Bxc3 Qxc4+ 0-1

Kevin Bonham
28-10-2006, 06:43 PM
Thanks. It looks like ...d4 is playable after seeing that, not sure if it is better than ...c5. Computers may well reject 16...Qc7 because after 17.cxd5 c5 18.Qg3 (white played Be4? in the game) they think that white is winning. But he isn't, because ...cxd4 19.Qxg7? Rg8 20.Qxh6 dxe3 23.Rhg1 Qc6 24.Qh3 Rxd3! and Black is winning. So white should instead consider 19.Rac1 Qd7 20.Bf2. I would say white is slightly better here because Black's overstretched d-pawn is a long-term liability but Black has good chances too.

qpawn
28-10-2006, 07:26 PM
Yes. My computer engine gave 18. Be4 ?? but suggested no alternative. It gave an annotation of...wait for it...18. dxc5 Bxf3

What the ?????????????

You are right about the lines with Qxg7 being a dead loss for white; I had a look at those at the time and my analysis was similar to yours.

qpawn
28-10-2006, 07:49 PM
I add that this whole game is not my usual "playing style". I usually play very solid chess. Basically, I won here by complicating the position to a greater degree than what my opponent could cope with. To be honest, I was like a "kid in a lolly shop" when I saw 11...Rd8!? even though the solid ...Nh6 was probably better to enable g7 to be protected by castling in 1 move. When you are as woeful as what I am at crossboard tactics and suddenly you play CC and get all these "juicy" tactics that you can analyse longhand...self-restraint goes out the window :D

For the rest of my CC games I will be my usual solid self again :D

I all goes to show that in positions where "appearances deceive" and it takes many ply to find the move like 24 Rxd3 ! that computer engines are pretty hopeless. No computer would ever have played the moves in that superb CC game between Estrin and Berliner.

Kevin Bonham
28-10-2006, 11:32 PM
Yes. My computer engine gave 18. Be4 ?? but suggested no alternative. It gave an annotation of...wait for it...18. dxc5 Bxf3

What the ?????????????

Did it do this in some kind of full game analysis mode? I have never been at all impressed with those.

qpawn
29-10-2006, 10:19 AM
There is that full analysis mode as you say. And that is what I used. I used my best PC in all these analysis; I know that it says above 28 mb but that is the old computer I have hooked to the net. I used my newer computer which is 150MB or somethintg like that. I have noticed a clear difference in the analysis quality depending on the computer power of my PC. Someone told me this at Elwood CC and I was skeptical; but it is true - the more pC power the better.

Frankly, I am a real amateur at this computer analysis caper. I set full analysis to about 40 mins with about 30 depth whatever that means. Fritz is very user-unfriendly in this respect; its features are not explained in a way that a dummy like myself can understand.

How do you get a list of the candidates that the computer considers in order from its preferred choice downwards? And how do you get a numerical evaluation [-4.5 etc] of each move? I know that plus or minus 3 means a piece of difference.

Kevin Bonham
29-10-2006, 06:43 PM
How do you get a list of the candidates that the computer considers in order from its preferred choice downwards? And how do you get a numerical evaluation [-4.5 etc] of each move? I know that plus or minus 3 means a piece of difference.

For a given move what I do is use Deep Position Analysis but set the branching in first move to, say, 20.

WhiteElephant
29-10-2006, 09:28 PM
Fritz is very user-unfriendly in this respect

Yeah, agree with that...I've been using Fritz for a long time and there are still features in the menus that I have no idea what they are. For example, I still don't know whether you can set Fritz to play at a particular rating...

Kevin Bonham
29-10-2006, 09:45 PM
Yeah, agree with that...I've been using Fritz for a long time and there are still features in the menus that I have no idea what they are. For example, I still don't know whether you can set Fritz to play at a particular rating...

When you play "rated games" against it there is a sliding scale within a certain range for each engine if you uncheck the unleashed option. (I assume this function is still there on recent versions?)

qpawn
29-10-2006, 09:50 PM
If you go to rated game I think you can change the strength of Fritz; but it's not obvious how to do it.

Another thing which took me a long time to find on Fritz is how to turn the engine off!! It is weird that there is no "referee" mode on Fritz like there used to be on my old Novag. For CC I want to bang in the moves only; I don't want to see any of the computer analysis window!!! Also, sometimes Fritz does this awkward "thing" where you try to take a move back in analysis and it won't let you unless you take the current move back. That is so annoying. And it led me to use chessbase light instead for working through my CC games.

WhiteElephant
29-10-2006, 09:59 PM
I know how to handicap Fritz by making it play in drunk mode, patzer mode etc. I know about the sparring option where you can set it to easy, hard,etc. But to play at a specific rating...hmmm...I will experiment..thanks for the suggestions.

Edit: Just worked it out...awesome. Thanks!

ElevatorEscapee
02-11-2006, 07:33 PM
Fritz can get drunk!? :eek: