PDA

View Full Version : Challenge your chess computer: position 2!



qpawn
28-10-2006, 04:45 PM
3r1rk1/p2qbppp/4p3/3nP3/1PNpQP2/P2K2PP/3B4/R1R5 b - - 0 25

Here is a position that DID kerflummox my Fritz 8 engine. It is later on in the same game as last time but after move 25. I played 25Qb5 a natural move which many people below 1000 ELO would play as well. Here it isnt the human player at test; it is the engine. You will not believe what Fritz 8 recommends instead in deep analysis and gives a -+ evaluation to; it is the WORST computer recommendation I have ever seen. If the engine had a urinary system it would have been drug tested.

Test your engine: does your engine recommend the same tomfoolery as mine?

Can your chess engine withstand the challenge? Or will it, like Fritz 8, crack under the pressure.

The move recommended by Fritz 8 was so bad that I am strongly inclined to never do a post-mortem with a chess engine again. Yep. I said never.

Kevin Bonham
28-10-2006, 07:00 PM
The move recommended by Fritz 8 was so bad that I am strongly inclined to never do a post-mortem with a chess engine again. Yep. I said never.

Computers will often recommend a silly move by a small margin. They're most useful in post-mortems for finding errors and calculating tactics. Mine preferred ...Qb7 (but not by much) over ...Qb5, the very strange looking ...f5, ...Qc6 and ...Rfe8.

Let me guess - yours recommends ...f5 ? :eek:

qpawn
28-10-2006, 07:19 PM
Interesting.

My Fritz 8 engine gives 25...Qc6-+ on full analysis

Gulp.

That is bizarre. I mean, if my idea was to try to launch a discovered attack on the white queen why not ...Qb7 when the black queen is not in the line of fire from the rook on c1? What's even stranger is that there is no continuation of the line; it just gives 25...Qc6-+ and that's it. :eek:



I considered ...Qb7 in the game. I never thought of ...Qc6 :eek:

But...Qb5 is the most "human" move that people would play.

But here is where ....f5 is interesting; Fritz gave my 24... Nd5 ?? and recommended ...f5
weird. And totally antipositional; ...f5 weakens the e pawn while ....Nd5 is surely a great centralisation.

If I played crossboard chess I would not use a computer for any opening preparation, novelties etc; their analysis is so unreliable.

IF I were a serious chessplayer with a title I wouldn't dream of preapring with a computer. At least not Fritz 8 anyway.

Kevin Bonham
29-10-2006, 12:41 AM
Interesting.

My Fritz 8 engine gives 25...Qc6-+ on full analysis

Gulp.

That is bizarre. I mean, if my idea was to try to launch a discovered attack on the white queen why not ...Qb7 when the black queen is not in the line of fire from the rook on c1?

This has me stumped too, but the "threat" from the Rc1 is not serious, so I would not say ...Qc6 is bad, just that I don't see the point of it cf. ...Qb7.


But here is where ....f5 is interesting; Fritz gave my 24... Nd5 ?? and recommended ...f5
weird. And totally antipositional; ...f5 weakens the e pawn while ....Nd5 is surely a great centralisation.

I've been looking at some lines to try to get a handle on why Fritz thinks ...f5 is playable. It turns out that there is a reason for it, which is that it half-opens up the f-file for attacks on White's suspect king. If Black has played ...f5 and ...Bh4 then in lines where White plays Bxe3 dxe3+, Kxe3 becomes unplayable because ...Rxf4!! results in a winning attack for black via ...Qf7+. However if Black has not played ...f5 then the f-file stays closed and none of this works. So the point of ....f5 is it either forces white's queen to move, or else restricts White's ability to win a pawn.

Good players follow positional rules. Great players know when to break them. Fritz seems to realise that there are strong tactical reasons to allow the backwards pawn on e6. It is a complex position so that may not be the last word on it, but I think I understand why Fritz finds the strange-looking ...f5 attractive.


If I played crossboard chess I would not use a computer for any opening preparation, novelties etc; their analysis is so unreliable.

IF I were a serious chessplayer with a title I wouldn't dream of preapring with a computer. At least not Fritz 8 anyway.

You can still use them a lot for tactics when analysing novelties. And the issue is knowing how to make the best of them despite their limitations. (The classic game in which Leko defeated Kramnik was a good example of what a failure in this regard looks like!)

Aaron Guthrie
29-10-2006, 09:47 AM
I can think of a few reasons why a computer would prefer Qc6 over Qb7. On c6 it has more squares. On c6 its closer to the king, and it may be programmed to increase this value if the king is in the center with a certain number of peices on, for example. If it's calculation includes moving the rook to c8, then the queen on c6 is protected.

So while the Rook on c1 would just eliminate Qc6 as an option to humans, the computer would rather have a specific value that it assigns to putting the queen on the same file as the rook. Im guessing this value would probably be quite low, and so a number of other factors can overide it.

Of course this doesnt address why it would like Qc6, or why it would assign such a high value to it, just why its possible for a computer to like Qc6 in preference to Qb7.