PDA

View Full Version : Some CC Analysis Blind Spots



compchess
25-07-2006, 08:02 PM
Mate in One (Yes it's true)

8/8/pppppppK/NBBR1NRp/nbbrqnrP/PPPPPPPk/8/6Q1 w - - 0 1

The length of time to solve this position depends on if the chess engine you analyse with generates capture sequences first or does a quick scan for the obvious mate.

You could be there a while with Fritz 9........

:cool:

Chris

EZBeet
27-07-2006, 12:30 PM
are you serious?

It took Hiarcs running on my PDA 0 secs to find it.
Fritz 8 took <1 sec
I took 2 secs

:hmm:

MichaelBaron
27-07-2006, 04:11 PM
are you serious?

It took Hiarcs running on my PDA 0 secs to find it.
Fritz 8 took <1 sec
I took 2 secs

:hmm:


Agreed. Computers never miss forced mate

ElevatorEscapee
27-07-2006, 08:02 PM
Agreed. Computers never miss forced mate

It may depend what level the chess computer is set on.

About 8-10 years ago, I used to publish 'mate in 2' problems once a fortnight in my chess column in the local newspaper, with a $10 book voucher prize (kindly donated by a local bookshop) going to the first correct entry drawn.

I would check (pardon the pun) these problems with an old version of Fritz called "Fritz2" which ran on my 486.

Unless I specifically set Fritz2 to its 'solve mate in x number of moves' level, then it would sometimes tend to immediately play a move that lead to a mate in three, (missing the quicker win) no matter what other time limit it was set on.

However if I set it specifically to the 'solve mate' level it would 'think' a little longer on its first move, but always find the mate in two.

(In fact I even had one entrant complaining that a 'mate in two' position I published was invalid, because his copy of Fritz3 could only find a mate in three! However he hadn't set his version of Fritz to the 'solve mate' level either! :lol: )

Maybe there's something with the algorithms in Fritz that are a little different when set at a 'playing' level as opposed to a 'solve mate' level. :)

MichaelBaron
28-07-2006, 12:32 AM
I was referring to the top level...sorry about the missunderstanding

Steve K
30-07-2006, 05:28 PM
On infinite analysis Shredder 9 first considered dxe4 then looked at fxe4 and finally after 34 seconds 'found' Qh1 mate. Goodness knows what's going on in that tiny little brain.

qpawn
01-08-2006, 02:17 PM
So if there had been less than 34 secs left on the clock with no increment then shredder would have lost on time :D

If the wiring can't get alzheimers, CJD or a hangover then it is not a brain: it is a silicon machine's imitation of the superior human abilities :)

Kevin Bonham
01-08-2006, 02:53 PM
So if there had been less than 34 secs left on the clock with no increment then shredder would have lost on time :D

More likely it would have just played another move, albeit an inferior one. Most engines these days are quite good at avoiding losses on time. Has anyone here ever beaten a strong engine/computer on time?

Kevin Bonham
01-08-2006, 03:01 PM
Hahaha. I just decided to give Fritz a game at 1 minute plus 1 second per move to see how it played under such conditions. Playing the exchange "draw" line of the Rubenstein Four Knights I sucked it into an endgame and kicked its silicon butt!

Aaron Guthrie
01-08-2006, 06:13 PM
Junior 8, no tablebases, from the initial position infinite analysis reaches depth 57 in about 5 seconds and thinks it is a draw. After Ne2, it suddenly has a heart attack, changing between the different promotions quickly, settling on h8=R as being in Blacks favour by 90 points, while the other promotions all lead to mate. Clearly this is absurd, h8=N is the obviously the best move. In any case, after promoting to a Rook, Junior recognises the mate, it gives it as the 12th option, and evaluates it as equal. No other engine I have tried this with has this problem.

1.h4 Nf4 2.h5 Kf2 3.h6 Nd4 4.h7 Nf3+ 5.Kh1 Ne2 0-1

Brian_Jones
02-08-2006, 08:16 AM
No comment!

ElevatorEscapee
03-08-2006, 11:51 PM
Do you still sell Chessmaster for the Commodore 64 and Amiga, Brian? ;)

Brian_Jones
04-08-2006, 10:29 AM
Do you still sell Chessmaster for the Commodore 64 and Amiga, Brian? ;)

CP/M and other operating systems were cleared out some years ago with floppy disks of various shapes and sizes.

Everything on CD and DVD for PC today!

pax
08-08-2006, 01:41 PM
More likely it would have just played another move, albeit an inferior one. Most engines these days are quite good at avoiding losses on time. Has anyone here ever beaten a strong engine/computer on time?

Computers are also generally better at time budgeting than they used to be. If a computer knows it only has ten seconds to make a move, it will not end the time period with a very lopsided search tree - it does this by choosing an appropriate search depth for the amount of computation time available. On the other hand, if it has an infinite search time it will probably have a lopsided search tree after ten seconds, and it may take significant time to find a short mate.

The root cause for these sorts of issues is the alpha-beta search algorithm, which is a depth first search. This means you search to a specified search depth in a single branch before exploring alternative moves. It is much more efficient in time and memory than breadth first algorithms, where you look at all the first moves, then all of the replies, then all of the second moves and so on.

The cost is that mate in one can take a while :)