Go back
Strange Fritz Analysis

Strange Fritz Analysis

Only Chess

e

Joined
19 Nov 05
Moves
3112
Clock
29 Mar 06
1 edit
Vote Up
Vote Down

Here's a position I've reached from the Maroczy bind:



Fritz evaluates the position at -5.26 after 30 minutes.

e

Joined
19 Nov 05
Moves
3112
Clock
29 Mar 06
Vote Up
Vote Down

I moved Nc5, not realizing that this would lead to:

e

Joined
19 Nov 05
Moves
3112
Clock
29 Mar 06
2 edits
Vote Up
Vote Down

As a result, the forced moves led to:



Strangely, Fritz evaluates this position at -5.37! This is better than the initial evaluation!

I thought I blundered and to my surprise, the following evaluation is actually better than the initial after the forced moves. It also doesn't seem to be a fluke as I have left it to analyze each position for about 30 minutes.

So going from being up a knight and a pawn (~+4 total pawn points) to losing a rook for a bishop (~+2 total pawn points), yields almost the same result? That's odd, I would have thought that the evaluation would have gone down at least a pawn and a half.

Someone please explain this to me....

X
Cancerous Bus Crash

p^2.sin(phi)

Joined
06 Sep 04
Moves
25076
Clock
29 Mar 06
Vote Up
Vote Down

In the first position 1. ... Nc5 is the second best move at a depth of 16 ply according to Fritz 8.

What you may be seeing is a horizon effect, in the first position Fritz can't see far enough to see just how bad white's position is. By the time it reaches the final position you posted it can see further and give a score based on a deeper look.

e

Joined
19 Nov 05
Moves
3112
Clock
29 Mar 06
Vote Up
Vote Down

Originally posted by XanthosNZ
In the first position 1. ... Nc5 is the second best move at a depth of 16 ply according to Fritz 8.

What you may be seeing is a horizon effect, in the first position Fritz can't see far enough to see just how bad white's position is. By the time it reaches the final position you posted it can see further and give a score based on a deeper look.
That seems to make sense, but I left it for 30 minutes on each position using Fritz 9 on a single 2.4 Ghz AMD processor. I thought Fritz could calculate millions of positions in just a second, shouldn't it see far enough at with such specs?

X
Cancerous Bus Crash

p^2.sin(phi)

Joined
06 Sep 04
Moves
25076
Clock
29 Mar 06
Vote Up
Vote Down

Originally posted by exigentsky
That seems to make sense, but I left it for 30 minutes on each position using Fritz 9 on a single 2.4 Ghz AMD processor. I thought Fritz could calculate millions of positions in just a second, shouldn't it see far enough at with such specs?
No. If you watch the depth to which Fritz has searched it gets to 9 ply even on a slow computer within seconds. Then each successive ply takes longer.

The number of positions that must be checked for a given ply depth increases exponentially (i.e. in the order of 2^n). From this it's easy to see that no matter how long you leave it you won't get far enough to see everything in a position (unless you manage to reach tablebase positions in your search).

Since I first checked the starting position you gave I haven't stopped Fritz from analysing, it's currently approaching 18 ply after 22 minutes on a very fast computer.

X
Cancerous Bus Crash

p^2.sin(phi)

Joined
06 Sep 04
Moves
25076
Clock
29 Mar 06
Vote Up
Vote Down

To further illustrate my point after nearly 5 hours Fritz has reached a depth of 20 ply looking at the best two moves (this changes speed to a degree).
In doing so Fritz has looked at 12.5 billion nodes (individual positions).

GP

Joined
10 Mar 06
Moves
4933
Clock
29 Mar 06
Vote Up
Vote Down

I'm missing something here......Fritz is showing white down almost 5 and a half points, what's the problem? How badly do you think white should be rated?

e

Joined
19 Nov 05
Moves
3112
Clock
29 Mar 06
Vote Up
Vote Down

Well, I only expected Fritz to give White a lower score when it lost two pawns , rather than a higher one.

W
Angler

River City

Joined
08 Dec 04
Moves
16907
Clock
29 Mar 06
Vote Up
Vote Down

Originally posted by XanthosNZ
To further illustrate my point after nearly 5 hours Fritz has reached a depth of 20 ply looking at the best two moves (this changes speed to a degree).
In doing so Fritz has looked at 12.5 billion nodes (individual positions).
Hiarcs 9 was at 13 ply almost instantly, but after four hours has extended it to 17 ply. It has examined just under 3 billion nodes.

current evaluation: -5.63

I have it examining only the best move.

GP

Joined
10 Mar 06
Moves
4933
Clock
29 Mar 06
Vote Up
Vote Down

Originally posted by exigentsky
Well, I only expected Fritz to give White a lower score when it lost two pawns , rather than a higher one.
Once again, I'm missing something.....exig, you said fritz scored the game at -5.3 or so...that is white's score. Fritz rates the white pieces, whether it's you or Fritz that is white.
So what's the problem?

GP

Joined
10 Mar 06
Moves
4933
Clock
30 Mar 06
Vote Up
Vote Down

I think I see the problem...did you think that -5+ score was yours?

Cookies help us deliver our Services. By using our Services or clicking I agree, you agree to our use of cookies. Learn More.