PDA

View Full Version : Fritz - analysis function

Southpaw Jim
06-06-2006, 10:21 PM
Not a major concern, just wondering if anyone has a rational explanation:

The other night I ran a blunder check on a recent game, focussing on white. I then ran a blunder check on the same game, but for black. Fritz's evaluations for various positions were not the same each time! Eg, the first time the evaluation given at a particular point might have been -20.3, the next time it was -24.4.

How could this be? :eh:

Basil
06-06-2006, 11:31 PM
the first time the evaluation given at a particular point might have been -20.3, the next time it was -24.4.

I'm not a Fritz user, but if I understand its assessment ratings, it appears the first [-20] is saying:

This position is very bad. Your mother has been sold into prostitution, your father into slavery. Your king is moments away from being sodomised, but before that happens, your opponent will eat all your children.

The second assessment [-24] appears to mean that your children will have to watch all the aforementioned happen, and then be eaten.

In other words, the advantage of the move [or disadvantage of it in this case] is sufficient for the assessment to be further developed from diabolical to devilishly diabolical.

Desmond
07-06-2006, 10:38 PM
Was the position analysed for the exact same period of time on both occasions?

Southpaw Jim
11-06-2006, 10:11 AM
Was the position analysed for the exact same period of time on both occasions?

The settings were identical for both occasions, with the exception of which side was being analysed.

I was wondering whether Fritz was remembering the previous analysis, and therefore able to spend more time on given positions the second time round? :hmm:

Desmond
11-06-2006, 10:29 AM
The settings were identical for both occasions, with the exception of which side was being analysed.

I was wondering whether Fritz was remembering the previous analysis, and therefore able to spend more time on given positions the second time round? :hmm:

I very much doubt that Fritz is able to learn in this manner. Have never heard of a chess program that can learn anything it didn't know out of the box. If anyone knows of evidence to the contrary I'd be very keen to read about it.

Thunderstone
14-06-2006, 11:27 AM
Apparently Hiarcs can learn.

"New enhancements mean that while playing or analysing games, Hiarcs is able to learn more about the positions and moves it sees to improve its chess strength in future games."

Above info from http://www.hiarcs.com/pc_chess_hiarcs.htm

How good this feature is I don't know but it would be interesting to hear from anyone that has Hiarcs.

Desmond
14-06-2006, 06:53 PM
Apparently Hiarcs can learn.

"New enhancements mean that while playing or analysing games, Hiarcs is able to learn more about the positions and moves it sees to improve its chess strength in future games."

Above info from http://www.hiarcs.com/pc_chess_hiarcs.htm

How good this feature is I don't know but it would be interesting to hear from anyone that has Hiarcs.

Hi Thunderstone
Yes, "apparently" is defintely the correct word. The site is very vague on this point. All it says it that there is a variable called Postion Learning and that it is switched on by default. Sounds like cheap sales talk to me.

Davidflude
14-06-2006, 07:47 PM
The settings were identical for both occasions, with the exception of which side was being analysed.

I was wondering whether Fritz was remembering the previous analysis, and therefore able to spend more time on given positions the second time round? :hmm:

Go to the section where it says change engine and clear the hash table after each set of analysis. I am fairly certain that the Chessbase family of programs keeps as much of the hash table as it can. Of course this shows when you have really big hash tables.

Trent Parker
16-06-2006, 10:14 AM
Ok ok ok. I have a question.

For those of you who put your games into fritz and use a full analysis. How long do you normally put them in for? What settings?

Southpaw Jim
16-06-2006, 12:17 PM
Of course this shows when you have really big hash tables.

What qualifies as really big? I have ~158MB :hmm:

Thunderstone
17-06-2006, 11:23 PM
I read that keeping hash tables in multiplies of 4 was better.
So having 156meg is better then 158 meg?
Is this really true?

queenspawn
20-06-2006, 06:53 PM
I'm not a Fritz user, but if I understand its assessment ratings, it appears the first [-20] is saying:

This position is very bad. Your mother has been sold into prostitution, your father into slavery. Your king is moments away from being sodomised, but before that happens, your opponent will eat all your children.

The second assessment [-24] appears to mean that your children will have to watch all the aforementioned happen, and then be eaten.

In other words, the advantage of the move [or disadvantage of it in this case] is sufficient for the assessment to be further developed from diabolical to devilishly diabolical.
Uhm, I just stopped laughing. That was very unfair of you... A perfectly good cup of coffee splattered down my front. :doh: :lol:

Basil
21-06-2006, 03:23 AM
Uhm, I just stopped laughing. That was very unfair of you... A perfectly good cup of coffee splattered down my front. :doh: :lol:
QP ... thanks for your reaction. As I was saying in the Shoutbox, I am completely wasted over here ... tragic ... pearl before swine ...