Thinker PV

[Gábor Szõts]

Geschrieben von: / Posted by: Gábor Szõts at 01 March 2004 07:37:21:

Hi,
There must be something wrong with Thinker's evaluation values under WinBoard. The values given for the PV are very small - I estimate they are about 4 times smaller than normal. Is this due to a wrong interpretation of data sent by the engine to WinBoard?
This was already reported by Rober Allgeuer.
Gábor

[Robert Allgeuer]

Geschrieben von: / Posted by: Robert Allgeuer at 01 March 2004 09:41:45:
Als Antwort auf: / In reply to: Thinker PV geschrieben von: / posted by: Gábor Szõts at 01 March 2004 07:37:21:

Hi,
There must be something wrong with Thinker's evaluation values under WinBoard. The values given for the PV are very small - I estimate they are about 4 times smaller than normal. Is this due to a wrong interpretation of data sent by the engine to WinBoard?
This was already reported by Rober Allgeuer.
Gábor

Hi,
there was a response by Lance on this subject. In fact in Thinker he scales down the material score in order to match the level of his positional scores, because scaling up the positional scores would have meant significantly more effort.
What it boils down, however, IMHO is that whatever Thinker outputs in its PV is not in pawn units - independent of the GUI. This does not interfere with Thinker's playing strength, but unfortunately this makes Thinker not suitable for analysis.
Robert

[Gábor Szõts]

Geschrieben von: / Posted by: Gábor Szõts at 01 March 2004 09:50:09:
Als Antwort auf: / In reply to: Re: Thinker PV geschrieben von: / posted by: Robert Allgeuer at 01 March 2004 09:41:45:

Hi,
There must be something wrong with Thinker's evaluation values under WinBoard. The values given for the PV are very small - I estimate they are about 4 times smaller than normal. Is this due to a wrong interpretation of data sent by the engine to WinBoard?
This was already reported by Rober Allgeuer.
Gábor

Hi,
there was a response by Lance on this subject. In fact in Thinker he scales down the material score in order to match the level of his positional scores, because scaling up the positional scores would have meant significantly more effort.
What it boils down, however, IMHO is that whatever Thinker outputs in its PV is not in pawn units - independent of the GUI. This does not interfere with Thinker's playing strength, but unfortunately this makes Thinker not suitable for analysis.
Robert

I don't understand this at all. Why do others not need this hocus-pocus?
Not only that. When the score fluctuates around equality, it is much more difficult to see if Thinker thinks he is ahead or behind.
Gábor

[Matthias Gemuh]

Geschrieben von: / Posted by: Matthias Gemuh at 01 March 2004 11:39:44:
Als Antwort auf: / In reply to: Re: Thinker PV geschrieben von: / posted by: Robert Allgeuer at 01 March 2004 09:41:45:

Hi,
there was a response by Lance on this subject. In fact in Thinker he scales down the material score in order to match the level of his positional scores, because scaling up the positional scores would have meant significantly more effort.
What it boils down, however, IMHO is that whatever Thinker outputs in its PV is not in pawn units - independent of the GUI. This does not interfere with Thinker's playing strength, but unfortunately this makes Thinker not suitable for analysis.
Robert

"Unfortunately this makes Thinker not suitable for analysis" is simply right.
I hope Lance will come back to the normal way of scaling.

/Matthias.


BigLion + Taktix