| |
Main
Date: 09 Oct 2006 09:49:15
From: Wilma
Subject: q questions about Chessmaster 9000
|
Question 1: When CM9k annotates a games, it gives a sumy table at the
end of the game which includes "Total Error" and "Relevant Error." What do
these terms mean?
Question 2: In determining "CM agreement," CM9k uses a tolerance of .5
which strikes me as too liberal. I couldn't find a setting for the
tolerance in the CM.ini file. Is there any way for a user to change that
tolerance?
Wilma
|
|
| |
Date: 11 Oct 2006 22:23:10
From:
Subject: Re: q questions about Chessmaster 9000 - 1 attachment
|
Wilma wrote:
> Okay it does work. I was looking for additional information instead of
> differences in the table. My bad. With the [annotator], I get a lower
> percentage of agreement.
>
> Now that it works, tell me what I've done. :) I used
> [annotator]
> slightly=5
> better=25
> much=50
>
> What do these numbers mean? What did I actually do?
>
> Wilma
The values that are used define the minimum difference in score (in
centipawns) between the move that was played and the move that CM
thinks is best for each annotation disagreement level.
So, using the above values, let's say that the move that CM thinks is
best is given a score of +1.00.
If the move that was actually played in the game results in a score
between +0.95 and +0.76, then the annotator will show a "slightly
better is..." annotation for the preferred move.
If the move that was played results in a score between +0.75 and +0.51,
then the annotator will show a "better is..." annotation for the
preferred move.
If the move that was played results in a score of +0.50 or lower, then
the annotator will show a "much better is..." annotation for the
preferred move.
Understand?
jm
|
| | |
Date: 12 Oct 2006 19:16:32
From: Wilma
Subject: Re: q questions about Chessmaster 9000 - 1 attachment
|
Excellent! Thanks for being here, John. We're pretty lucky in this
group.
Wilma
[email protected] wrote in news:1160630590.701909.18280
@h48g2000cwc.googlegroups.com:
>
> Wilma wrote:
>> Okay it does work. I was looking for additional information instead
of
>> differences in the table. My bad. With the [annotator], I get a lower
>> percentage of agreement.
>>
>> Now that it works, tell me what I've done. :) I used
>> [annotator]
>> slightly=5
>> better=25
>> much=50
>>
>> What do these numbers mean? What did I actually do?
>>
>> Wilma
>
> The values that are used define the minimum difference in score (in
> centipawns) between the move that was played and the move that CM
> thinks is best for each annotation disagreement level.
>
> So, using the above values, let's say that the move that CM thinks is
> best is given a score of +1.00.
>
> If the move that was actually played in the game results in a score
> between +0.95 and +0.76, then the annotator will show a "slightly
> better is..." annotation for the preferred move.
>
> If the move that was played results in a score between +0.75 and +
0.51,
> then the annotator will show a "better is..." annotation for the
> preferred move.
>
> If the move that was played results in a score of +0.50 or lower, then
> the annotator will show a "much better is..." annotation for the
> preferred move.
>
> Understand?
>
> jm
>
>
|
| |
Date: 11 Oct 2006 15:10:47
From:
Subject: Re: q questions about Chessmaster 9000 - 1 attachment
|
Wilma wrote:
> Don't know if it makes a difference, but the Annotator secion you had me
> add was added as the first 4 lines of Wilma.ini.
>
> [Annotator]
> slightly=25
> better=50
> much=100
>
> Wilma
Hi Wilma,
This is odd, because I just verified that it works correctly. Here is
exactly what I put, just to make the test very obvious:
[annotator]
slightly=5
better=25
much=50
I put this into Merlin.ini, which should correspond to your Wilma.ini.
Perhaps the fact that you capitalized "annotator"?
Either way, analysing the game that you sent gave me the following
results:
Default values:
White Black
CM9000 Agrees 36 33
CM9000 Disagrees 3 6
Agreement Pct. 92% 85%
Total Error 1.48 3.53
Relevant Error 1.48 3.53
Values adjusted to above numbers:
White Black
CM9000 Agrees 24 19
CM9000 Disagrees 15 20
Agreement Pct. 62% 49%
Total Error 5.00 9.22
Relevant Error 5.00 9.22
So it definitely should work.
jm
|
| | |
Date: 12 Oct 2006 02:11:42
From: Wilma
Subject: Re: q questions about Chessmaster 9000 - 1 attachment
|
Okay it does work. I was looking for additional information instead of
differences in the table. My bad. With the [annotator], I get a lower
percentage of agreement.
Now that it works, tell me what I've done. :) I used
[annotator]
slightly=5
better=25
much=50
What do these numbers mean? What did I actually do?
Wilma
> Hi Wilma,
>
> This is odd, because I just verified that it works correctly. Here is
> exactly what I put, just to make the test very obvious:
> [annotator]
> slightly=5
> better=25
> much=50
>
> I put this into Merlin.ini, which should correspond to your Wilma.ini.
> Perhaps the fact that you capitalized "annotator"?
>
> Either way, analysing the game that you sent gave me the following
> results:
>
> Default values:
> White Black
> CM9000 Agrees 36 33
> CM9000 Disagrees 3 6
> Agreement Pct. 92% 85%
> Total Error 1.48 3.53
> Relevant Error 1.48 3.53
>
> Values adjusted to above numbers:
> White Black
> CM9000 Agrees 24 19
> CM9000 Disagrees 15 20
> Agreement Pct. 62% 49%
> Total Error 5.00 9.22
> Relevant Error 5.00 9.22
>
> So it definitely should work.
>
> jm
>
>
|
|
