Greetings.
I have been experimenting with the Annotate feature and have
some requests. Keep in mind, Annotate is working really well for
me, so these requests are in no way complaints. I see them as
potentially making a good program better. Roughly in order of
least to most difficult to implement:
1. Annotation "Move out of book" has a slight wobble at end of book.
2. Annotate... dialog "Add variations" new option "Never".
3. Annotate... dialog "Find opening errors" new option "Start-from"
move.
4. Annotate reverse the analysis (game-end back to beginning).
5. Annotate... dialog "Time per move in sec." accept and use 0.
Details follow:
1. Annotation "Move out of book" has a slight wobble at end of book.
This is a really long section for a very rare and minor case.
Feel free to skip over it if it's too boring.
When the game under Annotate is at the end of the game AND at
the end of the book, then instead of marking "Last book move"
Scid marks "Move out of book", backs up a move, and adds a
variation. I noticed this when I used the Book Tuning | Export
to create a game from a book, and then, _without changing the
location_ from the end of the game, used the Annotate feature
to create scores for all the moves. As a guess, Scid is seeing
the End-of-Game marker as out of book, which strictly speaking
it is.
Steps to reproduce:
terminal
================================================================
$ cat << 'EOF' > veryshort.pgn
> [Result "1/2-1/2"]
>
> 1.e4 e5 2.Nf3 Nc6 3.Nc3 Nf6 4.a3 g6
> 1/2-1/2
>
> EOF
$ cat veryshort.pgn
[Result "1/2-1/2"]
1.e4 e5 2.Nf3 Nc6 3.Nc3 Nf6 4.a3 g6
1/2-1/2
$ polyglot make-book -min-game 1 -pgn veryshort.pgn -bin veryshort.bin
PolyGlot 2.0.4 by Fabien Letouzey.
inserting games ...
2 game. ## off-by-one bug in polyglot
8 entries.
filtering entries ...
8 entries.
sorting entries ...
saving entries ...
all done!
$ ls -Gg veryshort.*
-rwxr-xr-x 1 180 May 30 12:58 veryshort.pgn
-rwxr-xr-x 1 128 May 30 13:00 veryshort.bin
$ cp veryshort.bin ~/scid_5_0_2/share/scid/books
$
================================================================
Scid
================================================================
- In the clipbase create a new game 1.e4 e5 2.Nf3 Nc6 3.Nc3 Nf6
4.a3 g6 *
- Leave the position after 4...g6 on the board
- In the Analysis Engine window click the Annotate... button
Annotate options
Game review Comments
Time per move in sec. [ 1.0 [^v]] [ ] Score all moves
< > Annotate all moves [ ] Annotate variations
<X> When game move is a blunder [ ] Short annotations
Threshold: 1.0 [ ] Add score to annotations
[X] Use book [ ] Add annotator tag
[veryshort.bin [v]] [ ] Mark tactical exercises
Add variations [ ] From actual game to game:
<X> For moves by both sides [1 [^v]]
< > For White moves only [ ] Find opening errors
< > For Black moves only [12 [^v]]
[OK] [Cancel]
- Click OK
- Results ==
1.e4 e5 2.Nf3 Nc6 3.Nc3 Nf6 4.a3 g6
{ Move out of book}
( 4...g6 {10:+0.69} 5.d4 exd4 6.Nxd4 Qe7 7.Ndb5 d6
8.Bg5 Bg7 9.Bc4 Be6 10.Nd5 $14 )
*
- Empty the clipbase
- In the clipbase create a new game 1.e4 e5 2.Nf3 Nc6 3.Nc3 Nf6
4.a3 g6 *
- << to the start position
- Click the Annotate... button , use the same Annotate options
- Click OK
- Results ==
1.e4 e5 2.Nf3 Nc6 3.Nc3 Nf6 4.a3 g6
{ Last book move}
*
================================================================
4...g6 should be "Last book move" as in the second scenario, not
"Move out of book" as in the first scenario.
2. Annotate... dialog "Add variations" new option "Never".
The use case for this is when annotating a pgn opening book, e.g.
we might want to add centipawn scores to the moves _without_
adding any new moves. Setting to annotate "When game move is a
blunder", Scid still adds a variation at the end of the book.
That's easy enough to delete. But setting to "Annotate all
moves", Scid will also add variations to internal moves, and in
a large pgn game which already has many variations these might
be overlooked.
3. Annotate... dialog "Find opening errors" new option "Start-from"
move.
In the Annotate... dialog choosing "From actual game to
game [ N]", we can operate in batch mode. Then in "Find opening
errors" we set a final move beyond which Scid will not check.
With a start move we could skip even more moves at the beginning,
speeding up the batch operation.
As an example, in Twic I open the Tree Window, and copy all the
games after 1.d4 Nf6 2.c4 e6 3.Nc3 Bb4 to the clipbase. Then I
use the Annotate feature with my Nimzo-Indian polyglot book. In
each game the engine is analyzing 1.d4, then analyzing 1...Nf6,
etc. This is not helpful. I want to skip all the moves up to
3...Bb4, analyzing and/or checking against my book starting
only at move 4. With other openings, e.g. Ruy Lopez, we might
start at move 8 or move 9.
4. Annotate reverse the analysis (game-end back to beginning).
There are two reasons for wanting this. First, the engine
analysis in reverse is slightly more accurate, due to cache hits
(as long as the GUI does not send ucinewgame at every move).
Second, checking for "Last book move" currently ignores
transpositions; backwards analysis will catch them. For example,
using the same veryshort.bin, and the game 1.a3 e5 2.e4 Nf6
3.Nc3 Nc6 4.Nf3 Bc5 *, current analysis (forwards) annotates
1.a3 as "Move out of book", whereas backwards analysis should
identify 4...Bc5 as "Move out of book". I see this in my
Nimzo-Indian book all the time, when a Twic game starts 1.c4
("Move out of book") despite the transposition back into the
Nimzo-Indian. That's useless to me, I want it to find the true
later "Move out of book".
5. Annotate... dialog "Time per move in sec." accept and use 0.
In my tests of Annotate, the minimum thinking time the engine
will use is 0.1 seconds. If we could set it to 0 and have the
engine actually think for 0 seconds, then Annotate with an
opening book could be used as a batch "Find Novelty". Well, it
can already be used that way now, but it takes a _long_ time to
go through a Twic. My Nimzo-Indian book goes up to move 20, so I
have to go that far for each game. I think the book checking is
done by the GUI and only after the "Move out of book" is the
engine asked to analyze. So if the GUI recognized 0 as "skip the
analysis" then it should take no time at all to run through the
games. Of course if the analysis is reversed as in (4), then
the logic should also be reversed and "Move out of book" stops
the engine analysis rather than starts it.
--
Alan
_______________________________________________
Scid-users mailing list
Scid-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/scid-users
