On 10/29/2010 10:47 PM, Joost 't Hart wrote:
Attached a patch. Seems to work fine. Please try and confirm...
I will commit it iff successful.
Cheers,
Joost.
On 10/29/2010 09:19 PM, Roy Brunjes wrote:
Hi!
Hilarious indeed. Your wording is slightly unclear, but you are right.
How to reproduce:
Enter the following 3 games in the clipbase
A 1.e4 e5 2.Nf3 Ba3 3.Nxa3
B 1.e4 e5 2.Ba6 Nxa6
C 1.e4 e4 2.Nf3 Nc6
Open sf and game A, and annotate games 1 to 3...
Surely black's A2..Ba3 is nagged "with ?? +-" and a +5 eval
At the start of game B, sf is totally confused.
B1. e4 is nagged as "?? =" and sf produces the variation 1.Na3 Nf6
2.d4 Ne4 +-. Lol, wow!
From your text I would expect the same problem to occur at the start
of game C, but it does not.
I will see what I can do shortly. Somehow things are not reset
properly on the transition from one game to the next....
The problems are not related to the Find Opening Errors thing, since
in the above I analyzed "complete" games.
Cheers,
Joost.
I need to correct things a bit. Actually, it's not Black's 20th move
and White's first move (1. e4) in the next game that cause the
problem. It's White's 20th move in game 1 and White's 1. e4 in game 2
that can cause this as well as Black's 20th move in game 1 and
Black's 1st move in game 2. I just saw an example where 1. e4 is
rated as within the blunder threshold but 1... c5 was more than 0.5
worse than the previous game where Black's 20th move was highly rated
by the engine.
Hope that's clear...
----- Forwarded Message ----
*From:* Roy Brunjes <roy_brun...@yahoo.com>
*To:* scid-users@lists.sourceforge.net
*Sent:* Fri, October 29, 2010 12:12:09 PM
*Subject:* SCID - Annotate using "find opening errors" - bug
I took a collection of games and wanted to analyse the openings
(defined by me as first 20 moves of a game) of the games in this
particular database. I selected "find opening errors" and set that
parameter to 20. I also set the blunder threshold to 0.5 pawns. I set
the options to analyze both white and black moves but to not analyze
any variations that may be present.
As the engine (Stockfish-1.9.1 in my case) analyses away, when it
moves from the end of one game (it has analysed the first 20 moves of
the first game and thus skips the remainder of that game and starts
analysing the next game in the database. Sometimes the eval generated
for the first move of the second game is significantly different from
the eval of the previous game's last analysed position (black's move
20 of game 1 had eval of 0.8 for example and white's first move of
game 2 (let's say the move is "e4") is perhaps 0.2 and this triggers
the "this move (1. e4) is a blunder because it is more than 0.5 pawns
worse than the previous move.
This obviously is a bug and I would appreciate thoughts as to when
this might be fixed.
Thanks,
Roy
<zip>
Index: tcl/main.tcl
===================================================================
RCS file: /cvsroot/scid/scid/tcl/main.tcl,v
retrieving revision 1.48
diff -U3 -r1.48 main.tcl
--- tcl/main.tcl 30 Jun 2010 20:41:45 -0000 1.48
+++ tcl/main.tcl 29 Oct 2010 22:25:59 -0000
@@ -1667,12 +1667,14 @@
if { $::isBatch && $annotateMode && $::isBatchOpening && \
[sc_pos moveNumber] > $::isBatchOpeningMoves } {
sc_game save [sc_game number]
+ toggleEngineAnalysis 1 1
if {[sc_game number] < $::batchEnd} {
sc_game load [expr [sc_game number] + 1]
if {$::addAnnotatorTag} {
appendAnnotator " $analysis(name1)"
}
set ::wentOutOfBook 0
+ toggleEngineAnalysis 1 1
updateMenuStates
updateStatusBar
updateTitle
@@ -1706,12 +1708,14 @@
}
if {$::isBatch && [sc_game number] != 0} {
sc_game save [sc_game number]
+ toggleEngineAnalysis 1 1
if {[sc_game number] < $::batchEnd} {
sc_game load [expr [sc_game number] + 1]
if {$::addAnnotatorTag} {
appendAnnotator " $analysis(name1)"
}
set ::wentOutOfBook 0
+ toggleEngineAnalysis 1 1
updateMenuStates
updateStatusBar
updateTitle
------------------------------------------------------------------------------
Nokia and AT&T present the 2010 Calling All Innovators-North America contest
Create new apps & games for the Nokia N8 for consumers in U.S. and Canada
$10 million total in prizes - $4M cash, 500 devices, nearly $6M in marketing
Develop with Nokia Qt SDK, Web Runtime, or Java and Publish to Ovi Store
http://p.sf.net/sfu/nokia-dev2dev
_______________________________________________
Scid-users mailing list
Scid-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/scid-users
