pichy wrote:I also tested Critter 1.4, Houdini, Spark, Spike 1.4 and none of those underpromote to Rook in the given position of the Saavedra study( Rook versus pawn). I believe that the main reason is that the author of those programs thought that the Queen covers all the moves that the rook covers plus the diagonals, but for this case promoting to a Queen only get the engine to draw (stalemate) instead of winning
[d]8/8/1KP5/3r4/8/8/8/k7 w - - 0 1
Are you sure that underpromotion is the right move? The line from tablebase shows a queen promotion.
Hello mr Vida, sorry if i reply but the underpromotion to rook is the right solution. Of course black can avoid to manouevre keep on checking but in that way this position would be lost anyway driving in a easy endgame Rook vs Queen.
In other words, black tries to give always check to white king speculating that white king cannot go the the ''c'' file cause Rook d1! and draw but after all the checkes, white king goes to c2 and the only move for black is Rook d4 (idea c8= Q, Rook d4 +, Queen x d4 stalemate!)
Of course black can change this variation but in this study black is lost anyway.
pichy wrote:I also tested Critter 1.4, Houdini, Spark, Spike 1.4 and none of those underpromote to Rook in the given position of the Saavedra study( Rook versus pawn). I believe that the main reason is that the author of those programs thought that the Queen covers all the moves that the rook covers plus the diagonals, but for this case promoting to a Queen only get the engine to draw (stalemate) instead of winning
[d]8/8/1KP5/3r4/8/8/8/k7 w - - 0 1
Are you sure that underpromotion is the right move? The line from tablebase shows a queen promotion.
Hello mr Vida, sorry if i reply but the underpromotion to rook is the right solution. Of course black can avoid to manouevre keep on checking but in that way this position would be lost anyway driving in a easy endgame Rook vs Queen.
In other words, black tries to give always check to white king speculating that white king cannot go the the ''c'' file cause Rook d1! and draw but after all the checkes, white king goes to c2 and the only move for black is Rook d4 (idea c8= Q, Rook d4 +, Queen x d4 stalemate!)
Of course black can change this variation but in this study black is lost anyway.
Best Regards and wishes for your engine
Yes, I understand, but most engines discard black's stalemate plan because it leads to an even faster mate if white underpromotes. That is why c8=R does not show up in PV.
Komodo is an exception because it thinks black can force a draw.
pichy wrote:I also tested Critter 1.4, Houdini, Spark, Spike 1.4 and none of those underpromote to Rook in the given position of the Saavedra study( Rook versus pawn). I believe that the main reason is that the author of those programs thought that the Queen covers all the moves that the rook covers plus the diagonals, but for this case promoting to a Queen only get the engine to draw (stalemate) instead of winning
[d]8/8/1KP5/3r4/8/8/8/k7 w - - 0 1
Are you sure that underpromotion is the right move? The line from tablebase shows a queen promotion.
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
I played it out versus Komodo, letting komodo search up to depth 32 in each move and look at the result:
pichy wrote:I also tested Critter 1.4, Houdini, Spark, Spike 1.4 and none of those underpromote to Rook in the given position of the Saavedra study( Rook versus pawn). I believe that the main reason is that the author of those programs thought that the Queen covers all the moves that the rook covers plus the diagonals, but for this case promoting to a Queen only get the engine to draw (stalemate) instead of winning
[d]8/8/1KP5/3r4/8/8/8/k7 w - - 0 1
Are you sure that underpromotion is the right move? The line from tablebase shows a queen promotion.
pichy wrote:I also tested Critter 1.4, Houdini, Spark, Spike 1.4 and none of those underpromote to Rook in the given position of the Saavedra study( Rook versus pawn). I believe that the main reason is that the author of those programs thought that the Queen covers all the moves that the rook covers plus the diagonals, but for this case promoting to a Queen only get the engine to draw (stalemate) instead of winning
[d]8/8/1KP5/3r4/8/8/8/k7 w - - 0 1
Are you sure that underpromotion is the right move? The line from tablebase shows a queen promotion.
Hello mr Vida, sorry if i reply but the underpromotion to rook is the right solution. Of course black can avoid to manouevre keep on checking but in that way this position would be lost anyway driving in a easy endgame Rook vs Queen.
In other words, black tries to give always check to white king speculating that white king cannot go the the ''c'' file cause Rook d1! and draw but after all the checkes, white king goes to c2 and the only move for black is Rook d4 (idea c8= Q, Rook d4 +, Queen x d4 stalemate!)
Of course black can change this variation but in this study black is lost anyway.
Best Regards and wishes for your engine
Yes, I understand, but most engines discard black's stalemate plan because it leads to an even faster mate if white underpromotes. That is why c8=R does not show up in PV.
Komodo is an exception because it thinks black can force a draw.
You are right, it was more or less what i tried to mean but you were more clear than me.
As regards Komodo, sadly i think their programmers underestimate too much the underpromotion (in general) although it happens not so rarely even in real games (especially to knight).
pichy wrote:I also tested Critter 1.4, Houdini, Spark, Spike 1.4 and none of those underpromote to Rook in the given position of the Saavedra study( Rook versus pawn). I believe that the main reason is that the author of those programs thought that the Queen covers all the moves that the rook covers plus the diagonals, but for this case promoting to a Queen only get the engine to draw (stalemate) instead of winning
[d]8/8/1KP5/3r4/8/8/8/k7 w - - 0 1
Are you sure that underpromotion is the right move? The line from tablebase shows a queen promotion.
I don't think that the tablebase will show that the first move for Black is Rd3? since the correct move is Rd6+ and continue checking etc... watch the video on youtube
pichy wrote:I don't think that the tablebase will show that the first move for Black is Rd3? since the correct move is Rd6+ and continue checking etc... watch the video on youtube
pichy wrote:I also tested Critter 1.4, Houdini, Spark, Spike 1.4 and none of those underpromote to Rook in the given position of the Saavedra study( Rook versus pawn). I believe that the main reason is that the author of those programs thought that the Queen covers all the moves that the rook covers plus the diagonals, but for this case promoting to a Queen only get the engine to draw (stalemate) instead of winning
[d]8/8/1KP5/3r4/8/8/8/k7 w - - 0 1
Are you sure that underpromotion is the right move? The line from tablebase shows a queen promotion.
I don't think that the tablebase will show that the first move for Black is Rd3? since the correct move is Rd6+ and continue checking etc... watch the video on youtube
Richard Vida is right....the continuos checking leads to a quick checkmate; for an engines it's better to go in a slow lost endgame queen against rook.
pichy wrote:I also tested Critter 1.4, Houdini, Spark, Spike 1.4 and none of those underpromote to Rook in the given position of the Saavedra study( Rook versus pawn). I believe that the main reason is that the author of those programs thought that the Queen covers all the moves that the rook covers plus the diagonals, but for this case promoting to a Queen only get the engine to draw (stalemate) instead of winning
[d]8/8/1KP5/3r4/8/8/8/k7 w - - 0 1
Are you sure that underpromotion is the right move? The line from tablebase shows a queen promotion.
I don't think that the tablebase will show that the first move for Black is Rd3? since the correct move is Rd6+ and continue checking etc... watch the video on youtube
Richard Vida is right....the continuos checking leads to a quick checkmate; for an engines it's better to go in a slow lost endgame queen against rook.
Best Regards
Continuous checking leads to a quick checkmate only if Black makes several inferior moves. Like many studies, this one needs to be revisited now that tablebases exist.
