Don Dailey you need to teach underpromotion to Komodo!

[MM]

There are clearly positions where promotion to rook is the best move, but unlike promotion to knight, they do not occur frequently enough in real games to be an elo contributor (or even elo neutral). Of course, supporting it is a very very small elo hit, and my instinct would be to support it if I was commercial engine developer (which I am not ).

-Sam

Every programmer, commercial or not, should always consider underpromotions, no matter if it implies elo gain or not. Chess engines, especially commercial, should consider any variable of chess, sometimes elo is not important.

Best Regards

[zullil]

There are clearly positions where promotion to rook is the best move, but unlike promotion to knight, they do not occur frequently enough in real games to be an elo contributor (or even elo neutral). Of course, supporting it is a very very small elo hit, and my instinct would be to support it if I was commercial engine developer (which I am not ).

-Sam

Every programmer, commercial or not, should always consider underpromotions, no matter if it implies elo gain or not. Chess engines, especially commercial, should consider any variable of chess, sometimes elo is not important.

Best Regards

Commercial products are subject to the preferences of the market. If consumers care about lack of underpromotions in q-searches, they can avoid buying engines with such "deficiencies."

[pichy]

There are clearly positions where promotion to rook is the best move, but unlike promotion to knight, they do not occur frequently enough in real games to be an elo contributor (or even elo neutral). Of course, supporting it is a very very small elo hit, and my instinct would be to support it if I was commercial engine developer (which I am not ).

-Sam

You can say the same for using Tablebase or not using tablebase, since nowadays engines are so strong like Houdini, Critter, or Komodo that you don't have the need to use tablebase to increase the elo.

PS: If all that programmers care about is a small Elo gain in order to have a small edge over other engines then they should not support Rook underpromotion since it does not occur so frefuently, but people do NOT only purchase a program because it is a little bit stronger then X program, unless they are only going to use it to match it against other engines. But what about discouraging people buying an engine that does not know how to beat them with White in a position like the Saavedra endgame study.

And here is the famous Saavedra ending study game that I played versus Komodo:
[d]8/8/1KP5/3r4/8/8/8/k7 w - - 0 1
[Event " Rook vs Pawn-Ending"]
[Site "?"]
[Date "2012.03.29"]
[Round "?"]
[White "Pichy"]
[Black "Komodo3-64"]
[Result "1-0"]
[BlackElo "3000"]
[WhiteElo "2200"]
[SetUp "1"]
[FEN "8/8/1KP5/3r4/8/8/8/k7 w - - 0 1"]
[WhiteType "human"]
[BlackType "program"]


1. c7 Rd6+ 2. Kb5 Rd5+ 3. Kb4 Rd4+ 4. Kb3 Rd3+ 5. Kc2 Rd4 6. c8=R Rd2+ 7.
Kxd2 Kb2 8. Rc3 Kb1 9. Rb3+ Ka1 10. Kc3 Ka2 {Black resigns} 11. Kc2 Ka1
{Black resigns} 12. Ra3# *

[zullil]

But what about discouraging people buying an engine that does not know how to beat them with White in a position like the Saavedra endgame study.

[d]8/8/1KP5/3r4/8/8/8/k7 w - - 0 1

You play Black. Choose an engine for White, with no tablebases if you prefer. Play a game from this position. Can you ever get a draw? If so, please post the game.

[Robert Flesher]

There are clearly positions where promotion to rook is the best move, but unlike promotion to knight, they do not occur frequently enough in real games to be an elo contributor (or even elo neutral). Of course, supporting it is a very very small elo hit, and my instinct would be to support it if I was commercial engine developer (which I am not ).

-Sam

You can say the same for using Tablebase or not using tablebase, since nowadays engines are so strong like Houdini, Critter, or Komodo that you don't have the need to use tablebase to increase the elo.

PS: If all that programmers care about is a small Elo gain in order to have a small edge over other engines then they should not support Rook underpromotion since it does not occur so frefuently, but people do NOT only purchase a program because it is a little bit stronger then X program, unless they are only going to use it to match it against other engines. But what about discouraging people buying an engine that does not know how to beat them with White in a position like the Saavedra endgame study.

And here is the famous Saavedra ending study game that I played versus Komodo:
[d]8/8/1KP5/3r4/8/8/8/k7 w - - 0 1
[Event " Rook vs Pawn-Ending"]
[Site "?"]
[Date "2012.03.29"]
[Round "?"]
[White "Pichy"]
[Black "Komodo3-64"]
[Result "1-0"]
[BlackElo "3000"]
[WhiteElo "2200"]
[SetUp "1"]
[FEN "8/8/1KP5/3r4/8/8/8/k7 w - - 0 1"]
[WhiteType "human"]
[BlackType "program"]


1. c7 Rd6+ 2. Kb5 Rd5+ 3. Kb4 Rd4+ 4. Kb3 Rd3+ 5. Kc2 Rd4 6. c8=R Rd2+ 7.
Kxd2 Kb2 8. Rc3 Kb1 9. Rb3+ Ka1 10. Kc3 Ka2 {Black resigns} 11. Kc2 Ka1
{Black resigns} 12. Ra3# *

I am not familiar with this study. However, c7 leads to queen versus rook? Why is this even a study? White wins with ease!

[michiguel]

There are clearly positions where promotion to rook is the best move, but unlike promotion to knight, they do not occur frequently enough in real games to be an elo contributor (or even elo neutral). Of course, supporting it is a very very small elo hit, and my instinct would be to support it if I was commercial engine developer (which I am not ).

-Sam

You can say the same for using Tablebase or not using tablebase, since nowadays engines are so strong like Houdini, Critter, or Komodo that you don't have the need to use tablebase to increase the elo.

PS: If all that programmers care about is a small Elo gain in order to have a small edge over other engines then they should not support Rook underpromotion since it does not occur so frefuently, but people do NOT only purchase a program because it is a little bit stronger then X program, unless they are only going to use it to match it against other engines. But what about discouraging people buying an engine that does not know how to beat them with White in a position like the Saavedra endgame study.

And here is the famous Saavedra ending study game that I played versus Komodo:
[d]8/8/1KP5/3r4/8/8/8/k7 w - - 0 1
[Event " Rook vs Pawn-Ending"]
[Site "?"]
[Date "2012.03.29"]
[Round "?"]
[White "Pichy"]
[Black "Komodo3-64"]
[Result "1-0"]
[BlackElo "3000"]
[WhiteElo "2200"]
[SetUp "1"]
[FEN "8/8/1KP5/3r4/8/8/8/k7 w - - 0 1"]
[WhiteType "human"]
[BlackType "program"]


1. c7 Rd6+ 2. Kb5 Rd5+ 3. Kb4 Rd4+ 4. Kb3 Rd3+ 5. Kc2 Rd4 6. c8=R Rd2+ 7.
Kxd2 Kb2 8. Rc3 Kb1 9. Rb3+ Ka1 10. Kc3 Ka2 {Black resigns} 11. Kc2 Ka1
{Black resigns} 12. Ra3# *

I am not familiar with this study. However, c7 leads to queen versus rook? Why is this even a study? White wins with ease!

No, you cannot win promoting to Queen. 6. c8=Q Rc4+ 7. Qxc4 stalemate
After 6. c8=R Ra4 (forced) 7. Kb3 wins

Miguel

[Robert Flesher]

There are clearly positions where promotion to rook is the best move, but unlike promotion to knight, they do not occur frequently enough in real games to be an elo contributor (or even elo neutral). Of course, supporting it is a very very small elo hit, and my instinct would be to support it if I was commercial engine developer (which I am not ).

-Sam

You can say the same for using Tablebase or not using tablebase, since nowadays engines are so strong like Houdini, Critter, or Komodo that you don't have the need to use tablebase to increase the elo.

PS: If all that programmers care about is a small Elo gain in order to have a small edge over other engines then they should not support Rook underpromotion since it does not occur so frefuently, but people do NOT only purchase a program because it is a little bit stronger then X program, unless they are only going to use it to match it against other engines. But what about discouraging people buying an engine that does not know how to beat them with White in a position like the Saavedra endgame study.

And here is the famous Saavedra ending study game that I played versus Komodo:
[d]8/8/1KP5/3r4/8/8/8/k7 w - - 0 1
[Event " Rook vs Pawn-Ending"]
[Site "?"]
[Date "2012.03.29"]
[Round "?"]
[White "Pichy"]
[Black "Komodo3-64"]
[Result "1-0"]
[BlackElo "3000"]
[WhiteElo "2200"]
[SetUp "1"]
[FEN "8/8/1KP5/3r4/8/8/8/k7 w - - 0 1"]
[WhiteType "human"]
[BlackType "program"]


1. c7 Rd6+ 2. Kb5 Rd5+ 3. Kb4 Rd4+ 4. Kb3 Rd3+ 5. Kc2 Rd4 6. c8=R Rd2+ 7.
Kxd2 Kb2 8. Rc3 Kb1 9. Rb3+ Ka1 10. Kc3 Ka2 {Black resigns} 11. Kc2 Ka1
{Black resigns} 12. Ra3# *

I am not familiar with this study. However, c7 leads to queen versus rook? Why is this even a study? White wins with ease!

No, you cannot win promoting to Queen. 6. c8=Q Rc4+ 7. Qxc4 stalemate
After 6. c8=R Ra4 (forced) 7. Kb3 wins

Miguel

I glanced at the position without playing any moves. I am checking the line Jorge provided now to see if it is forced. I am now very interested.

[Robert Flesher]

There are clearly positions where promotion to rook is the best move, but unlike promotion to knight, they do not occur frequently enough in real games to be an elo contributor (or even elo neutral). Of course, supporting it is a very very small elo hit, and my instinct would be to support it if I was commercial engine developer (which I am not ).

-Sam

You can say the same for using Tablebase or not using tablebase, since nowadays engines are so strong like Houdini, Critter, or Komodo that you don't have the need to use tablebase to increase the elo.

PS: If all that programmers care about is a small Elo gain in order to have a small edge over other engines then they should not support Rook underpromotion since it does not occur so frefuently, but people do NOT only purchase a program because it is a little bit stronger then X program, unless they are only going to use it to match it against other engines. But what about discouraging people buying an engine that does not know how to beat them with White in a position like the Saavedra endgame study.

And here is the famous Saavedra ending study game that I played versus Komodo:
[d]8/8/1KP5/3r4/8/8/8/k7 w - - 0 1
[Event " Rook vs Pawn-Ending"]
[Site "?"]
[Date "2012.03.29"]
[Round "?"]
[White "Pichy"]
[Black "Komodo3-64"]
[Result "1-0"]
[BlackElo "3000"]
[WhiteElo "2200"]
[SetUp "1"]
[FEN "8/8/1KP5/3r4/8/8/8/k7 w - - 0 1"]
[WhiteType "human"]
[BlackType "program"]


1. c7 Rd6+ 2. Kb5 Rd5+ 3. Kb4 Rd4+ 4. Kb3 Rd3+ 5. Kc2 Rd4 6. c8=R Rd2+ 7.
Kxd2 Kb2 8. Rc3 Kb1 9. Rb3+ Ka1 10. Kc3 Ka2 {Black resigns} 11. Kc2 Ka1
{Black resigns} 12. Ra3# *

I am not familiar with this study. However, c7 leads to queen versus rook? Why is this even a study? White wins with ease!

No, you cannot win promoting to Queen. 6. c8=Q Rc4+ 7. Qxc4 stalemate
After 6. c8=R Ra4 (forced) 7. Kb3 wins

Miguel

Very nice! A clever attempt at a swindle. I am convinced and ever surprised how simple looking positions can have such beautiful gems.

[bob]

Here's the final test results. R13 is the best current version. R14 is R13 but with rook/bishop under-promotions removed from the basic search (except at the root). Never had 'em in q-search so that is unchanged for both..

Code: Select all

   Crafty-23.5R13-1     2638    4    4 30000   63%  2531   22%
   Crafty-23.5R14-1     2638    4    4 30000   63%  2531   22%
   Crafty-23.5R13-2     2637    4    4 30000   63%  2531   22%
   Crafty-23.5R14-2     2637    4    4 30000   63%  2531   22%

Seems to make absolutely no difference...

[pichy]

Here's the final test results. R13 is the best current version. R14 is R13 but with rook/bishop under-promotions removed from the basic search (except at the root). Never had 'em in q-search so that is unchanged for both..

Code: Select all

   Crafty-23.5R13-1     2638    4    4 30000   63%  2531   22%
   Crafty-23.5R14-1     2638    4    4 30000   63%  2531   22%
   Crafty-23.5R13-2     2637    4    4 30000   63%  2531   22%
   Crafty-23.5R14-2     2637    4    4 30000   63%  2531   22%

Seems to make absolutely no difference...

Where can I download the version that consider Rook underpromotion instead of Queen