promotion: value of promote-to?

[flok]

Hi,

I was wondering: do the piece-types to what you promote a pawn to also have some specific value? Or is it the same value a piece has in general?
e.g. normally queen 9, rook 5
or is it totally depending on the situation?

[Stan Arts]

Hi Folkert,

Indeed the same value a piece has in general. (so, 9 for a queen, etc.)
You often get to keep the new piece, the promotion wins.

I'm not sure what you mean with depending on situation, a newly promoted piece is just as deadly as an original piece.
Even when it's fairly sure the promotion will be captured right away, it still needs the full value else the other side could incorrectly neglect the capture.

Regards,
Stan

[flok]

Hi Folkert,

Indeed the same value a piece has in general. (so, 9 for a queen, etc.)
You often get to keep the new piece, the promotion wins.

Ah yes, that's also what I found. Well mostly: it seems that a promote to rook is almost as powerfull as the queen:

Code: Select all

white n 54.664
white r 72.516
white b 42.031
white q 71.3619
black n 46.8379
black r 67.8354
black b 39.2912
black q 68.8133

first column: which color did the promotion, second colum promoted to what type, third column what % of time the game then was won.

I'm not sure what you mean with depending on situation, a newly promoted piece is just as deadly as an original piece. Even when it's fairly sure the promotion will be captured right away, it still needs the full value else the other side could incorrectly neglect the capture.

ok thanks![/code]

[hgm]

Indeed, statistics cn be very misleading. Last month I was on a party of people that had to run across a mine field in WWII. And guess what: they not only all survived it, but they all lived to an age of 85+ years. It must be extraodinary healty to run through mine fields...

[flok]

Well it definately might be correlated. I mean: the choice of chess-type to promote to can heavily influence the game.

[Edmund]

Promotion to Rook or Bishop makes only sense if you would get a draw, promoting to a queen ... promotion to knights is a more complicated issue.

These are very rare cases.

Most of the time one sees underpromotions in games, if it doesn't matter what piece to promote to. Either, because the piece is captured in the next move anyway or a mate is threatened.

[hgm]

Well, only computers seem to suffer from this. Humans always promote to Queen when the piece is immeditely lost anyway, to keep the opponent under maximum pressure to indeed take it.

The point is that it only makes sense to promote to Rook if you can see that a Rook obviously wins, and a Queen does not. If both draw, people would still prefer to choose the Queen. In the overwhelming number of positions where a Queen wins, and a Rook draws or loses, people would never even consider chosing a Rook. So the sample is extremely biased.

[Dann Corbit]

When you go through the promotion cycle, generate in this order:

Promote to queen
Promote to knight
Promote to rook
Promote to bishop

That is the order of frequency of the best promotion move.
Chess programs will often do silly underpromotions just because they see the chessman is going to get captured. E.g. "Haha! You only captured a rook and not a queen!"

For obvious reasons, the queen is the best move almost all of the time. Once in a while a promotion to knight will lead to checkmate when promotion to queen will not.

Promotion to rook and bishop have some rare value but it will add less than one Elo to your program which is why some programs never even bother implementing it (though I don't like it).

[Aleks Peshkov]

There are problems with real game chess wood multiple queens, especially in blitz.

I think chess would be more dynamic if we allowed to promote to only already captured pieces. Stronger side had to find a way to exchange queens before it can promote a new one.

[Sven]

This leads me to a question that I wanted to ask for a while already.

Has anyone an example of a position where promoting to a bishop is the best move, and definitely better than promoting to anything else?

This is my logic:

1. If promoting to a queen loses then promoting to a bishop can't be better.

2. If promoting to a queen draws then it is either
a) due to stalemate or
b) due to the fact that the promotion does not give check so that the opponent can enforce draw somehow.

In case a) promoting to a bishop can avoid stalemate but in order to win, the promoting side "W" must have more pieces. The only case I can imagine where the additional bishop is not redundant is that "W" has only a knight, so two knights would draw but bishop+knight win. But I failed to build a case with KNP-K where promoting to bishop avoids stalemate and promoting to rook or queen does not.

In case b) promoting to a bishop does not give check either, so it can't be better than queening.

3. If promoting to a queen wins then a bishop can't be better, obviously ...

Did anyone follow up to here? Did I miss anything?

To avoid any misunderstanding: even if my theory is right, I do not like to skip generating certain underpromotions, similar to Dann.

Sven