Stockfish underpromotes much more often than Komodo

[Jouni]

OK found some in Krabbe page finally may be only 3 correct:

8
7
6
5
4
3
2
1
	a	b	c	d	e	f	g	h

8/8/5b2/8/K7/8/kp6/3N4 b - - 0 0

Boniface - Pugh
Bristol 1995 -b1B wins!

[Karlo Bala]

I tried to see initially which promotes where, but the most striking result was about underpromotions. I equalized the engines to have similar strength. 10,000 games between K and SF without adjudication at close to 10'' + 0.1'' time control (adjusted for strength)

Komodo 10.1 generally promotes more often than Stockfish dev, an important thing in itself, but underpromotes almost an order of magnitude less:

Code: Select all

Komodo                 SF dev
promotions             promotions
(under-pr)             (under-pr)  

2484                   1901
(18)                   (141)

Here are the details on which square promotions (and under-promotions) occur:

Code: Select all

      Komodo    SF dev
a8    482(0)    357(17)    
b8    320(4)    261(12)  
c8    279(3)    227(17)   
d8    220(3)    168(19)  
e8    216(3)    175(20) 
f8    286(3)    184(16)   
g8    322(2)    227(16)  
h8    359(0)    302(24)

And here to what pieces underpromotions occur:

Code: Select all

Komodo:  N=9,  B=4,  R=5
SF dev:  N=35, B=13, R=93

If I am not doing something wrong (I didn't go through the games manually), it seems Stockfish underpromotes a lot, especially to Rook.

IMO there is a logical explanation for that. When engine promote to queen it is immediately taken, and after few forced moves line is cut off by the null move. However, after underpromotion, opponent may choose not to take promoted piece (because of SEE or something else), search deep and (if true score is close to beta) find that it is better not to take. After a fail low, the engine eventually finds that it is still better to take promoted piece. Because of flipping between fail high and fail low, underpromotion is saved in HT over the Q promotion as the best move, yet stay in PV.

[bob]

I tried to see initially which promotes where, but the most striking result was about underpromotions. I equalized the engines to have similar strength. 10,000 games between K and SF without adjudication at close to 10'' + 0.1'' time control (adjusted for strength)

Komodo 10.1 generally promotes more often than Stockfish dev, an important thing in itself, but underpromotes almost an order of magnitude less:

Code: Select all

Komodo                 SF dev
promotions             promotions
(under-pr)             (under-pr)  

2484                   1901
(18)                   (141)

Here are the details on which square promotions (and under-promotions) occur:

Code: Select all

      Komodo    SF dev
a8    482(0)    357(17)    
b8    320(4)    261(12)  
c8    279(3)    227(17)   
d8    220(3)    168(19)  
e8    216(3)    175(20) 
f8    286(3)    184(16)   
g8    322(2)    227(16)  
h8    359(0)    302(24)

And here to what pieces underpromotions occur:

Code: Select all

Komodo:  N=9,  B=4,  R=5
SF dev:  N=35, B=13, R=93

If I am not doing something wrong (I didn't go through the games manually), it seems Stockfish underpromotes a lot, especially to Rook.

Here's a key question. How often is that promotion immediately captured? Those should not be counted since the promotion is irrelevant regarding promotion piece, if the piece is immediately removed because it was unsafe where it sat.

[bob]

That makes sense.

SF is winning more games through king attacks, and Komodo endgames other engines more, hence Komodo tends to promote more frequently.

Considering underpromotions, I have seen those SF 'underpromotions', and yes, they are very funny, many other engines also do so, probably because of some bug in the search/SEE/MVV-LVA.

Almost all SF underpromotions to rook are in positions where a queen promotion is changes nothing, as the opponent is obliged to capture on the very next move, either the promoted queen, or the underpromoted rook.

SF just prefers that a smaller-value piece(the rook) is captured(attacked)

Nothing changes, but I find SF approach aesthetically uglier for a human to look at, though it basically changes nothing, just evoking a funny feeling...

MVV/LVA might introduce this to a point. I treat promotions as if they were captures. If you did this using the promoted piece, you would try knight before rook/queen since it is the "least valuable attacker". But I have noticed that many programs will choose something other than a queen on a futile promotion. For example, if you play f8=q and the opponent must recapture followed by your playing exf8=q#, then you could promote to rook and change nothing. And it is possible that the second promotion could also be to a rook if this is a back-rank type mate.

I don't think those should count. under-promoting to a knight to check is a little more common and often it is the only way to win. Promoting to a bishop rather than a queen is usually nonsensical unless it is a forced mate where the queen produces a stalemate.

[Ajedrecista]

Hello Jouni:

Can somebody post position, where bishop promotion is THE ONLY winning or drawing move? It is not so easy to find any!

What about this position?

8
7
6
5
4
3
2
1
	a	b	c	d	e	f	g	h

q7/1p3N1P/1p6/1P5p/7k/4p2P/4P2K/8 w - -

It is reached after playing Nf7+, Kh4 in this position:

8
7
6
5
4
3
2
1
	a	b	c	d	e	f	g	h

q6N/1p5P/1p6/1P4kp/8/4p2P/4P2K/8 w - -

https://chessdailynews.com/a-very-tricky-puzzle-2/

http://rybkaforum.net/cgi-bin/rybkaforu ... =1042;pg=2

Regards from Spain.

Ajedrecista.

[rabbits23]

Chessmaster 10 found the bishop promotion in both instances.
Allan

[Jouni]

Actually my first position was mate in 70 = DRAW as Nalimov says! But here
Kholmov - Ehlvest Wolgodonsk 1983

8
7
6
5
4
3
2
1
	a	b	c	d	e	f	g	h

2k5/2P5/1K2p3/7r/8/8/7p/R7 b - - 0 0

obviously -h1B wins. What says Lomosov? In rybkaforum position queen promotion also wins!

[MikeB]

8
7
6
5
4
3
2
1
	a	b	c	d	e	f	g	h

5r1k/7P/6P1/6KN/8/8/8/8 w - -

Only bishop wins...