velmarin wrote:With all the respect we all deserve.
This is for beginner player. look for something with more substance or more flavor.
There is no motor that does not see a pawn on seventh.
Except for Skipper.
Hi Jose.
I bet Bouquet could gain 10 elo with such parameter added, if it is well tuned.
Think about it, most engines have a term for pawns attacking enemy pieces, and derive measurable elo gain from that.
As far as I know, most engines also do not have a specific term for pawns defending own pieces, but the 2 are closely related and strictly non-redundant.
If the first term gives you some added value, chances are the second one could also do so. I do not know if it will work and where, but it is a reasonable eval term.
Eval helps to pick the right moves, and on occasion it matters if a piece is defended by a pawn or not. Couple of centipawns more or less could distinguish between different positions. It might not tell immediately, there might be not tactical resolution in sight, but the difference in eval will certainly tell at some point.
All the engines Ippolít have that in the code, it is childish, really.
You never looked at the evaluation of an Ippolít ?
velmarin wrote:With all the respect we all deserve.
This is for beginner player. look for something with more substance or more flavor.
There is no motor that does not see a pawn on seventh.
Except for Skipper.
Hi Jose.
I bet Bouquet could gain 10 elo with such parameter added, if it is well tuned.
Think about it, most engines have a term for pawns attacking enemy pieces, and derive measurable elo gain from that.
As far as I know, most engines also do not have a specific term for pawns defending own pieces, but the 2 are closely related and strictly non-redundant.
If the first term gives you some added value, chances are the second one could also do so. I do not know if it will work and where, but it is a reasonable eval term.
Eval helps to pick the right moves, and on occasion it matters if a piece is defended by a pawn or not. Couple of centipawns more or less could distinguish between different positions. It might not tell immediately, there might be not tactical resolution in sight, but the difference in eval will certainly tell at some point.
All the engines Ippolít have that in the code, it is childish, really.
You never looked at the evaluation of an Ippolít ?
Never.
Do they give the same bonus for all defended pieces?
Do they make a distinction between outposts and non-outposts?
Is there a rank-based approach?
What is the base value?
I have thought about this some.
I think the rule won't work.
Let's start with the queen. Here it is obvious that support by a pawn is irrelevant. Any piece to threaten a queen is less valueable than the target.
With rooks similar logic applies. E.g. in position 1 the pawn on b3 looks pretty ugly to human eyes. It shouldn't get a bonus for supporting a4.
Supporting a bishop with a pawn is potentially useful. But it also weakens its mobility. It is easy to imagine a scenario where this parameter would make an engine weaken its own bishops.
In general far-reaching pieces don't need support by pawns in general IMHO.
It might make some sense with knights though. Knights clearly gain in strength in general if supported by pawns, especially if we consider advanced posts that can't be attacked by enemy pawns.
Supporting a queen by a pawn isn't irrelevant in case the opponent also threatens with the queen. Same for the rook if it is threatened by a queen or rook. What you say about bishop mobility makes sense to me, but overall I think it is not a bad idea at all. In fact in SF there is such a bonus and it is proven to work, but only for already threatened pieces. So at least in SF there should be an overlap with an already existing bonus.
Peter Berger wrote:I have thought about this some.
I think the rule won't work.
Let's start with the queen. Here it is obvious that support by a pawn is irrelevant. Any piece to threaten a queen is less valueable than the target.
With rooks similar logic applies. E.g. in position 1 the pawn on b3 looks pretty ugly to human eyes. It shouldn't get a bonus for supporting a4.
Supporting a bishop with a pawn is potentially useful. But it also weakens its mobility. It is easy to imagine a scenario where this parameter would make an engine weaken its own bishops.
In general far-reaching pieces don't need support by pawns in general IMHO.
It might make some sense with knights though. Knights clearly gain in strength in general if supported by pawns, especially if we consider advanced posts that can't be attacked by enemy pawns.
Peter
A small bonus will never hurt.
As said, the bonus is given because of tactical considerations, as this saves you time for defence or escape, if one of your pieces is attacked.
If pawn-supported outposts give added value, why should not other pawn-supported pieces that are not minor outposts also not give?
For me, the point is to precisely tune the bonus in the different cases.
Queen or minor, it does not quite matter, it saves you time.
Lyudmil Tsvetkov wrote:Well, this is a very important rule.
It is always well when your pieces are defended, primarily because of tactical considerations, but engines usually do not consider this, as it is expensive. And indeed, defence among pieces themselves might be skipped, leaving to the search to do its job instead, however, I think pawn defence is an absolute must with a big added value.
Just to chime in here: the old version of Sjaak used to have a term like this, and I have recently added it to SjaakII as well. The test is still running, but results so far confirm what I remember:
1. In orthochess, this term is not worth very much. It currently sits at ~5 elo (with a 5cp bonus for a minor defended by a pawn).
2. In gothic chess, this term is worth considerably more. It currently sits at > 100 elo (the bonus depends on the number of super pieces, see below).
This is at short time controls and using self-play. The main difference between gothic chess and orthochess is that gothic chess has two extra super (queen-class) pieces: the Archbishop (B+N) and Chancellor (R+N), which makes it that much more dangerous to have undefended minors around. The scaling is with (#super + 1)/2, which can probably be tuned as well.
