"Agressive Chess" not equal to Lethal chess.

[fern]

Once and again a thread appears with the following issue: which engine past or present could be lined as the most aggressive in his playing behavior?
Let me offer a clue to solve the question with something more than just feelings about that, which are valuable but not very precise.
In the first place, an engine capable of cutting us to pieces due to phenomenal tactical power is NOT aggressive by that sole simple fact. ANY engine these days can do that, more or less. Some of them can do it faster because are better and so they does not miss a chance to kill us.
But aggression goes along a different path. In my book, an aggressive player is the one that takes chances to corner us against the wall choosing not the most accurate moves, but even bluffing a bit.
It is in that sense that CSTAL is the most aggressive creature. It is NOT better tactician than any other engine, in fact on the contrary is less good on that than ANY current top30 engine. So it can happen that this aggressive entity can take more time to kill you than a conventional current engine. Nevertheless, you feel it as lot more aggressive because he try all the time to push you. He goes beyond the calculations, he throw on you a cloud of dust, he threaten you and that make of it , to this day, the more funny chess engine to play. And the more aggressive.

My opinion, just that
Fern

[Don]

Once and again a thread appears with the following issue: which engine past or present could be lined as the most aggressive in his playing behavior?
Let me offer a clue to solve the question with something more than just feelings about that, which are valuable but not very precise.
In the first place, an engine capable of cutting us to pieces due to phenomenal tactical power is NOT aggressive by that sole simple fact. ANY engine these days can do that, more or less. Some of them can do it faster because are better and so they does not miss a chance to kill us.
But aggression goes along a different path. In my book, an aggressive player is the one that takes chances to corner us against the wall choosing not the most accurate moves, but even bluffing a bit.
It is in that sense that CSTAL is the most aggressive creature. It is NOT better tactician than any other engine, in fact on the contrary is less good on that than ANY current top30 engine. So it can happen that this aggressive entity can take more time to kill you than a conventional current engine. Nevertheless, you feel it as lot more aggressive because he try all the time to push you. He goes beyond the calculations, he throw on you a cloud of dust, he threaten you and that make of it , to this day, the more funny chess engine to play. And the more aggressive.

My opinion, just that
Fern

I think the English word you are looking for is "provactively." An interesting engine wants to "provoke" you into making a mistake and setting all sorts of problems in front of you that you must solve.

A thought experiment: If you had a program that could search full width to the end of them game and thus play "perfect" chess, would it be as effective as one that is simply very strong (but not perfect) but knows how to play "provactively?"

I think a perfect program might tend to simplify and draw too much - it would not know how to complicate the position or increase it's winning chances in position that are technically drawn. Imagine a position that is very difficult to defend but still a draw with best play. The "perfect" program would know it's a draw, assume best play by the opponent and might easily fail to keep the pressure on, perhaps quickly simplifying to something that is easy to draw.

So if such a program were ever created we might have to modify it to play more provactive moves when it does not affect the game theoretic result.

[fern]

Thanks for your comments, Don. Yes, provocative and proactive are the exact words here instead of aggressive. And as you say a perfect machine would be the pinnacle of boredom.
Why we play chess after all? To get amused facing challenges, being attacked, attacking ourselves, face dangers, etc, as seeing a thriller in the cinema.
In all that I still believe CSTAl was a paradigm.
What about Komodo?
I cannot say yet. I have not played enough games. It is too strong and if weakened, then it is not Komodo anymore.
Perhaps you could add a function to make of it more spe3culative in the sense you described?

my best
Fern

[Don]

Thanks for your comments, Don. Yes, provocative and proactive are the exact words here instead of aggressive. And as you say a perfect machine would be the pinnacle of boredom.
Why we play chess after all? To get amused facing challenges, being attacked, attacking ourselves, face dangers, etc, as seeing a thriller in the cinema.
In all that I still believe CSTAl was a paradigm.
What about Komodo?
I cannot say yet. I have not played enough games. It is too strong and if weakened, then it is not Komodo anymore.
Perhaps you could add a function to make of it more spe3culative in the sense you described?

my best
Fern

You can make a program more sacrificial quite easily. Primarily one would lower all the material values in relation to the other things, especially the dynamic features such as mobility and king safety. While lowering all the piece values you would lower the non-pawn values a bit more. In addition you could look at the standard sacrificial motifs and enhance the program to prefer getting into the attacking side of them. You might lower the scoring for weak pawns a bit in addition.

It might require more than just changing a few values to do it well, we might have to make adjustments to prefer different piece postings over others a bit more.

If your goal is not absolute soundness, I don't think this would be that difficult to do. Some of the changes I suggested would not just cause it to sacrifice more often but it would increase the aggressiveness of the program too in the general case. Lowering features like pawn weaknesses would cause it to prefer more aggressive postings of pieces and whether sacrifices were realized or not the program would "threaten" to sacrifice more so you would have to be more careful.

I'm pretty sure Komodo has most of this about right. In most gambit positions Komodo does see signficant compensation, but not quite enough which pretty much agrees with grandmaster theory. So it would be easy to push it a little bit more in that direction.

[rbarreira]

Thanks for your comments, Don. Yes, provocative and proactive are the exact words here instead of aggressive. And as you say a perfect machine would be the pinnacle of boredom.
Why we play chess after all? To get amused facing challenges, being attacked, attacking ourselves, face dangers, etc, as seeing a thriller in the cinema.
In all that I still believe CSTAl was a paradigm.
What about Komodo?
I cannot say yet. I have not played enough games. It is too strong and if weakened, then it is not Komodo anymore.
Perhaps you could add a function to make of it more spe3culative in the sense you described?

my best
Fern

You can make a program more sacrificial quite easily. Primarily one would lower all the material values in relation to the other things, especially the dynamic features such as mobility and king safety. While lowering all the piece values you would lower the non-pawn values a bit more. In addition you could look at the standard sacrificial motifs and enhance the program to prefer getting into the attacking side of them. You might lower the scoring for weak pawns a bit in addition.

It might require more than just changing a few values to do it well, we might have to make adjustments to prefer different piece postings over others a bit more.

If your goal is not absolute soundness, I don't think this would be that difficult to do. Some of the changes I suggested would not just cause it to sacrifice more often but it would increase the aggressiveness of the program too in the general case. Lowering features like pawn weaknesses would cause it to prefer more aggressive postings of pieces and whether sacrifices were realized or not the program would "threaten" to sacrifice more so you would have to be more careful.

I'm pretty sure Komodo has most of this about right. In most gambit positions Komodo does see signficant compensation, but not quite enough which pretty much agrees with grandmaster theory. So it would be easy to push it a little bit more in that direction.

What about asymmetry in the eval? For example, if the program had very different weights for king safety depending on the king's side (Komodo or opponent), wouldn't that be the simplest way to increase its aggressiveness?

[Don]

Thanks for your comments, Don. Yes, provocative and proactive are the exact words here instead of aggressive. And as you say a perfect machine would be the pinnacle of boredom.
Why we play chess after all? To get amused facing challenges, being attacked, attacking ourselves, face dangers, etc, as seeing a thriller in the cinema.
In all that I still believe CSTAl was a paradigm.
What about Komodo?
I cannot say yet. I have not played enough games. It is too strong and if weakened, then it is not Komodo anymore.
Perhaps you could add a function to make of it more spe3culative in the sense you described?

my best
Fern

You can make a program more sacrificial quite easily. Primarily one would lower all the material values in relation to the other things, especially the dynamic features such as mobility and king safety. While lowering all the piece values you would lower the non-pawn values a bit more. In addition you could look at the standard sacrificial motifs and enhance the program to prefer getting into the attacking side of them. You might lower the scoring for weak pawns a bit in addition.

It might require more than just changing a few values to do it well, we might have to make adjustments to prefer different piece postings over others a bit more.

If your goal is not absolute soundness, I don't think this would be that difficult to do. Some of the changes I suggested would not just cause it to sacrifice more often but it would increase the aggressiveness of the program too in the general case. Lowering features like pawn weaknesses would cause it to prefer more aggressive postings of pieces and whether sacrifices were realized or not the program would "threaten" to sacrifice more so you would have to be more careful.

I'm pretty sure Komodo has most of this about right. In most gambit positions Komodo does see signficant compensation, but not quite enough which pretty much agrees with grandmaster theory. So it would be easy to push it a little bit more in that direction.

What about asymmetry in the eval? For example, if the program had very different weights for king safety depending on the king's side (Komodo or opponent), wouldn't that be the simplest way to increase its aggressiveness?

Yes, asymmetry would probably be a major component of this, probably combined with some of the changes I mentioned already. For aggressive play you want the program to be willing to go on the hunt without being unduly concerned about it's own position.

To make Komodo more aggressive you would not have to make major changes to the evaluation, just fairly minor tweaks. It might be possible to make it a little more aggressive without impacting the strength much if any. That is someone for Larry to determine.

[Ferdy]

Once and again a thread appears with the following issue: which engine past or present could be lined as the most aggressive in his playing behavior?
Let me offer a clue to solve the question with something more than just feelings about that, which are valuable but not very precise.
In the first place, an engine capable of cutting us to pieces due to phenomenal tactical power is NOT aggressive by that sole simple fact. ANY engine these days can do that, more or less. Some of them can do it faster because are better and so they does not miss a chance to kill us.
But aggression goes along a different path. In my book, an aggressive player is the one that takes chances to corner us against the wall choosing not the most accurate moves, but even bluffing a bit.
It is in that sense that CSTAL is the most aggressive creature. It is NOT better tactician than any other engine, in fact on the contrary is less good on that than ANY current top30 engine. So it can happen that this aggressive entity can take more time to kill you than a conventional current engine. Nevertheless, you feel it as lot more aggressive because he try all the time to push you. He goes beyond the calculations, he throw on you a cloud of dust, he threaten you and that make of it , to this day, the more funny chess engine to play. And the more aggressive.

My opinion, just that
Fern

You may post some sample model games if you have, I will attempt to create that personality . By the way what is your estimated elo rating range?

[Master Om]

Once and again a thread appears with the following issue: which engine past or present could be lined as the most aggressive in his playing behavior?
Let me offer a clue to solve the question with something more than just feelings about that, which are valuable but not very precise.
In the first place, an engine capable of cutting us to pieces due to phenomenal tactical power is NOT aggressive by that sole simple fact. ANY engine these days can do that, more or less. Some of them can do it faster because are better and so they does not miss a chance to kill us.
But aggression goes along a different path. In my book, an aggressive player is the one that takes chances to corner us against the wall choosing not the most accurate moves, but even bluffing a bit.
It is in that sense that CSTAL is the most aggressive creature. It is NOT better tactician than any other engine, in fact on the contrary is less good on that than ANY current top30 engine. So it can happen that this aggressive entity can take more time to kill you than a conventional current engine. Nevertheless, you feel it as lot more aggressive because he try all the time to push you. He goes beyond the calculations, he throw on you a cloud of dust, he threaten you and that make of it , to this day, the more funny chess engine to play. And the more aggressive.

My opinion, just that
Fern

That is why i quoted Richard Vida that Critter 0.90 is not Aggressive rather Stockfish is. It is evident from its style.
There are 2 more engines who play like that. Zappa Mexico II at King safety = 500 and Rybka 3 Dynamic at contempt 35. Try and tell me.

[royb]

Thanks for your comments, Don. Yes, provocative and proactive are the exact words here instead of aggressive. And as you say a perfect machine would be the pinnacle of boredom.
Why we play chess after all? To get amused facing challenges, being attacked, attacking ourselves, face dangers, etc, as seeing a thriller in the cinema.
In all that I still believe CSTAl was a paradigm.
What about Komodo?
I cannot say yet. I have not played enough games. It is too strong and if weakened, then it is not Komodo anymore.
Perhaps you could add a function to make of it more spe3culative in the sense you described?

my best
Fern

You can make a program more sacrificial quite easily. Primarily one would lower all the material values in relation to the other things, especially the dynamic features such as mobility and king safety. While lowering all the piece values you would lower the non-pawn values a bit more. In addition you could look at the standard sacrificial motifs and enhance the program to prefer getting into the attacking side of them. You might lower the scoring for weak pawns a bit in addition.

It might require more than just changing a few values to do it well, we might have to make adjustments to prefer different piece postings over others a bit more.

If your goal is not absolute soundness, I don't think this would be that difficult to do. Some of the changes I suggested would not just cause it to sacrifice more often but it would increase the aggressiveness of the program too in the general case. Lowering features like pawn weaknesses would cause it to prefer more aggressive postings of pieces and whether sacrifices were realized or not the program would "threaten" to sacrifice more so you would have to be more careful.

I'm pretty sure Komodo has most of this about right. In most gambit positions Komodo does see signficant compensation, but not quite enough which pretty much agrees with grandmaster theory. So it would be easy to push it a little bit more in that direction.

What about asymmetry in the eval? For example, if the program had very different weights for king safety depending on the king's side (Komodo or opponent), wouldn't that be the simplest way to increase its aggressiveness?

Yes, asymmetry would probably be a major component of this, probably combined with some of the changes I mentioned already. For aggressive play you want the program to be willing to go on the hunt without being unduly concerned about it's own position.

To make Komodo more aggressive you would not have to make major changes to the evaluation, just fairly minor tweaks. It might be possible to make it a little more aggressive without impacting the strength much if any. That is someone for Larry to determine.

Don,

Though I'm confident you are focusing on MP right now, perhaps a few UCI parameters could be added to Komodo4.1 that would allow customers to tweak things as this thread discusses?

And, having that work in conjunction with an overall approximate ELO setting (play at roughly 1600 ELO for example) would allow a customer to have a "normal", aggressive, or even passive 1600 ELO computer opponent to play against.

Roy

[fern]

Well, then, as greeks used to say: "hic Rodas, Hic salta".

a Tal-style or even better, Andersen kind of Komodo would be a delight to play.

Fern