diep wrote: D) Do not give the bonus at all
In Diep i'm not giving a bonus at all. Some 10+ years ago i tested many games with and without, and it didn't help at all, also to my surprise.
I have the same experience.
From my (old) to-do-list:
1. King-safety and passed pawn: subtract 10-15-20-25% from the score, often the next ply the opponent can flatten the score.
2. In a quiet position after a series of captures add a penalty. Not seldom trouble starts afterwards.
Are you saying that the side wishing to stand pat after a series of captures should take a penalty for doing so? It's easy enough to test this, but it's hard for me to see why a capture series should favor the side making the last capture.
I meant the other way around. The one doing the last capture and entering a quiet position (assuming you have this kind of information at your disposal) could be given a penalty. I am thinking of positions the search delays a loss by captures.
That seems backwards to me. A typical situation is that I fork or pin you, you make a capture which forces a recapture, and now you have reached the depth where you can stand pat, although in reality you are losing. So the side making the last capture should get a bonus in this case, not a penalty as you suggest. So I actually now like the idea as I originally interpreted it, not as you explained above. Am I wrong, or did you mean to give a penalty to the side on-move, which is the opposite of the side who made the last capture?
If I understand correctly, Komodo already does something similar. In qsearch it gives a small bonus when the last move was not a capture.
I forgot about that, because we got rid of it some months ago, not being able to confirm that it helped at all. So thanks for saving us the time of re-trying that idea. One of the drawbacks of getting older is forgetting things, such as this.
Have you found that any of the original ideas in Komodo has worked in Critter?
diep wrote: D) Do not give the bonus at all
In Diep i'm not giving a bonus at all. Some 10+ years ago i tested many games with and without, and it didn't help at all, also to my surprise.
I have the same experience.
From my (old) to-do-list:
1. King-safety and passed pawn: subtract 10-15-20-25% from the score, often the next ply the opponent can flatten the score.
2. In a quiet position after a series of captures add a penalty. Not seldom trouble starts afterwards.
Are you saying that the side wishing to stand pat after a series of captures should take a penalty for doing so? It's easy enough to test this, but it's hard for me to see why a capture series should favor the side making the last capture.
I meant the other way around. The one doing the last capture and entering a quiet position (assuming you have this kind of information at your disposal) could be given a penalty. I am thinking of positions the search delays a loss by captures.
That seems backwards to me. A typical situation is that I fork or pin you, you make a capture which forces a recapture, and now you have reached the depth where you can stand pat, although in reality you are losing. So the side making the last capture should get a bonus in this case, not a penalty as you suggest. So I actually now like the idea as I originally interpreted it, not as you explained above. Am I wrong, or did you mean to give a penalty to the side on-move, which is the opposite of the side who made the last capture?
If I understand correctly, Komodo already does something similar. In qsearch it gives a small bonus when the last move was not a capture.
Yes, I think that was removed but used to be in the program. It is one of those things that is almost impossible to measure. I don't remember why we took it out, but I think we did.
It may be because as an engineer I"m always pushing for simplicity and reducing the complexity of the program so I don't like stuff in the program that I am not fairly sure is an improvement. Larry tends to be more superstitious than I am about stuff like this and trusts his intuition so he likes to keep things he believe might help based on credulity. So this might have been one of those that I won out. But I don't remember for sure.
Capital punishment would be more effective as a preventive measure if it were administered prior to the crime.
diep wrote: D) Do not give the bonus at all
In Diep i'm not giving a bonus at all. Some 10+ years ago i tested many games with and without, and it didn't help at all, also to my surprise.
I have the same experience.
From my (old) to-do-list:
1. King-safety and passed pawn: subtract 10-15-20-25% from the score, often the next ply the opponent can flatten the score.
2. In a quiet position after a series of captures add a penalty. Not seldom trouble starts afterwards.
Are you saying that the side wishing to stand pat after a series of captures should take a penalty for doing so? It's easy enough to test this, but it's hard for me to see why a capture series should favor the side making the last capture.
I meant the other way around. The one doing the last capture and entering a quiet position (assuming you have this kind of information at your disposal) could be given a penalty. I am thinking of positions the search delays a loss by captures.
That seems backwards to me. A typical situation is that I fork or pin you, you make a capture which forces a recapture, and now you have reached the depth where you can stand pat, although in reality you are losing. So the side making the last capture should get a bonus in this case, not a penalty as you suggest. So I actually now like the idea as I originally interpreted it, not as you explained above. Am I wrong, or did you mean to give a penalty to the side on-move, which is the opposite of the side who made the last capture?
I somehow think we expressing the same idea and code
bob wrote: No attacked pieces and such, I believe that is the task of the search and q-search to resolve, hence the name "static evaluation" which excludes "dynamic considerations"
Threats evaluation is used to avoid erroneously stand pat when some threat could be pending.
That's fine, but I don't call that "evaluation". That is a trigger that says "this position is not quiet, we need to let search sort it all out, and give the evaluation code a quiet position it understands."
bob wrote: No attacked pieces and such, I believe that is the task of the search and q-search to resolve, hence the name "static evaluation" which excludes "dynamic considerations"
Threats evaluation is used to avoid erroneously stand pat when some threat could be pending.
That's fine, but I don't call that "evaluation". That is a trigger that says "this position is not quiet, we need to let search sort it all out, and give the evaluation code a quiet position it understands."
"Static Evaluation" just means the eval is not calculating moves ahead, it certainly should not exclude "dynamic considerations", by any chess definition of dynamic I can think of. The evaluation function is suppose to give us a rough prediction of future winning chance from a given position. The higher the value, the more winning chance (that is why our search works when trying to aim for the leaves with the highest score). If attacking pieces translates into an improved probability of winning (an intuitive outcome) even in situations where the Q-search does not see an immediately profitable capture, then it should give a bonus in eval.
I do find it useful to keep in mind that the eval score is more important in leaves, and what is a leaf or not is quite affected by q-search, but it doesn't pay to over-weight this issue when designing the eval function in my opinion.
-Sam
p.s. I hate it how spell checkers are always translating eval -> veal. It makes typing things like this a bother.
diep wrote: D) Do not give the bonus at all
In Diep i'm not giving a bonus at all. Some 10+ years ago i tested many games with and without, and it didn't help at all, also to my surprise.
I have the same experience.
From my (old) to-do-list:
1. King-safety and passed pawn: subtract 10-15-20-25% from the score, often the next ply the opponent can flatten the score.
2. In a quiet position after a series of captures add a penalty. Not seldom trouble starts afterwards.
Are you saying that the side wishing to stand pat after a series of captures should take a penalty for doing so? It's easy enough to test this, but it's hard for me to see why a capture series should favor the side making the last capture.
I meant the other way around. The one doing the last capture and entering a quiet position (assuming you have this kind of information at your disposal) could be given a penalty. I am thinking of positions the search delays a loss by captures.
That seems backwards to me. A typical situation is that I fork or pin you, you make a capture which forces a recapture, and now you have reached the depth where you can stand pat, although in reality you are losing. So the side making the last capture should get a bonus in this case, not a penalty as you suggest. So I actually now like the idea as I originally interpreted it, not as you explained above. Am I wrong, or did you mean to give a penalty to the side on-move, which is the opposite of the side who made the last capture?
If I understand correctly, Komodo already does something similar. In qsearch it gives a small bonus when the last move was not a capture.
