high values of pst

[Joost Buijs]

Hi,

This phenomena is something I see in my engine as well.
During the opening phase it produces very high scores, on many occasions the scores are getting so high that it will sacrifice a pawn without any real compensation.

In the current TCEC tourney my engine lost several games due to this.
My mistake is that I never tested the engine without opening book, and I never tuned the engine for opening play.

At the moment I'm looking at it, these high scores are indeed coming from the PST.
To solve this I have to scale the PST a lot lower, or the piece values a lot higher, during the opening phase.
This means that I have to retune everything.
I will probably leave it like it is, and spend my time to continue working on my new engine.

Sometimes I have the feeling that the whole concept of PST is flawed, and that the (added) value of a piece should solely be determined by the positional evaluation.

[cdani]

Sometimes I have the feeling that the whole concept of PST is flawed, and that the (added) value of a piece should solely be determined by the positional evaluation.

Of course is conceptually flawed. Is just a helper that with the advance of the engine is less and less desirable. Then is really difficult to remove it. Anyway it can help a little in stronger engines, for example to centralize king in endgames, or to do something when there is nothing to do.

In fact in my rewrite of Andscacs I did not put them on it.

[Robert Pope]

Hi all
I noticed in a lot of engines sources (and at CPW too) pst values are always so high.
I tried that in Isa and didn't work very well. With high values Isa " create " an artificial advantage say +1.00 or even +1.5 ( even higher sometimes lol) and lose .
Have you some expérience/advices with this phenomenon?
Is there some reasons for that?
Isa have a lot of parameters in his eval and I think it's the main reason

I tried to decrease some values yesterday and the scores are less crazy lol
Must be tune again
We'll see..
Bests
Dany

Someone pointed out that the levels in these tables are completely arbitrary. You can lower all the piece square values by 100 and then increase the material factor by 100 and get the exact same scores. Or eliminate the material weights entirely and represent them completely within the piece square tables.

[bob]

Hi all
I noticed in a lot of engines sources (and at CPW too) pst values are always so high.
I tried that in Isa and didn't work very well. With high values Isa " create " an artificial advantage say +1.00 or even +1.5 ( even higher sometimes lol) and lose .
Have you some expérience/advices with this phenomenon?
Is there some reasons for that?
Isa have a lot of parameters in his eval and I think it's the main reason

I tried to decrease some values yesterday and the scores are less crazy lol
Must be tune again
We'll see..
Bests
Dany

The value of a particular square (PST-wise) is not nearly so important as the delta between squares. Assuming symmetry with black and white, of course. Larger base PST values simply increase the basic material value of that piece. One could simply set all material values to zero, and set the queen PST to 900+something, assuming queen = 900 in your code, and you would get the exact same scores and skill level.

[Joost Buijs]

Hi all
I noticed in a lot of engines sources (and at CPW too) pst values are always so high.
I tried that in Isa and didn't work very well. With high values Isa " create " an artificial advantage say +1.00 or even +1.5 ( even higher sometimes lol) and lose .
Have you some expérience/advices with this phenomenon?
Is there some reasons for that?
Isa have a lot of parameters in his eval and I think it's the main reason

I tried to decrease some values yesterday and the scores are less crazy lol
Must be tune again
We'll see..
Bests
Dany

The value of a particular square (PST-wise) is not nearly so important as the delta between squares. Assuming symmetry with black and white, of course. Larger base PST values simply increase the basic material value of that piece. One could simply set all material values to zero, and set the queen PST to 900+something, assuming queen = 900 in your code, and you would get the exact same scores and skill level.

Indeed, the absolute value of the PST has no influence.

In my case the difference between the minimum and maximum values (delta) in the opening phase is too high with net result that my engine is sacrificing position and/or material in the very early moves of the game.
It is a mistake that I never tested the engine without opening book, and that I tuned most of the parameters with midgame positions only.

At the moment my PST has fixed values, so the easiest remedy would be to increase the piece values and the positional scores during the opening phase.
Not very difficult to do, but I probably have to retune everything.
I don't want to spend too much time on 15 year old code, I'm working on a new version of the engine which is 90% finished, I have the feeling it is better to spend my time on that.

[Daniel Anulliero]

Hi,

This phenomena is something I see in my engine as well.
During the opening phase it produces very high scores, on many occasions the scores are getting so high that it will sacrifice a pawn without any real compensation.

In the current TCEC tourney my engine lost several games due to this.
My mistake is that I never tested the engine without opening book, and I never tuned the engine for opening play.

At the moment I'm looking at it, these high scores are indeed coming from the PST.
To solve this I have to scale the PST a lot lower, or the piece values a lot higher, during the opening phase.
This means that I have to retune everything.
I will probably leave it like it is, and spend my time to continue working on my new engine.

Sometimes I have the feeling that the whole concept of PST is flawed, and that the (added) value of a piece should solely be determined by the positional evaluation.

Ok then I'm not alone ( sorry for you and Nightmare ) and tcec stage 1 was without book...
So I tried to decrease the values of pst by 2 (the delta seems better too)
Test running , we'll see...
Good luck with your new engine !
Thanks to all to your advices
Bests
Dany

[Daniel Anulliero]

Too late for édition. ..
Also I would say I never like the pst's concept very much ... Now I want to add more positional knowledges for pièces and pawns and Keep my tables at lower values
We'll see...

[thomasahle]

This is why I require the PST tables to sum to zero. Always nice to be able to remove some redundant dimensions when optimizing.

[DustyMonkey]

Who is doing automated PST tuning, and which direction are the values headed?

[D Sceviour]

While trying Texel's tuning method, I noticed a big reduction in material values due to PSQT. For example, the delta for a knight could be in the range of +/-25. Thus, the Texel tuning material value for knight averaged only 272 when using PSQT, but averaged 314 with PSQT turned off. As mentioned by others, I find continual adjustments seem to needed to reduce the values of PSQT. For kings, I use a center-gravity formula instead of PSQT. Also, PSQT is very dangerous to use in asymmetric positions. In knight versus bishop, the knight material value can be over-rated compared to the bishop due to PSQT. On the other hand, PSQT is very useful for scaling move ordering.