Are PST necessary for a good HCE?

Discussion of chess software programming and technical issues.

Moderator: Ras

KhepriChess
Posts: 93
Joined: Sun Aug 08, 2021 9:14 pm
Full name: Kurt Peters

Are PST necessary for a good HCE?

Post by KhepriChess »

In my mind, and I could be very wrong here, it seems like it would be possible to encourage good play (through a good evaluation) by coding in the various ideas for pieces and patterns. I get that PST implicitly do this, but don't they also add a level of arbitrariness?

Is it just that PST are an easy way to add those things (rather than a potentially longer eval function to individually add various patterns, etc.) and the search function will ruling out any quirks introduced by potential arbitrary square scores?

Could an effective HCE eval be done without PSTs?
Puffin: Github
KhepriChess: Github
syzygy
Posts: 5694
Joined: Tue Feb 28, 2012 11:56 pm

Re: Are PST necessary for a good HCE?

Post by syzygy »

KhepriChess wrote: Fri Mar 17, 2023 6:46 pm In my mind, and I could be very wrong here, it seems like it would be possible to encourage good play (through a good evaluation) by coding in the various ideas for pieces and patterns. I get that PST implicitly do this, but don't they also add a level of arbitrariness?

Is it just that PST are an easy way to add those things (rather than a potentially longer eval function to individually add various patterns, etc.) and the search function will ruling out any quirks introduced by potential arbitrary square scores?

Could an effective HCE eval be done without PSTs?
You could implement knight mobility by looking up a mobility value in an array indexed by the knight's current square. Oh, we have invented PSTs :)

PSTs are evaluation patterns like any other evaluation pattern.
smatovic
Posts: 3225
Joined: Wed Mar 10, 2010 10:18 pm
Location: Hamburg, Germany
Full name: Srdja Matovic

Re: Are PST necessary for a good HCE?

Post by smatovic »

...some people pointed out that PSTs with Texel Tuning Method are a kind of SLP NN with backpropagation, the precursor of NNUE so to speak.

--
Srdja
abulmo2
Posts: 465
Joined: Fri Dec 16, 2016 11:04 am
Location: France
Full name: Richard Delorme

Re: Are PST necessary for a good HCE?

Post by abulmo2 »

Olithink is a relatively strong program with a mobility based evaluation and no PST, so you can avoid PST and still have a competitive engine. Note that Olithink's evaluation is 100% hand crafted; whereas strong PST based engines, like PeSTO or my more modest engine Dumb, are automatically tuned by computer in a very similar way done for NNUE evaluations.
Richard Delorme
Mike Sherwin
Posts: 965
Joined: Fri Aug 21, 2020 1:25 am
Location: Planet Earth, Sol system
Full name: Michael J Sherwin

Re: Are PST necessary for a good HCE?

Post by Mike Sherwin »

The PST can also be algorithmically created before each search like is done in RomiChess. Just do it better than I did.