If limited evaluation what would be your choice ?

[kiroje]

Hi

While fiddling and rewriting(for the billionth time) this question popped into my head :

If you could only choose 5 evaluation terms, what would they be ?

It should be in identical program with regards on search etc.

/Kim Jensen

[BubbaTough]

1. Value of Queen
2. Value of Rook
3. Value of Bishop
4. Value of Knight
5. Value of Pawn

[kiroje]

1. Value of Queen
2. Value of Rook
3. Value of Bishop
4. Value of Knight
5. Value of Pawn

Yes that would seem like a good start

Let me rephrase : What would be the 5 evaluation terms beside material value ?

/Kim

[Uri Blass]

Hi

While fiddling and rewriting(for the billionth time) this question popped into my head :

If you could only choose 5 evaluation terms, what would they be ?

It should be in identical program with regards on search etc.

/Kim Jensen

I am not sure what is the meaning of evaluation term.

Do you need different terms for the values of different pieces or can you use material difference as one evaluation term(or maybe you can use piece square table evaluation as one evaluation term)?

Uri

[kiroje]

Hi

While fiddling and rewriting(for the billionth time) this question popped into my head :

If you could only choose 5 evaluation terms, what would they be ?

It should be in identical program with regards on search etc.

/Kim Jensen

I am not sure what is the meaning of evaluation term.

Do you need different terms for the values of different pieces or can you use material difference as one evaluation term(or maybe you can use piece square table evaluation as one evaluation term)?

Uri

I was thinking something in the lines of

Rook on open file
Rook on halfopen file
Piece attacking lesser value piece
Double pawns
Isolated pawns

These "simple" evaluation terms was what I was thinking about.

/kim

[Uri Blass]

Hi

While fiddling and rewriting(for the billionth time) this question popped into my head :

If you could only choose 5 evaluation terms, what would they be ?

It should be in identical program with regards on search etc.

/Kim Jensen

I am not sure what is the meaning of evaluation term.

Do you need different terms for the values of different pieces or can you use material difference as one evaluation term(or maybe you can use piece square table evaluation as one evaluation term)?

Uri

I was thinking something in the lines of

Rook on open file
Rook on halfopen file
Piece attacking lesser value piece
Double pawns
Isolated pawns

These "simple" evaluation terms was what I was thinking about.

/kim

My opinion is that mobility is more important than pawn structure factors and number of squares that pieces control is more important.

Uri

[Eelco de Groot]

Hi

While fiddling and rewriting(for the billionth time) this question popped into my head :

If you could only choose 5 evaluation terms, what would they be ?

It should be in identical program with regards on search etc.

/Kim Jensen

Hello Kim,

I could say what evaluation terms would be my ideal, if you were to approach your question from the other end, 5(or 6 or so) possible evaluation terms, but without respect to their simplicity... I have this note in my copy of the Toga Log Manual under the paragraph about Piece Square Tables, for my own reference:


4.3 Position Table
For each piece there is a table with 64 entries, one for each square. Each pieces position is just looked up in the appropriate table, and the score is summed up. I append the position tables at the end of the document (see Appendix).#²

#²: Net result is fairly small. Mainly intended to give direction absent other factors. But if there are other factors, tactical for instance, this [PST-]method possibly can’t be much enhanced. Other [more long term/strategical] factors [for piece evaluation] to consider would be Safety, Function, (Influence/Control), Overloading, Flexibility (development is similar, related: centre control), Cooperation between pieces, Plan and consistency.

The difficulty is implementing those terms of course, but it is a bit like my draft for a 'Positional Evaluation Prime Directive' for me, to stay in SF terms...

Eelco

[hgm]

I addressed exactly this question in micro-Max. The most important evaluation terms turned out to be:

* Centralization of light pieces (incl. King and Pawns)
* penalty for not having Pawn control over squares
* penallty for King not being in its 'original' place in the middle-game (c1 and g1 being considered original squares after losing castling rights)
* penalty for not having a Pawn directly in front of a King without castling rights
* bonuses for pusing Pawns (increasing during the game), including very large bonuses for having Pawns on 6th and 7th rank.

The most obvious missing thing after this is a penalty for doubled Pawns.

If you want to define 'evaluation term' more broadly, so that you could group some of the specific descriptions together, you could say the most important terms are centralization/mobility, Pawn structure, King safety, passer pushing.

Open files/7th-rank bonuses did not seem to have very much impact. I did not investigate more complicated terms, like undefended pieces when facing a Queen, and soft-pinned pieces. My expectation is that they are very important, but also very costly.

[Harald Johnsen]

I was thinking something in the lines of

Rook on open file
Rook on halfopen file
Piece attacking lesser value piece
Double pawns
Isolated pawns

These "simple" evaluation terms was what I was thinking about.

/kim

Not so simple terms :
- material
- activity
- king safety
- pawn structure

HJ.

[mjlef]

Hi

While fiddling and rewriting(for the billionth time) this question popped into my head :

If you could only choose 5 evaluation terms, what would they be ?

It should be in identical program with regards on search etc.

/Kim Jensen

My guess:
material
simple "if you have no pawns you need a rook advantage to win-divider" endgame rule (amazing what a simple rule like this catches)
mobility
king safety
pawn structure

But I do not know. From time to time I try turning off some parts of the evaluation to see what effect they have. An interesting experiment would be to take a strong program (say Fruit/Toga) and run fixed depth matches with various evaluation terms turned off, then see how many ELO the terms get you. This would be a worst case test, since it would negate the cost (slowdown) a rule gives. Probably 5 ply searches would do for this test, and it would not even take very long. If you do this, I would love to see the impact of various terms. You could just add ifdefs around various terms and see what happens.

Mark