romantic-style play

[smcracraft]

I am always looking for ideas to make my chess program more "romantic" by making speculative sacrifices which lead into huge variations to confuse HUMAN (not computer) opponents, with the only player of the program being me, your 'umble programmer.

If anyone has suggestions on how to do SPECULATIVE sacrifices into large variation trees with many likely branches but not easy to compute in the usual 10-30 ply, I am all ears.

Thanks.

[syzygy]

Overevaluate the opponent's "king unsafety" by a lot.

[mclane]

Ask Chris w.

[Rebel]

1. What syzygy said.

2. From the final evaluation score subtract the material, what is left is the pure positional evaluation of the position. Multiply this value by a parameter and add it to the final evaluation score. The effect of a parameter value < 0.00 will result in a more defensive playing style while values > 0.00 will make an engine more aggressive. For instance, when 1.e4 normally scores (say) 0.20, a parameter of 0.30 will produce a score for 1. e4 as 0.26, a parameter of -0.60 and 1.e4 will be 0.08.

3. Count advantances. playing normal chess - when your king safety is better than the opponent PLUS your mobility is better PLUS you have a good passed pawn and your opponent has not PLUS you have a strong knight output PLUS you have the bishop pair than usually you have a won position. The basic idea is to count the number of advantages and apply a bonus with the number of advantages as an index to a table. Do this for white and black.

[chrisw]

I am always looking for ideas to make my chess program more "romantic" by making speculative sacrifices which lead into huge variations to confuse HUMAN (not computer) opponents, with the only player of the program being me, your 'umble programmer.

If anyone has suggestions on how to do SPECULATIVE sacrifices into large variation trees with many likely branches but not easy to compute in the usual 10-30 ply, I am all ears.

Thanks.

Is a question of balance between immediate dynamism and long term static value. Your rook may have a static score of 500 centipawns, but that is a measure of “potential” energy. You also measure its current kinetic energy, what it can immediately do.
Most of the features you need are in the SF eval already. You just need to change the weightings. More for kinetic, less for potential. Sorry for the somewhat obvious answer.
Practically speaking, you could try texel tuning everything, save those weightings, then tune again but only with certain dynamic features. Texel will try and pull apart those dynamic weights and since it has nothing else to work with it will take them more extreme. Then add the extreme weights you just calculated with the weights you saved for everything else.