Testing methodology

[Edsel Apostol]

Hi Tony,

I don't know what you mean by this.

Off coarse, you give them for free.

About what you have suggested:

In your test setup, searching all non-capture moves the first 5 ply in quiescence will give good results.

It is equivalent to doing a depth 5 search so it defeat the purpose.

[Edsel Apostol]

I will try to test further on this setup. I am not yet convinced of your argument, though I am not dismissing it.

[Mangar]

Hi,

I agree with Aleks Peshkov;
Search and eval are interacting with each other. An eval for a verry shallow search will be best if it has many knowledge for tactics. The deeper the search is the less important are tactics in eval.

You can see this very well with your fruit example. Do the following test: play a tournament with depth 1 vs. fruit (that has a futility pruning in quiescese for loosing captures). Then play the same tournament against fruit on fixed depth 10. You will see that the pruning in quiescense does not harm fruit that much any more.

Greetings Volker

[Edsel Apostol]

Hi Volker,

Thanks for your input. Now it is clear to me. In short, a simple eval can be compensated by the quality of the search just like in Fruit. I am now back to my old tournament setup for testing.

By the way, I hope Spike is doing well.