Time controls for testing when changing evaluation weights

[AndrewGrant]

It seems to me that I should be able to test evaluation weight (changed values of PSQT, changed values of outposts... ect) with incredibly low time controls. Something like 1+.01s.

This will be my first time playing around with just evaluation weights, so I've never done this testing before. Is there any reason that 1+.01s won't cut it?

[bob]

It seems to me that I should be able to test evaluation weight (changed values of PSQT, changed values of outposts... ect) with incredibly low time controls. Something like 1+.01s.

This will be my first time playing around with just evaluation weights, so I've never done this testing before. Is there any reason that 1+.01s won't cut it?

All you really want is a time control where you don't see many losses on time...

[Ferdy]

It seems to me that I should be able to test evaluation weight (changed values of PSQT, changed values of outposts... ect) with incredibly low time controls. Something like 1+.01s.

This will be my first time playing around with just evaluation weights, so I've never done this testing before. Is there any reason that 1+.01s won't cut it?

What do you mean by the following?

Is there any reason that 1+.01s won't cut it?

[AndrewGrant]

What i mean is this:

Using 1+.01s time control will not be useful when you make a change regarding null pruning above depth 12. Or any other search features that are not triggered due to low time = low depth. That is a case I would say does NOT cut it

[Ferdy]

What i mean is this:

Using 1+.01s time control will not be useful when you make a change regarding null pruning above depth 12. Or any other search features that are not triggered due to low time = low depth. That is a case I would say does NOT cut it

Most programs use their eval at depth 0 or below, sometimes at higher depths. Even if tc is low, what matters are the eval values that are produced as a result of tuning. The tuned values should be tested at longer TC, can be TC 40/4 or TC 40/40 and others. There are eval parameters that would scale well with time and there are others that would not (king safety for example) probably depends on the program.