combining pvs with null move

[Robert Pope]

Hmmm it is indeed. I'll let it play a bit longer for a while. Am curious what will happen.
For the last 10k games the 3 elo has been steady by the way.

Personally I wouldn't bother with a 3 Elo patch at this level. There must be bigger gains that only take a fraction of the time to test.

The problem is finding out where those gains are. I gained more from one round of tuning than I did from a year of testing algorithm patches. Even things that seem like no-brainers, like adding passed pawn code, often don't seem to pan out for me.

[Evert]

The problem is finding out where those gains are. I gained more from one round of tuning than I did from a year of testing algorithm patches. Even things that seem like no-brainers, like adding passed pawn code, often don't seem to pan out for me.

Making sure other terms are appropriately tuned is an important precondition. They don't have to be perfect, but they do have to be reasonable, otherwise the evaluation just misleads. Note that algorithm improvements are small fish compared to the evaluation, epecially if the engine is weaker (you need a good enough evaluation for more advanced search improvements to pay off).
If passed pawn evaluation doesn't help, it may be that the implementation is buggy (for instance awarding a penalty instead of a bonus), the weight is too small or too big, or the term never has a chance to affect the outcome o the game because the engine has a critical weakness that makes it lose before getting to the end game.
To find out about the latter requires that you actually look at and interpret the games, and self-testing will not reveal the problem (being blind to the issue means the engine will also not exploit it).