Verify null move

[benstoker]

Would implementing null move verification benefit long time control games more than blitz games?

[michiguel]

Would implementing null move verification benefit long time control games more than blitz games?

It could, in theory, go both ways. More verification could mean a worse branch factor, hence, worse at long time controls. Lack of verification, could OTOH mean systematic errors (e.g. Zugzwang, threats etc.) are present in the search that will stay no matter how deep you search, making long searches less effective in certain cases.

Miguel

[diep]

Would implementing null move verification benefit long time control games more than blitz games?

it's called 'double nullmove' and real interesting to implement. eats more nodes in openingspositions and start middlegame, benefits in endgame, as it can avoid you failing to miss zugzwangs, meanwhile still being able to use everywhere nullmove (except for pawnendgame).

the principle is simple: never allow 3 consecutive nullmoves in a row.

e4 null null e5 allowed
e4 null Nf3 null null Nf6 null Nxe4 allowed
e4 null null null NOT allowed as it is 3 times in a row

Thanks,
Vincent

[benstoker]

Would implementing null move verification benefit long time control games more than blitz games?

it's called 'double nullmove' and real interesting to implement. eats more nodes in openingspositions and start middlegame, benefits in endgame, as it can avoid you failing to miss zugzwangs, meanwhile still being able to use everywhere nullmove (except for pawnendgame).

the principle is simple: never allow 3 consecutive nullmoves in a row.

e4 null null e5 allowed
e4 null Nf3 null null Nf6 null Nxe4 allowed
e4 null null null NOT allowed as it is 3 times in a row

Thanks,
Vincent

Better to implement only in the endgame phase?

A program that dare not speak its name has this feature now.

[metax]

I tried the double null move idea in prior versions of ChessMind but it didn't work for me. It ate some nodes but needed extremely high depths to see zugzwang situations so that it would hurt in 99% of the positions where there is no zugzwang, but take long time in positions where zugzwang was important.

My conclusion: _If_ you want to do a verification (I do not do this), rather use a verification search at depth >= 6 or something like that. Double null move could be worth a try, but for me it did not work very well, so I threw it out.