longer King move

[lech]

[D]8/1p1b2k1/p1p1p1p1/2P1P2p/1P3P1P/P2B1K2/8/8 w - -
The problem in such endgames concerns Kings. They are too lazy and it is impossible to get the long threat.
The solution can be to make possible the king move longer (ONLY in threat variants and endgames).
The default move is x-x1, and longer (double) x-x1-x2 (y-y1-y2 for the opponent king).
Some limitations for longer move are:
a) the distance x-x2 = 2
b) x1 and x2 no capture
c) x1 and x2 are not attacked
d) the distance beetwen kings y-x2 >= 4
Of course kings may play normal or longer move. It seems to be safe (threat) and only method to solve it.
What do you think about it?

[Uri Blass]

I do not think that it is a good idea because sometimes the long moves are not possible because the opponent may play a pawn move that prevent the long move.

It may be better to build some table for distance between squares in king moves in the endgame and extend lines when the king moves to the target in the smallest number of moves(when this smallest number is above some number for example above 2).

It is also possible to avoid null move pruning in lines when the king follow a plan(meaning making a "long move" in the smallest number of moves).

[lech]

I do not think that it is a good idea because sometimes the long moves are not possible because the opponent may play a pawn move that prevent the long move.

It makes rather nothing. It happens in a threat variant (plies from null < plies) and gives one threat more only. The threat variant is verified double if there is no threat and once if it is. It should be enough to verify any wrong effect of longer move.

[Don]

I do not think that it is a good idea because sometimes the long moves are not possible because the opponent may play a pawn move that prevent the long move.

It may be better to build some table for distance between squares in king moves in the endgame and extend lines when the king moves to the target in the smallest number of moves(when this smallest number is above some number for example above 2).

It is also possible to avoid null move pruning in lines when the king follow a plan(meaning making a "long move" in the smallest number of moves).

I experimented with "path distance" code in Komodo so that it has a real feel for intrusion squares. But it seems to be a slight regression even though it really seems like a sensible idea. Path distance assumes that only pawns are on the board so in a closed position, or a position like this the king is going to want to migrate to a5.

But in this position it needs to do much more than that - because even you advance the the position where the king is on a5 it takes a good bit of time to see the answer. Komodo solves the problem without any help but it takes 12 hours but without a long test I don't know if this helps or not. I assume that it will take at least a couple of less ply to solve as there will likely be less reductions on the path to getting to a5.

Don