I'm having a hard time believing Stockfish endgame technique sucks, what program are you running it in? Isn't there some way to improve it's "thinking" process by adjusting the depth of ply search, time spent on each move (or entire game) etc?
Try Houdini and Firebird, they're both free although I can't say if either one will work better for you.
A good old Shredder (7 or 8) might do the trick too. This engine is extremely strong in practical endgames.
For anyone still looking for this 13 years later:
Disable tablebase if you have it enabled and use Stockfish 16. Tablebase knows it is lost and goes for the quickest way to lose it seems.
Example:
7R/8/1K6/8/pk6/8/8/8 w - - 0 1
1. Rh4+ Kb3
2. Kb5 And here is where it just gives up: ...Kc3
Without the tablebase it still tries with ...Kb2 instead
I'm currently enjoying studying Silman's complete endgame course. I usually test myself against an engine (Stockfish) to see if I have learnt a particular endgame. The problem I'm having is that Stockfish plays lost (from the engine's perspective) or drawn endgames poorly. As you know, in many technical endgames there are certain tricks and traps to look out for such as stalemate possibilities etc, but Stockfish doesn't try to reach these position and trap me, and instead plays meaningless and harmless moves when it understands that the position is lost (or drawn) if I play accurately.
Is there an engine that better emulates a (strong) human player in these situations, that will try to set up these traps so that I can practise the methods for avoiding them?
Get Komodo 14 free. Komodo is better for endgame practice.
I'm currently enjoying studying Silman's complete endgame course. I usually test myself against an engine (Stockfish) to see if I have learnt a particular endgame. The problem I'm having is that Stockfish plays lost (from the engine's perspective) or drawn endgames poorly. As you know, in many technical endgames there are certain tricks and traps to look out for such as stalemate possibilities etc, but Stockfish doesn't try to reach these position and trap me, and instead plays meaningless and harmless moves when it understands that the position is lost (or drawn) if I play accurately.
Is there an engine that better emulates a (strong) human player in these situations, that will try to set up these traps so that I can practise the methods for avoiding them?