Why most engines can't find this mate in 3 ?

[Martin Hertz]

[d] 1R6/pPk1r3/2p5/K7/8/2B1b3/8/8 w - - 0 1

Ka6 draws, but almost no engine can find it...

But if you move Ka6 by yourself and let the engine think after this, they will find a draw-score nearly instantly !

I've tested some DOS programs, for finding Ka6 in a reasonable time @ 3.3 GHz.

Chess Genius 3 needs about 7 minutes (depth 15).

Also Wchess 1.05 finds it, but the time varies very strong on each try. The fastest try was under
10 seconds for switching to Ka6 (depth 14) and showing +0.00 in about 40 seconds (depth 16).

[Codesquid]

[d] 1R6/pPk1r3/2p5/K7/8/2B1b3/8/8 w - - 0 1

Ka6 draws, but almost no engine can find it...

But if you move Ka6 by yourself and let the engine think after this, they will find a draw-score nearly instantly !

I really wonder why this happens. In particular Ka6 is found quickly if I let Octochess search all root moves with a full window.

[Marek Soszynski]

I'm running octochess-linux-generic-r4984 within Scid 4.4, and after 492.75s (depth 26) it still prefers 1.Rg8.