Hi,
I've just upgraded all stockfish nodes (82 instances!) to the very latest commit from development repository.
Not only analysis got better and faster, but we can now play ThreeCheck games with the computers, and also request computer analysis for any ThreeCheck game:
http://en.lichess.org/uC1nWNf7Enjoy!
not bad for a ... COMMUNIST WEBSITE !!!!
<jk, jk, I'm just kidding!> :)
Thanks to thibault, the Stockfish team, the scidb-Stockfish team, and 3+ developer ThyChief/ChiefTalu for bringing this together! As always, constructive feedback (preferably with repeatable tests and suggestions) is welcome!
Hooray!
I am so excited about this.
Thank you for the update!
This is a very nice addition. It gives up early checks too easily IMO. Still, it is very fun to play (and hard to calculate consequences). I like how my Nf5 changes its eval from -1.6 to #9 here!
http://en.lichess.org/Sk7H1Sj9#32A bit sad that it fell into a beginner's trap (2. d4?) when I gave it White.
http://en.lichess.org/XltGD77IL3zTMaybe upon rematch it was using "normal" Stockfish?
Well, it did the same (with reversed colors) when I asked for a fresh new game:
http://en.lichess.org/u0hs2ZF8This makes it less likely a rematch problem occurred the first time.
So maybe it's choice of randomness is faulty?
i feel like Christmas has come early! cheers thibault
#5 #6 #7 Stockfish's atomic variant cousin "atomkraft" suffers from a similar randomness/move ordering issue (which causes it to take longer to converge on the correct move):
http://en.lichess.org/forum/team-atomic-chess-theoreticians/can-someone-check-these-positions-with-atomkraftto what you observed on move 14:
http://en.lichess.org/Sk7H1Sj9#28and on move 1:
http://en.lichess.org/u0hs2ZF8#2I'll work on tuning parameters so in these positions Stockfish finds best moves faster. (During my testing I allocated too much thinking time to the engine & therefore didn't notice this.)