wonderingabout

i would always use the native engine whenever i can @renanpablocruz plus leela zero is known to have many blindspot issue both at low and high playouts :+1:

i was thinking about pondering too, but i thought maybe phoenixgo is smart i used 6x6 in .conf file of phoenixgo but on grp settings i used max variations 3,...

then i'll retry with same version and ponderingoff this time, same as this file : https://github.com/wonderingabout/PhoenixGo/blob/faqv2-bazel-master/etc/mcts_1gpu_grp.conf

oh wow @pnprog !!! its 1000 times better with pondering off lol http://www.mediafire.com/file/vcenr3jzq2vaxqs/game2v3.rsgf native phoenixgo 3200 simulations per move :  compare it with LZ-phoenixgo-orig -v 3200 :  now...

@alreadydone i have yet to check if they suggest the same moves and variations though

compare it to lz-phoenixgo-orig : https://github.com/Tencent/PhoenixGo/issues/86#issuecomment-456697268

now i want to test width 3 + depth 15 (i think seeing 3 alternate moves is reasonable + with enough diversity + they will be meaningful enough (enough simulations...

result v4 :+1: http://www.mediafire.com/file/83ajv4mytdw3ld3/game2v4.rsgf/file  compare it with LZ-phoenixgo-orig -v 3200 : 

isnt it what is already done here ? when i set variations to display to 3 ?

btw i sent you this yesterday @pnprog :+1: https://github.com/pnprog/goreviewpartner/pull/100 i think it will help clear away many of the confusion :)