Jump to content

  • Log In with Google      Sign In   
  • Create Account


#Actualpatishi

Posted 16 July 2013 - 08:52 AM

right now, from checking the time i get an answer from the engine there is not a crucial difference between the TT with the single entry array &&  always replace scheme, and the scheme i am using now  (deep+always).    like i said,so far,the only big difference i noticed is using the killer moves instead of the history heuristic (killer moves are better), but there is a chance that my HH implementation is not the "best".     

 

as for the TT , so far i haven't been using the hash move,  and right now i am working on implementing it,  to see if this will speed up the process.   
I must say that now,  after i implemented the 8 ply "book" moves (e.g database) my engine is playing perfect, and against week engines, the answer of the first engine move (the 9th ply) is getting pretty fast, something like 5 6 seconds..  but i want it to be faster!    
 


#3patishi

Posted 16 July 2013 - 07:34 AM

right now, from checking the time i get an answer from the engine there is not a crucial difference between the TT with the single entry array &&  always replace scheme, and the scheme i am using now  (deep+always).    like i said,so far,the only big difference i noticed is using the killer moves instead of the history heuristic (killer moves are better), but there is a chance that my HH implementation is not the "best".     

 

as for the TT , so far i haven't been using the hash move,  and right now i am working on implementing it,  to see if this will speed up the process.   
I must say that now,  after i implemented the 8 ply "book" moves (e.g database) my engine is playing perfect, and against week engines, the answer of the first engine move (the 9th ply) is getting pretty fast, something like 5 6 seconds..  but i want it to be faster!    
but when testing the engine against a perfect player in black,  it somehow suddenly taked much (much!) more time for the first move.  sometimes even 30+ seconds.  
this is because that there are not a lot of cut offs when facing strong moves (correct me if i am wrong).           so i  hope that the hash move will help solving this


#2patishi

Posted 16 July 2013 - 07:33 AM

right now, from checking the time i get an answer from the engine there is not a crucial difference between the TT with the single entry array &&  always replace scheme, and the scheme i am using now  (deep+always).    like i said,so far,the only big difference i noticed is using the killer moves instead of the history heuristic (killer moves are better), but there is a chance that my HH implementation is not the "best".     

 

as for the TT , so far i haven't been using the hash move,  and right now i am working on implementing it,  to see if this will speed up the process.   
I must say that now,  after i implemented the 8 ply "book" moves (e.g database) my engine is playing perfect, and against week engines, the answer of the first engine move (the 9th ply) is getting pretty fast, something like 5 6 seconds..  but i want it to be faster!    
but when testing the engine against a perfect player in black,  it somehow suddenly taked much (much!) more time for the first move.  sometimes even 30+ seconds.  
this is because that there are not a lot of cut offs when facing a strong engine (correct me if i am wrong).           so i am hope that the hash move will help solving this


#1patishi

Posted 16 July 2013 - 07:32 AM

right now, from checking the time i get an answer from the engine there is not a crucial difference between the TT with the single entry array &&  always replace scheme, and the scheme i am using now  (deep+always).    like i said,so far,the only big difference i noticed is using the killer moves instead of the history heuristic (killer moves are better), but there is a change that my HH implementation is not the "best".     

 

as for the TT , so far i haven't been using the hash move,  and right now i am working on implementing it,  to see if this will speed up the process.   
I must say that now,  after i implemented the 8 ply "book" moves (e.g database) my engine is playing perfect, and against week engines, the answer of the first engine move (the 9th ply) is getting pretty fast, something like 5 6 seconds..  but i want it to be faster!    
but when testing the engine against a perfect player in black,  it somehow suddenly taked much (much!) more time for the first move.  sometimes even 30+ seconds.  
this is because that there are not a lot of cut offs when facing a strong engine (correct me if i am wrong).           so i am hope that the hash move will help solving this


PARTNERS