🚀 go-pugleaf

The current patchlevel 70 exhibit some undesirable alpha-beta          
inconsistencies.  Take a look at the following post output.



Your move is? Enter file name: 


White 1:00  Black 1:00

 - - R Q - K R -
 B P - - - P - -
 P - - P B - - -
 - - - - - n - P
 - p - P b - - p
 p - - - - - p -
 - - p - q p - -
 - - - r - r k -


Your move is? 
Your move is? 
Your move is? 
Move# 1    Target= 6000    Clock: 6000
 2.   -142    0     7620   d6d5  e4d3  e6f5  d3f5 
 3&   -142    0     7897   d6d5  e4d3  e6f5 
 3+    -31    0    11104   e6f5  c2c4  c8c4  e4f5  d4d3  d1d3  c4h4  e2h5 
 3&    133    0    17297   e6f5  e4f5  g8g3  f2g3  d4d3  e2f2  a7f2  f1f2 
                           d3c2  f5c2 
 3.    133    1    30227   e6f5  e4f5  g8g3  f2g3  d4d3  e2f2  a7f2  f1f2 
                           d3c2  f5c2 
 4+    228    1    39084   e6f5  e2c4 
 4-    220    4    98342   e6f5  e4f5  g8g7 
 4-    223    4    98560   g8g4 
 4-    223    4    99171   g8g5 
 4.    223    4   107422   g8g5 
 5-     35    5   126487   g8g5  h4g5  d8g5  f5d4  a7d4 
 5-     49    6   141692   g8g4  f1e1  g4g5 
 5&    149   11   265770   e6f5  e4f5  g8g3  f2g3  d4d3  e2f2  a7f2  f1f2 
                           c8c2  f2c2  d3c2  f5c2  d8b6  g1f1  b6e3  d1d5 
 5.    149   15   356544   e6f5  e4f5  g8g3  f2g3  d4d3  e2f2  a7f2  f1f2 
                           c8c2  f2c2  d3c2  f5c2  d8b6  g1f1  b6e3  d1d5 
 6&    150   21   510973   e6f5  e4f5  g8g3  f2g3  d4d3  e2f2  a7f2  f1f2 
                           c8c2  f2c2 
 6.    150   33   760747   e6f5  e4f5  g8g3  f2g3  d4d3  e2f2  a7f2  f1f2 
                           c8c2  f2c2 
 7.    150   60  1398467   e6f5  e4f5  g8g3  f2g3  d4d3  e2f2  a7f2  f1f2 
                           c8c2  f2c2 
1. ... e6f5
Gen 5528171 Node 1398467 Tree 860 Eval 62870 Rate 23307 EC 37502/24315 RS hi 0 lo 2 
Hin/Hout/Tcol/Coll/Fin/Fout = 42479/11016/0/0/0/0
My move is: e6f5

Your move is? 
==========================================================================

Take a look at depth=4.  First the search fail-high with a score of 228
points.  The value of beta at this point is 223 (not shown, but I did
print it out during some other test runs).  What happens next is that a
lower bound value of beta+1=224 is stored in the hash table.  The search
is now restarted with a window of (-9999, 9999).  

However, at ply=1, the hashtable is probed and the lowerbound value of
224 is found to which alpha is set.  Now a search of the position
suddenly yields a score of 220, which is less than 224.  So an
alpha-beta inconsistency has occurred at ply=1 and the final move picked
for depth=4 is Rg5 which is absolutely wrong.

The simplest way of solving this problem is not to probe the table at
ply==1 (which is what is done with pl68).  I tried this and here is what
I obtained.


White 1:00  Black 1:00

 - - R Q - K R -
 B P - - - P - -
 P - - P B - - -
 - - - - - n - P
 - p - P b - - p
 p - - - - - p -
 - - p - q p - -
 - - - r - r k -


Your move is? 
Your move is? 
Your move is? 
Move# 1    Target= 6000    Clock: 6000
 2.   -142    0     7620   d6d5  e4d3  e6f5  d3f5 
 3&   -142    0     7897   d6d5  e4d3  e6f5 
 3+    -31    0    11104   e6f5  c2c4  c8c4  e4f5  d4d3  d1d3  c4h4  e2h5 
 3&    133    1    18072   e6f5  e4f5  g8g3  f2g3  d4d3  e2f2  a7f2  f1f2 
                           d3c2  f5c2 
 3.    133    1    30066   e6f5  e4f5  g8g3  f2g3  d4d3  e2f2  a7f2  f1f2 
                           d3c2  f5c2 
 4+    228    2    38929   e6f5  e2c4 
 4&    224    6   105875   e6f5  e4f5  g8g7 
 4.    224    6   115636   e6f5  e4f5  g8g7 
 5&    125   10   183693   e6f5  e4f5  g8g3  f2g3  d4d3  e2f2  a7f2  f1f2 
                           c8c2  f2c2  d3c2  f5c2 
 5.    125   14   257775   e6f5  e4f5  g8g3  f2g3  d4d3  e2f2  a7f2  f1f2 
                           c8c2  f2c2  d3c2  f5c2 
 6&    126   22   384699   e6f5  e4f5  g8g3  f2g3  d4d3  e2f2  a7f2  f1f2 
                           c8c2  f2c2 
 6.    126   36   603067   e6f5  e4f5  g8g3  f2g3  d4d3  e2f2  a7f2  f1f2 
                           c8c2  f2c2 
 7.    126   60  1014040   e6f5  e4f5  g8g3  f2g3  d4d3  e2f2  a7f2  f1f2 
                           c8c2  f2c2 
1. ... e6f5
Gen 4349666 Node 1014040 Tree 928 Eval 50819 Rate 16900 EC 30723/19063 RS hi 0 lo 2 
Hin/Hout/Tcol/Coll/Fin/Fout = 35324/9925/0/0/0/0
My move is: e6f5

Your move is? 

==========================================================================

In this case, the search is able to respect the (-9999, 9999) window
at ply=1 and no alpha-beta inconsistencies at ply=1 occurs.

Kong Sian