@Mich
I guess one small change that caused huge change in evals.
was using beta cutoff as
alpha>=beta
You make your debug hard to read because your positions go from 0-8 rather than the expected 1-9.
Well now you can see it from 1-9.
Worst of all though you don't appear to be even looking at the debug output and attempting to understand what it means ie in the first example it played initially in the corner and these were the score returned for a 6 ply search for the 2nd move by the 2nd player.
# Player2
# Eval= -1 played at 1
# Eval= -1 played at 2
# Eval= -1 played at 3
# Eval= 0 played at 4
# Eval= 0 played at 5
# Eval= 0 played at 6
# Eval= 0 played at 7
# Eval= 0 played at 8
However all moves other than in the centre result in a loss within 6 ply ie 0,7,4,8,6,3,2 or perhaps 0,6,8,4,2,1,5 etc etc. Each of these wins is forced yet your program returns a draw
This was due to the way I implemented beta cutoff . I referred the code at
www.hamedahmadi.com
I put there alpha>=beta(On site, it said that it will speed up the search a bit, so I thought the evals of -1 changed to 0 was part of the algorithm) Now I changed it to alpha>beta and its giving proper evaluations.
For your test case.
Here is the complete game set played at 6plys.
Board format:
123
456
789