These statistics are from the starting position up to ply 12.
My engine has the following move ordering implementation:
1) hashmove
2) winning captures (SEE-score > -50)
3) killer-moves
4) quiet-moves, using the history heuristic
5) losing captures (SEE-score < -50)
The evaluation function does not recognize hanging pieces. These will be found the next ply so this is bad for move ordering.
SEE does not recognize illegal moves (pinned-pieces), again bad for move ordering.
Image:
https://drive.google.com/file/d/0BxPfDW ... 9UWVk/view
The following can be seen in the data:
- the chart of the EXACT moves is not 100% accurate because of the 0 values
- total number of moves LOWER > UPPER > EXACT
- move ordering for LOWER moves > EXACT > UPPER
- the rise of UPPER-moves between move 20 and 45 are caused by the sorting of loosing captures. If I disable this, this peak is not seen, but the total number of nodes is higher (needs tuning?)
- smetimes the bestmove is movenumber 50!
What do you think of these numbers? I am still not 100% familiar with CUT-nodes, ALL-nodes, fail-high fail-low, etc...
Data:
Code: Select all
EXACT UPPER LOWER
ply # % # % # %
1 182 57,24 83321 43,81 573940 94,7
2 57 17,93 31173 16,39 21999 3,63
3 20 6,29 21524 11,32 4744 0,79
4 11 3,46 13349 7,02 1779 0,3
5 13 4,09 8137 4,28 1026 0,17
6 2 0,63 3590 1,89 667 0,12
7 5 1,58 2139 1,13 383 0,07
8 4 1,26 1380 0,73 242 0,04
9 1 0,32 908 0,48 172 0,03
10 2 0,63 761 0,41 152 0,03
11 0 0 647 0,35 84 0,02
12 4 1,26 599 0,32 90 0,02
13 0 0 482 0,26 84 0,02
14 1 0,32 483 0,26 81 0,02
15 0 0 406 0,22 50 0,01
16 0 0 405 0,22 47 0,01
17 1 0,32 375 0,2 53 0,01
18 0 0 350 0,19 55 0,01
19 1 0,32 331 0,18 42 0,01
20 1 0,32 322 0,17 52 0,01
21 1 0,32 418 0,22 34 0,01
22 0 0 454 0,24 28 0,01
23 0 0 533 0,29 20 0,01
24 2 0,63 625 0,33 26 0,01
25 3 0,95 684 0,36 31 0,01
26 1 0,32 784 0,42 16 0,01
27 0 0 865 0,46 13 0,01
28 2 0,63 1064 0,56 13 0,01
29 2 0,63 1171 0,62 16 0,01
30 1 0,32 998 0,53 18 0,01
31 0 0 1167 0,62 15 0,01
32 0 0 1153 0,61 15 0,01
33 0 0 1013 0,54 18 0,01
34 0 0 1064 0,56 13 0,01
35 1 0,32 970 0,51 9 0,01
36 0 0 1010 0,54 7 0,01
37 0 0 987 0,52 7 0,01
38 0 0 881 0,47 7 0,01
39 0 0 805 0,43 7 0,01
40 0 0 713 0,38 8 0,01
41 0 0 619 0,33 12 0,01
42 0 0 482 0,26 4 0
43 0 0 319 0,17 3 0
44 0 0 288 0,16 3 0
45 0 0 185 0,1 1 0
46 0 0 112 0,06 1 0
47 0 0 62 0,04 1 0
48 0 0 44 0,03 0 0
49 0 0 31 0,02 0 0
50 0 0 18 0,01 0 0
51 0 0 2 0,01 0 0
52 0 0 1 0 0 0
53 0 0 3 0,01 0 0