Strategy¶

Random¶

First randomly selects a checker from all checkers of the player, then randomly selects a legal move for that checker.

\(score = f_{sn} + f_{dp} + f_{ks} + f_p\)
Solider Number. \(f_{sn} = c_{w} * N_{w\_solider} + c_{b} * N_{b\_solider}\) , \(c_w = 160/8, c_b = -160/16\). For black, reverse the sign.
Direct Path (to escape). For white,

\[\begin{split} f_{dp} = \begin{cases} c_{escape}, & n\_path = 1 \\ \frac{BIG\_NUM}{2}, & n\_path \geq 1 \end{cases}, {c}_{escape} = 20 \end{split}\]

For black,

\[\begin{split} f_{dp} = \begin{cases} \frac{BIG\_NUM}{4}, & n\_path = 1 \\ \frac{BIG\_NUM}{2}, & n\_path \geq 1 \end{cases} \end{split}\]
King Safety \(f_{ks}\). For white, the coefficient is negative, for black, positive.
- If the king is in the castle, for each adjacent black soldier, -20.
- If the king is adjacent to the castle, for each adjacent black soldier or camp, -30.
- If the king is not adjacent to the castle, for each adjacent black soldier or camp, -50.
- If the king is in these four areas, the FLAT_AREA, the king safety score adds depends on the number of black soldiers in the same area with table below:
```
    area_safety_table = {
        0: 50.0,
        1: 30.0,
        2: 0.0,
        3: -50.0,
        4: -100.0
    }
```
Position Score \(f_p\): precomputed positional score table for both roles.
- \(f_p = \sum_{i,j=0}^{i,j=9}{weight_{ij} * on_{ij}}\)

Early Termination
- makes Alpha-Beta pruning much more efficient, check captures first and eval a BIG_NUMBER to Quick fall
- For later win solution, eval a number less than BIG_NUMBER but still big, like BIG_NUMBER/2
Multiprocess Searching
- send tasks to 4 process to use 4 cores of CPU. (can be the number of cores of the machine)
- handle timeout easily. But one problem is that sub-tree state is lost if break from timeout with recursion search. To handle timeout more efficiently could implement search functions with loop.
- default depth is 3.