上一篇我们完成了整个程序的基础框架，那么在讲到真正的搜索算法前，我们先来看看五子棋如何评估当前局势，以及如何计算某个位置的价值。
一、五子棋
在五子棋中，包括成五，活三，活二等定势，下图为山东师范大学董红安在2005年的硕士毕业论文中使用的的评分表，可以供我们来参考。
但是对于四子棋来说，上述评分却并不适用，因为棋盘空间大小的原因，任何一个维度只有4子的空间，一旦没有落成，或是任意一个位置被对方下了，那么该位置将没有任何价值。
二、潜在可能性评估
我们以这张图来举例，当黑棋在(0,3,0)这个位置落子后，我们来分析之后的可能性。
首先在xy平面内，有如下三种方式取胜：
除xy平面内之外，我们还需要考虑立体斜着四子连成的情况：
最后还要考虑垂直四子连成的情况：
并且在计算价值时，我们要注意，在上述任何一种情况中，只要预计的路线上有对方棋子出现，那么这条线（仅仅是单条线，不是整体）的评分将为0，因为他已经不能实现胜利。
三、展平化
为了更方便的计算，我们通过将chessBoard[x][y][z]符合规则的任意连续的四个子编入序列，并通过计数的方式实现计分。
我们通过一段代码来了解这个过程：
struct flatData{
int a;
int b;
int c;
int d;
};
typedef std::vector<flatData> PicesFlatDataList;
int isWin(int board[][][]);
int ChessBoard::isWin(int board[][][])
{
PicesFlatDataList flats;
for(int y = ;y < ;y++)
{
for(int z = ;z < ;z++)
{
flatData data;
data.a = board[][y][z];
data.b = board[][y][z];
data.c = board[][y][z];
data.d = board[][y][z];
flats.push_back(data);
}
}
for(int x = ;x < ;x++)
{
for(int z = ;z < ;z++)
{
flatData data;
data.a = board[x][][z];
data.b = board[x][][z];
data.c = board[x][][z];
data.d = board[x][][z];
flats.push_back(data);
}
}
for(int x = ;x < ;x++)
{
for(int y = ;y < ;y++)
{
flatData data;
data.a = board[x][y][];
data.b = board[x][y][];
data.c = board[x][y][];
data.d = board[x][y][];
flats.push_back(data);
}
}
for(int y = ;y < ;y++)
{
flatData data;
data.a = board[][y][];
data.b = board[][y][];
data.c = board[][y][];
data.d = board[][y][];
flats.push_back(data);
}
for(int x = ;x < ;x++)
{
flatData data;
data.a = board[x][][];
data.b = board[x][][];
data.c = board[x][][];
data.d = board[x][][];
flats.push_back(data);
}
for(int y = ;y < ;y++)
{
flatData data;
data.a = board[][y][];
data.b = board[][y][];
data.c = board[][y][];
data.d = board[][y][];
flats.push_back(data);
}
for(int x = ;x < ;x++)
{
flatData data;
data.a = board[x][][];
data.b = board[x][][];
data.c = board[x][][];
data.d = board[x][][];
flats.push_back(data);
}
for(int z = ;z < ;z++)
{
flatData data;
data.a = board[][][z];
data.b = board[][][z];
data.c = board[][][z];
data.d = board[][][z];
flats.push_back(data);
}
for(int z = ;z < ;z++)
{
flatData data;
data.a = board[][][z];
data.b = board[][][z];
data.c = board[][][z];
data.d = board[][][z];
flats.push_back(data);
}
flatData data;
data.a = board[][][];
data.b = board[][][];
data.c = board[][][];
data.d = board[][][];
flats.push_back(data);
data.a = board[][][];
data.b = board[][][];
data.c = board[][][];
data.d = board[][][];
flats.push_back(data);
data.a = board[][][];
data.b = board[][][];
data.c = board[][][];
data.d = board[][][];
flats.push_back(data);
data.a = board[][][];
data.b = board[][][];
data.c = board[][][];
data.d = board[][][];
flats.push_back(data);
}
不难看出逻辑简单粗暴，直接遍历整个棋盘，并且将所有横竖斜的可能性加入FlatData的abcd四个位置中，再把该条加入到整个list中，为后续其他功能提供数据。
四、局势评分及输赢判断
在上一步的基础上，我们要做的是根据每组FlatData（展平后的数据格式）来给出我们评估的分数。
int whiteNum = ,blackNum = ;
for(auto iter = flats.begin();iter != flats.end();iter++)
{
whiteNum = ;
blackNum = ;
if(iter->a == chessPicesStatus::black)
blackNum++;
else if(iter->a == chessPicesStatus::white)
whiteNum++;
if(iter->b == chessPicesStatus::black)
blackNum++;
else if(iter->b == chessPicesStatus::white)
whiteNum++;
if(iter->c == chessPicesStatus::black)
blackNum++;
else if(iter->c == chessPicesStatus::white)
whiteNum++;
if(iter->d == chessPicesStatus::black)
blackNum++;
else if(iter->d == chessPicesStatus::white)
whiteNum++;
if(whiteNum == )
return chessPicesStatus::white;
if(blackNum == )
return chessPicesStatus::black;
}
return chessPicesStatus::empty;
以上为判断输赢的代码，可以看出就是在上一步的基础上添加了abcd四个位置白棋黑棋的统计，并且判断是否已经获胜。对于获取当前局面分评估的逻辑，其实只是在最后一步统计的时候更加细分一些：
int ChessBoard::getFlatPicesValue(PicesFlatDataList flats, chessPicesStatus status)
{
int value = ,whiteNum = ,blackNum = ; for(auto iter = flats.begin();iter != flats.end();iter++)
{
whiteNum = ;
blackNum = ; if(iter->a == chessPicesStatus::black)
blackNum++;
else if(iter->a == chessPicesStatus::white)
whiteNum++; if(iter->b == chessPicesStatus::black)
blackNum++;
else if(iter->b == chessPicesStatus::white)
whiteNum++; if(iter->c == chessPicesStatus::black)
blackNum++;
else if(iter->c == chessPicesStatus::white)
whiteNum++; if(iter->d == chessPicesStatus::black)
blackNum++;
else if(iter->d == chessPicesStatus::white)
whiteNum++; if(status == chessPicesStatus::white)
{
//Calculating White Picess if(blackNum != )
continue; if(whiteNum == )
value += ;
else if(whiteNum == )
value += ;
else if(whiteNum == )
value += ;
else if(whiteNum == )
value += ;
else if(whiteNum == )
{
value += ;
} }
else
{
//Calculating Black Picess if(whiteNum != )
continue; if(blackNum == )
value += ;
else if(blackNum == )
value += ;
else if(blackNum == )
value += ;
else if(blackNum == )
value += ;
else if(blackNum == )
{
value += ;
} } //cout<<iter->a<<" "<<iter->b<<" "<<iter->c<<" "<<iter->d<<" "<<value<<endl;
}
return value;
}
可以看出我们对于任何一个FlatData，如果该条数据有对方棋子存在，价值即为零，进入下一条。价值从本方棋子数目为零到四分别为1，5，100，5000，10000。
五、单个位置价值评分
为了实现后续的启发式搜索算法，我们需要计算出每个可下位置的得分，单个位置价值评分大体逻辑与上面两步基本一样，是先获取当前棋盘可落子的位置，接着遍历与之相连的各种可能性。但由于立体四子棋支持斜着四子连成，所以需要额外注意，此时展开需要分为SimpleFlat与Non-SimpleFlat。具体区别在于：
红色位置需要计算斜着连成的各种情况，而黑色区域的棋子不需要。所以在遍历时如果判断需要Non-SimpleFlat，则需要把更多的（斜着的若干种可能）添加到list中进行计算。
bool ifSimpleFlat(PicesPos pos)
{
if(pos.x == && pos.y == ||
pos.x == && pos.y == ||
pos.x == && pos.y == ||
pos.x == && pos.y == ||
pos.x == && pos.y == ||
pos.x == && pos.y == ||
pos.x == && pos.y == ||
pos.x == && pos.y == )
return true;
else
return false;
} int ChessBoard::getPosValue(int board[][][],PicesPos *pos, chessPicesStatus side)
{
PicesFlatDataList flatList; // cout<<"x:"<<pos->x<<" y:"<<pos->y<<" z:"<<pos->z<<endl;
if(ifSimpleFlat(*pos))
{
/*
[0] -
[1] |
[2] /-
[3] -\
[4] /_
[5] _\
*/
flatData temp0;
temp0.a = board[][pos->y][pos->z];
temp0.b = board[][pos->y][pos->z];
temp0.c = board[][pos->y][pos->z];
temp0.d = board[][pos->y][pos->z];
flatList.push_back(temp0); flatData temp1;
temp1.a = board[pos->x][][pos->z];
temp1.b = board[pos->x][][pos->z];
temp1.c = board[pos->x][][pos->z];
temp1.d = board[pos->x][][pos->z];
flatList.push_back(temp1); flatData temp2;
temp2.a = board[][pos->y][];
temp2.b = board[][pos->y][];
temp2.c = board[][pos->y][];
temp2.d = board[][pos->y][];
flatList.push_back(temp2); flatData temp3;
temp3.a = board[][pos->y][];
temp3.b = board[][pos->y][];
temp3.c = board[][pos->y][];
temp3.d = board[][pos->y][];
flatList.push_back(temp3); flatData temp4;
temp4.a = board[pos->x][][];
temp4.b = board[pos->x][][];
temp4.c = board[pos->x][][];
temp4.d = board[pos->x][][];
flatList.push_back(temp4); flatData temp5;
temp5.a = board[pos->x][][];
temp5.b = board[pos->x][][];
temp5.c = board[pos->x][][];
temp5.d = board[pos->x][][];
flatList.push_back(temp5); flatData temp6;
temp6.a = board[pos->x][pos->y][];
temp6.b = board[pos->x][pos->y][];
temp6.c = board[pos->x][pos->y][];
temp6.d = board[pos->x][pos->y][];
flatList.push_back(temp6);
}
else
{
/*
[0] -
[1] |
[2] /-
[3] -\
[4] /_
[5] _\
[6] -|/
[7] -|\
*/
flatData temp0;
temp0.a = board[][pos->y][pos->z];
temp0.b = board[][pos->y][pos->z];
temp0.c = board[][pos->y][pos->z];
temp0.d = board[][pos->y][pos->z];
flatList.push_back(temp0); flatData temp1;
temp1.a = board[pos->x][][pos->z];
temp1.b = board[pos->x][][pos->z];
temp1.c = board[pos->x][][pos->z];
temp1.d = board[pos->x][][pos->z];
flatList.push_back(temp1); flatData temp2;
temp2.a = board[][pos->y][];
temp2.b = board[][pos->y][];
temp2.c = board[][pos->y][];
temp2.d = board[][pos->y][];
flatList.push_back(temp2); flatData temp3;
temp3.a = board[][pos->y][];
temp3.b = board[][pos->y][];
temp3.c = board[][pos->y][];
temp3.d = board[][pos->y][];
flatList.push_back(temp3); flatData temp4;
temp4.a = board[pos->x][][];
temp4.b = board[pos->x][][];
temp4.c = board[pos->x][][];
temp4.d = board[pos->x][][];
flatList.push_back(temp4); flatData temp5;
temp5.a = board[pos->x][][];
temp5.b = board[pos->x][][];
temp5.c = board[pos->x][][];
temp5.d = board[pos->x][][];
flatList.push_back(temp5); flatData temp6;
temp6.a = board[pos->x][pos->y][];
temp6.b = board[pos->x][pos->y][];
temp6.c = board[pos->x][pos->y][];
temp6.d = board[pos->x][pos->y][];
flatList.push_back(temp6); if(pos->x == && pos->y == ||
pos->x == && pos->y == ||
pos->x == && pos->y == ||
pos->x == && pos->y == )
{
flatData temp7;
temp7.a = board[][][];
temp7.b = board[][][];
temp7.c = board[][][];
temp7.d = board[][][];
flatList.push_back(temp7); flatData temp8;
temp8.a = board[][][];
temp8.b = board[][][];
temp8.c = board[][][];
temp8.d = board[][][];
flatList.push_back(temp8);
}
else
{
flatData temp7;
temp7.a = board[][][];
temp7.b = board[][][];
temp7.c = board[][][];
temp7.d = board[][][];
flatList.push_back(temp7); flatData temp8;
temp8.a = board[][][];
temp8.b = board[][][];
temp8.c = board[][][];
temp8.d = board[][][];
flatList.push_back(temp8);
} } int val = getFlatPicesValue(flatList,side); return val; }
至此，我们已经完成了棋盘的评估函数，下一章我们将讨论极值搜索算法，也就是我们真正的博弈树函数。