Here are some notes and summary of a few corrections for a3: (All errors are updated in the specification)

1)Make sure to submit Board.java as well. (Originally this was missing from the list of files to submit at the end)

2)The constants NUMROWS and NUMCOLS should be made public so that other classes have access to them. 
Normally, it is considered bad style to make variables public, and it is better to use getters and setters
to set them. However, with a final, it is not quite as bad. (Also practically to work around this would require
adding a new method getRows() and getCols() so this is easier)

3)The projection equation was originally wrong. The right equation is 

u dot v       
-------    v
|v| |v|


This has been updated.

Also note that there is no test case for projection. You should write one yourself.

4)The normalize() method should do nothing if the vector magnitude is 0. (i.e. everything should stay the same) . 
The projectOnto() method should just return the vector v in this case. (Or any vector that is 0,0,0)

Since this method does NOT create a new vector and returns void, you will have to use it a bit differently:

Vector v1 = new Vector(1,1,1);
Vector v2 = new Vector(2,3,5);
Vector v3 = v1.Add(v2) ;  //returns a new vector so OK
 System.out.println( v1.Add(v3).calculateMagnitude()); //v1.Add(v3) returns a new vector. 
      //Since it returns a new vector, you can call calculateMagnitude() on this new vector
      //since calculateMagnitude returns a double, you can print it

System.out.println(v3.normalize());
    //ERROR: v3.normalize() returns void. You can't print void.

The difference is that the normalize() method changes the original vector. It does not create a new one.
The other methods create a new vector.

5)The method should be called calculateMagnitude() (Be consistant with the test case file)

6)the static method GenerateEmptyBoard is allowed to use the literals 6 and 8 to create an empty board. 
The reason that it is okay here but not in other cases is this method is designed specifically for a connect four 
game. Perhaps a better name would be GenerateEmptyConnectFourBoard  . All other methods are meant to work
for any other kind of board size. (In fact extending it to something like checkers or even chess would not
be too difficult)

7)If your AI player is too slow, you may modify the line at the top of ComputerPlayer.java which say MAX_PLY_DEPTH=6 to be a smaller number, like 2.
You should change this back before submitting, but we won't be looking at the value of that variable anyway.

8)Your heuristic should return -900 in the case of a fork (a previous version of the slides said -990. If you do this, the test case will fail)

9)A couple of the heuristic tests use ``unrealistic" connect 4 boards. That is, they have positions that are not legal. Your heuristic should work
regardless of whether the board is a legal position:

e.g.
_ _ _ _ _ _ _ _
_ _ _ _ _ _ _ _
_ _ _ _ _ _ _ O
_ _ _ _ _ _ _ _
_ _ _ _ _ _ _ _
X X X _ _ _ _ _

is not technically a reachable position as the piece in the right most column is floating. In addition X has had too many
turns. However, your heuristic should still count this as 1 group of 3 connected.

10)Clarification on the getWinner() method.

Your method should detect any time there are 4 *consecutive* in one row, column or diagonal since that is what it means to win in connect four (hence the name :)  )