Transcript Poster(ppt)

Reversi
Meng Tran
Faculty Advisor: Dr. Barry Silverman
Abstract:


Corners






Brief History:

Approach:
Strategies:
The purpose of this project was to explore the
field of artificial intelligence using Reversi.
Reversi was invented around 1880 by Englishmen,
Lewis Waterman and John W. Mollett.
Mattel produced the game Reversi under the
name Othello.
Mattel was probably referencing the conflict in
Shakespeare's play.
Since 1978 there has been a World Othello
Championship.


Edges


Rules:








The game starts with 2 pieces of each color in the
center: Black on E4, D5 and White on E5, D4.
Each player takes alternate turns with Black going
first.
Each move must flank one of your opponent’s
color. There must be at least one of your
opponent’s chip between your current move and
one of your existing chip on the board.
The flanked chips changes to your color.
Each player must move if a valid move exists on
the board.
If a valid move does not exist then the player
must forfeit the turn.
The game ends when neither player can make a
valid move on the board either because there is
no more room or no valid moves exist.
The player with the most chips on the board wins.
New Game
Black moves E6 flanking
Whites E5, thus flipping
it to Black


Edges are the squares on the top and bottom rows
and leftmost and rightmost columns (marked with
an X).
Edges like corners gives the most potential in
flipping the most pieces across the board.
Edges are harder to flank as the only places that can
flank an edge is another edge.
Mobility





Corners are marked with an X and dangerous
squares are marked with an O.
Corners can not be flanked which can give you an
edge since it can flip pieces across the diagonal of
the board or to another corner.
Squares around the corner are dangerous as it
might give your opponent access to the corner.
One of the hardest strategies to pull off.
Limits your opponent’s moves by planning ahead.
Gains an extra move and also limits the squares
your opponent can play.
Extremely advantageous, but easier to pull off in the
later stages of the game.
End Game





At the end of the game, when 8 to 12 squares are
left anybody can perform a look-ahead to see which
squares are most advantageous.
Can predict where your opponent might move after
you.
Would be able to estimate if you can win with
certain moves.
Since there are only a few steps left, simulating
each move would not be too difficult.
Less time simulating what would happen.


1
49
5
17
18
6
50
2
51
57
41
33
34
42
58
52
7
43
13
25
26
14
44
8
19
35
27
28
36
20
21
37
29
30
38
22
9
45
15
31
32
16
46
10
53
59
47
39
40
48
60
54
3
55
11
23
20
12
56
4
Smart AI:





The Smart AI option plays using the End Game
strategy.
It uses a 12 step look-ahead.
Simulates the game and returns a heuristic of 1 for
win, -1 for loses, and 0 for draws.
Each node sums up the heuristic of all its children.
The move with the highest heuristic value is the
best possible move with the most chances of
ending in a win.
Conclusions:




Senior Project Poster Day 2007, CIS Dept. University of
Pennsylvania
Simple strategy
Instead of randomly choosing a square for the
default computer opponent, I used a priority matrix
developed by Peter Frey.
The top priority is 1 and lowest is 60.
This extensive depth-first search can only be done
near the end of the game as it takes a lot of time
and space.
Smart AI does not take into account Frey’s weights,
thus giving the opponent the upper hand when it
comes to corners.
End Game is not perfect unless it takes into
account the position’s weight.
Future research should go into the beginning of the
game where the search branches are extremely
long.