Scribd
Upload
15 views2 pages

Handout Gametheory

This document discusses finite two-player games with perfect information, defining key concepts such as winning positions and strategies. It presents a theorem stating that in any such game, one player has a winning strategy from any given position, along with a sketch of a recursive algorithm to determine the winner. The document includes specific examples, particularly focusing on the game Bridg-It, and proves that the starting player has a winning strategy in this game.

Uploaded by

yasserelmoussaed
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
15 views2 pages

Handout Gametheory

This document discusses finite two-player games with perfect information, defining key concepts such as winning positions and strategies. It presents a theorem stating that in any such game, one player has a winning strategy from any given position, along with a sketch of a recursive algorithm to determine the winner. The document includes specific examples, particularly focusing on the game Bridg-It, and proves that the starting player has a winning strategy in this game.

Uploaded by

yasserelmoussaed
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 2

Some facts on finite two-player games with

perfect information
Definition. For the purpose of this handout, a game will always be a game
between two players of the following form. The game starts in some starting
position. The player alternate in making their move, thus bringing the game in
a new position. It is part of the position which player is to move next, so the
position “empty board and Player 1 moves next” is different from “empty board
and Player 2 moves next”. Both players have at every time full information
about the rules of the game and of the current position. (This is called perfect
information. Many board games are perfect information games, while most card
games are not.)
Some positions are endings. Each ending implies either Player 1 or Player 2
as the winner of the game (so we do not allow draws). We call the game finite
if there is no infinite sequence of moves without reaching an ending.
Examples for such games are Bridg-It, Go, and Nim games (e.g.: from a pile of
100 coins, the players may take in their turn 1, 2, or 3 coins. The player who
takes the last coin wins).
Theorem. For every finite two-player information game as defined above, and
every position P , exactly one of the two players (call him W) has a winning
strategy meaning that for all strategies of the other player (call him L) there
exists a strategy of W that will lead to an ending where W wins. In this case,
we call P a winning position for W.

Proof (sketch). Let P be a position of the game. Assume that player 1 moves
next. Then it is easily determined whether P is a winning position for Player 1
by the following recursive algorithm.
DETERMINE_WINNER(P )
• For each possible move M of Player 1, let P 0 (M ) be the resulting position.
For each M , compute DETERMINE_WINNER(P 0 (M )).
• If DETERMINE_WINNNER(P 0 (M )) == Player 2 for all moves M then
return “Player 2”.
# No matter what Player 1 does, Player 2 can win in all cases.
• If there is a move M such that DETERMINE_WINNNER(P 0 (M )) ==
Player 1, then return “Player 1”.
# There is a winning move for Player 1.
Note that this algorithm is terribly inefficient, but for finite games it will termi-
nate and return the winner. It is also called “Minimax Algorithm”. To see that
this algorithm is well-defined, that is, it will eventually terminate, it is enough
to observe that we cannot reach position P from P 0 (M ), as otherwise there
would exist an infinite sequence of moves without reaching an end position.
Bridg-It
Recall that the board of Bridge-It consists of 25 “crossings” which can be either
claimed by the blue or the red player.
So we can describe a position by a triple (B, R, b), where B and R are two
disjoint subsets of [5]x[5], encoding the crossings that have been claimed so far
by the blue and the red player, respectively; and b is a bit that encodes which
player is the next to move. Note that in this way we even encode more positions
than we can reach from the standard starting position (B = R = ∅).
A move of the blue player (the red player) consists of adding one free crossing
(= a crossing neither contained in B nor R) to B (to R), and flipping the bit b.
An end position is reached when all 25 crossings have been claimed. Then blue
is a winner if and only if B contains a path from left to right.
Lemma. The game is monotone in the following sense. If (B, R, b) is a winning
position for blue, and B 0 ⊇ B is a set of crossings disjoint to R, then (B 0 , R, b)
is also a winning position for blue.

Proof. We use an inductive argument on the number of moves yet to play. In


an end position, the statement is trivially true.
If (B, R, b) is not an end position and it is red’s turn, then we need to show
that all moves of red lead to a winning position for blue. Since (B, R, b) is a
winning position for blue, for all free crossings c the position (B, R ∪ {c}, b̄) is
also a winning position for blue. Hence, by induction hypothesis, (B 0 , R ∪ {c}, b̄)
is also a winning position whenever B 0 and R ∪ {c} are disjoint. Thus (B 0 , R, b)
is a winning position for blue.
If (B, R, b) is not an end position and it is blue’s turn, then there exists a free
crossing c such that (B ∪ {c}, R, b̄) is a winning position for blue. Now we need
to find a winning move for position (B 0 , R, b). If c is not contained in B 0 , blue
takes c as his next move, otherwise he takes an arbitrary free crossing. (If no
free crossing exists then (B 0 , R, b) is already an end position won by blue.) In
both cases, blue can achieve a position (B 00 , R, b̄), where B 00 ⊇ B ∪ {c}. By
induction hypothesis, he has thus reached a winning position.
Theorem. The blue player (= the starting player) has a winning strategy for
Bridge-It.

Proof. Assume otherwise, i.e., assume the position P0 := (∅, ∅, blue) is a win-
ning position for red. Then by symmetry, (∅, ∅, red) must be a winning position
for blue. Let c be an arbitrary crossing. Then by the lemma, P1 := ({c}, ∅, red)
is also a winning position for blue. But blue can reach P1 from P0 by claiming
the crossing c. Thus P0 is also a winning position for blue (he has a winning
move), in contradiction to the assumption.

You might also like