Math 443/543 Graph Theory Notes 5: Digraphs, Tra C, and Tournaments
Math 443/543 Graph Theory Notes 5: Digraphs, Tra C, and Tournaments
1 Digraphs
Recall the de…nition of a digraph.
2 Tra¢ c problem
Suppose we have a small town for which we are planning a road system. We
have the following limitations:
1. We can usually have all of the roads bidirectional since there is not much
tra¢ c.
2. However, on Saturdays that the football team plays at home, there is a
great deal of tra¢ c and so we would like to make the roads one-way.
3. We sometimes have to do repairs on roads, so we would like to be able to
have a detour around any road segment so that drivers can still get from
any one place to any other place.
4. We would like to be able to get from any one place to any other place even
on Saturdays, when the roads are one-way.
We turn this problem into a graph in the following way. Let every road
intersection be a vertex, and connect the vertices if there is a road between
them. Limitation 1 says that usually we have a proper graph. Limitation 2
says that on Saturdays, we want to produce a digraph by assigning a direction
to each edge of the graph. Limitation 3 says that we do not want to have any
bridges in the graph. Limitation 4 says that the digraph is strongly connected,
as de…ned:
1
De…nition 2 A v0 vk -path in a digraph G is a list P = v0 ; e1 ; v1 ; : : : ; ek ; vk such
that vi 2 V (G) and ei = (vi 1 ; vi ) 2 E+ (G) is a directed edge and such that
the vertices and edges are distinct. A graph is strongly connected if for every
vertices u; v 2 V (G) there is a uv-path and a vu-path.
Proof. First we show that a connected, orientable graph has no bridges. Sup-
pose G is a connected graph with a bridge uv. Now suppose we have assigned di-
rections to the edges to produce a digraph G+ : Certainly G+ contains a uv-path
or a vu-path consisting of just the edge (u; v) or (v; u) (depending on our choice
of orientation of that edge). Without loss of generality, say (u; v) 2 E+ (G+ ) :
Then there is no vu-path in G+ ; since G uv is disconnected and so any path
in G between u and v must contain the edge uv; which is oriented the wrong
way. Thus G+ is not orientable.
Now we show that if G has no bridges and is connected, then it is orientable.
Since G has no bridges, every edge is contained in a cycle. The cycle can be
oriented in an obvious way. Now suppose that there is an edge not in the cycle
between two vertices in the cycle. We can assign those directions however we
wish, and if the cycle contains all of the vertices of G; it is clear that G would
be strongly connected. Now suppose that there is a vertex not in C: Then there
must be an edge vi w 2 E (G) such that w is not in C: This edge must lie on a
cycle C1 = w1 ; w2 ; w3 ; : : : ; wk ; where w1 = w and w2 = vi : We direct the edge
(w; vi ) = (w1 ; w2 ) and then direct the edges (wi ; wi+1 ) if they have not already
been directed. Finally, any edges between vertices in C1 and C but not in the
cycle, we can assign the direction arbitrarily. We claim that this is a strongly
connected subgraph. Given any vertex, there is a path to a point on the cycle
C and also there is a path from any point on the cycle C to the vertex. Now we
can continue with more cycles if all vertices are not contained in C and C1 :
Show picture like Fig 7.1.
3 Tournaments
De…nition 5 A tournament T is a directed graph such that for each pair of
vertices v; w 2 V (T ) ; either (v; w) or (w; v), not both, is in E+ (T ) : Thus
a tournament is a directed graph gotten by giving directions to the edges of a
complete graph.
2
tournament as v1 ; v2 ; : : : ; vp in such a way that (vi ; vi+1 ) 2 E+ (T ) ; where we
think of (v; w) 2 E+ (T ) means that v wins over w:
Proof. Suppose the vertices adjacent from v are fv1 ; : : : ; vn g and then the
vertices adjacent to v are fu1 ; : : : ; um g : We wish to show that for each vertex
ui ; there is a vertex vj adjacent to ui : If there were a vertex ui such that no
vertex vj is adjacent to ui for all j = 1; 2; : : : ; n; then o deg (ui ) n + 1; which
would mean that v is not maximal, a contradiction.
The interpretation is this. If there is a tournament where every team plays
every other tournament (a round-robin tournament), then the winner (the team
with the most wins; there may be several) has only lost to teams which have
lost to teams they have beaten.
Note that this means that we can rank teams in the tournament as suggested
above. Note that this may not be unique, and also that the winner might not
be what you expect!
Proof. We induct on the number of vertices p in the tournament. It is certainly
true for p = 1; and we can easily check it for p = 2; 3; 4: Now suppose it is true
for p n: Let T be a tournament with n vertices. We note that T v is still a
tournament (where T v removes the vertex v and all arcs to and from v). By the
inductive hypothesis, T v has a hamiltonian path P = v1 ; v2 ; : : : ; vn : If (v; v1 )
or (vn ; v) 2 E+ (T ) ; then T has a hamiltonian path. Otherwise, (v; v1 ) 2 E+ (T )
and there must be a minimal k n such that (vk ; v) 2 E+ (T ) : Then we have
that P 0 = v1 ; : : : ; vk 1 ; v; vk ; vk+1 ; : : : ; vn is a hamiltonian path.