2 Repeated trials and simple random walks

2.3 The reflection principle

We now look at a way of calculating the probability of ruin at any time t. Remember that if T denotes the time until ruin starting from X0=1, then T is odd, and

P(T=2m+1)=N2⁢m+1pmqm+1,

where N2⁢m+1 is the number of paths that start at X0=1 and first have Xt=0 at t=2⁢m+1. We will look at how to calculate N2⁢m+1.

Lemma 2.3.1 (Counting Paths).

Let Nt⁢(a,b) be the number of paths from X0=a to Xt=b. Let n=(b-a+t)/2. Then

Nt⁢(a,b)=t!n!⁢(t-n)!.
Proof.

Since a+n-(t-n)=b, n is the number of upwards moves and t-n is the number of downward moves needed for a path from X0=a to Xt=b. We can divide t into n upward moves and t-n downward moves and this is same as the number of possible combinations. ∎

Lemma 2.3.2 (The Reflection Principle).

Assume a>0 and b>0. Let Nt0⁢(a,b) be the number of paths from X0=a to Xt=b for which Xs=0 for some s=1,…,t-1. Then

Nt0⁢(a,b)=Nt⁢(-a,b).
Proof.

Consider a path from from X0=a to Xt=b for which Xs=0 for some s=1,…,t-1. Let s* be time when Xs is first 0. Then we can map this path onto a path from Y0=-a to Yt=b by reflecting the path between time 0 and Xs⁣*. Formally we define Ys=-Xs for s≤s* and Ys=Xs for s>s*.

For example consider a path with X0=1, X3=0 and X7=4:

Figure 2.10: Link, Caption: none provided

Now the key idea is that this mapping is a bijection between all paths from Y0=-a to Yt=b and all paths from X0=a to Xt=b that hit the value Xs=0 for some s=1,…,t-1. Hence the number of the two sorts of paths are identical. ∎

Theorem 2.3.3.

The number of paths that start at X0=1 and have ruin at T=2⁢m+1 is

N2⁢m+1=(2⁢m)!(m+1)!⁢m!.

Consider a picture of a path to ruin at T=7:

Figure 2.11: Link, Caption: none provided
Proof.

To prove this result, first notice that for such a path X2⁢m=1. So we wish to count the number of paths from X0=1 to X2⁢m=1 that do not have Xs=0 for any s=1,…,2⁢m-1. We count the number of such paths as:

‘the number of paths X0=1 to X2⁢m=1 ’ minus ‘the number of paths X0=1 to X2⁢m=1 which hit 0 ’

Thus

N2⁢m+1=N2⁢m⁢(1,1)-N2⁢m0⁢(1,1)

Using that N2⁢m0⁢(1,1)=N2⁢m⁢(-1,1) we get

N2⁢m+1=(2⁢m)!m!⁢m!-(2⁢m)!(m+1)!⁢(m-1)!=(2⁢m)!(m+1)!⁢m!⁢((m+1)-m)=(2⁢m)!(m+1)!⁢m!

∎