Description:
- A gambler to play potentially infite number of games, each wins 1 unit with probability p and lose 1 unit otherwise.
- Starting with i units, what is the probability that the gambler’s fortune will reach N before reaching 0
- Given that P00∞=PNN=1
- meaing he stops playing when he lose all or reaching N
- and Pi,i+1=p=1−q=1−Pi+1,i for i=1,2,...,N−1
Solve with Markov:
- We have 3 classes, {0} and {N} are transient while {1,..,N−1} is recurrent
- Define Pi as probability that the fortune will eventually reach N, Pi=p.Pi+1+q.Pi−1 for i=1,2,...,N−1
- Next:
- P0=0
- Pi+1−Pi=pq(Pi−Pi−1) for i=1,2,...,N−1
- P2−P1=pq(P1−P0)=pqP1
- P3−P2=pq(P2−P1)=(pq)2P1
- …
- Pi−Pi−1=pq(Pi−1−Pi−2)=(pq)i−1P1
- Sum from P2 to i−i terms →Pi−P1=P1[(qp)+(qp)2+...+(qp)i−1]
- Pi=1−(q/p)1−(q/p)iP1
- As PN=1=1−(q/p)1−(q/p)NP1→P1=1−(q/p)N1−(q/p)
- →Pi=1−(q/p)N1−(q/p)i
- Pi={1−(q/p)N1−(q/p)iNiif p=21if p=21
- For n→∞, Pi={1−(pq)i0if p>21if p≤21