How many dimensions is it safe to get drunk in? In Michael Lugo's blog post Variations on the drunken-bird theorem, and real-world sightings he wonders (without coming to a conclusion) what the maximum 'safe' number of dimensions to get drunk in might be. 

So where's the line between "always finds his way home" and "gets lost forever"? It seems to me that this should be at 2+ε dimensions, since that's what we need to make the sum convergent, but I don't know if anyone's studied random walks on fractal spaces.

In the comments Matt Noonan recalls an unverified claim that the line is at 2.5 dimensions. 
Can anybody shed any further light on these musings?
 A: Indeed, the answer is 2+ϵ in most senses. For example, consider a wedge in Z3 which is all the points (x,y,z) such that |z|<f(x) for some specific increasing function f:N->N .
These wedges are transient already for pretty slowly growing functions. If you take f(x)=xϵ which correspond, in the sense of volume growth, to 2+ϵ dimensions, then it is transient for any ϵ>0.
One can get even more refined results by taking f(x)=log(x)a, in which case the wedge is transient if and only if a>1.
You can find more information in the book of Lyons with Peres.
A: In that post, my intuition was simply that the probability of a random walk in d dimensions being at the origin at time t scales like t-d/2, and ∑t > 0 t-d/2 converges if and only if d > 2.  Obviously this can only be turned into a proof if d is an integer.
A: There are some "fractal" meanings to a question like this.
See the delightful little Carus Monograph Random Walks and Electric Networks by Doyle and Snell.  At the end of Chapter 6 they discuss a certain tree that has dimension log 6/log 2 = 2.5849... It is used to show that the 3-dimensional lattice is transient.
A: Have a look at
http://mathworld.wolfram.com/PolyasRandomWalkConstants.html
The answer turns out that at as soon as dim > 2 the probability of recurrence is lower than 100%.
