1: \begin{abstract}

2: We describe a novel algorithm for random sampling of freely reduced words equal 

3: to the identity in a finitely presented group. The algorithm is 

4: based on Metropolis Monte Carlo sampling. The algorithm samples from a 

5: stretched Boltzmann distribution

6: \begin{align*}

7:   \pi(w) &= (|w|+1)^{\alpha} \beta^{|w|} \cdot Z^{-1}

8: \end{align*}

9: where $|w|$ is the length of a word $w$, $\alpha$ and $\beta$ are parameters 

10: of the algorithm, and $Z$ is a normalising constant. It follows that words of 

11: the same length are sampled with the same probability. 

12: The distribution can be expressed in terms of the cogrowth series of the 

13: group, which then allows us to  relate statistical properties of words sampled by 

14: the algorithm to the cogrowth of the group, and hence its amenability.

15: 

16: We have implemented the algorithm and applied it to several  group 

17: presentations including the Baumslag-Solitar groups, some free 

18: products studied by Kouksov, a finitely presented  amenable  

19: group that is not subexponentially amenable (based on the 

20: {\em basilica group}), and  Richard Thompson's group~$F$.

21: \end{abstract}

22: