1: \begin{abstract}

2:     The usual way to investigate the statistical properties of

3:     finitely generated subgroups of free groups, and of finite

4:     presentations of groups, is based on the so-called word-based

5:     distribution: subgroups are generated (finite presentations are

6:     determined) by randomly chosen $k$-tuples of reduced words, whose

7:     maximal length is allowed to tend to infinity.  In this paper we

8:     adopt a different, though equally natural point of view: we

9:     investigate the statistical properties of the same objects, but

10:     with respect to the so-called graph-based distribution, recently

11:     introduced by Bassino, Nicaud and Weil.  Here, subgroups (and

12:     finite presentations) are determined by randomly chosen Stallings

13:     graphs whose number of vertices tends to infinity.

14: 

15:     Our results show that these two distributions behave quite

16:     differently from each other, shedding a new light on which

17:     properties of finitely generated subgroups can be considered

18:     \textit{frequent} or \textit{rare}.  For example, we show that

19:     malnormal subgroups of a free group are negligible in the

20:     graph-based distribution, while they are exponentially generic in

21:     the word-based distribution.  Quite surprisingly, a random finite

22:     presentation generically presents the trivial group in this new

23:     distribution, while in the classical one it is known to

24:     generically present an infinite hyperbolic group.

25: \end{abstract}

26: