in reply to monastery memory
in thread Troll Warning

Interesting node, Matt, particularly the notion of reporting on tension in threads. For kicks, I threw together a little code to see how well the binomial standard deviation would work as a measure of an individual post's tension. 
# Binomial standard deviation, from PerlMonks node [477227]
# Standard deviation = SQR{(N*p*(1-p)} 
# Represent upvotes as 1, downvotes as 0 (or vice-versa)
sub b_stddev {
  my $N = @_ or return 0;
  my $p = (grep $_, @_) / $N;
  sqrt($N * $p * (1-$p));
}

my @samples = (1, 3, 8, 20);

for (map {my $x=$_; map [$x, $_], grep $x >= $_, @samples} @samples) {
  printf "%2d %2d: %3.4f\n", @$_, b_stddev((1) x $_->[0], (0) x $_->[1
+]);
}

[download]
The results didn't appeal to me, because a lot of votes in one direction and just a few in the other can make for a comparatively high result. So using the stddev of reps in a thread as a measure of tension in the thread probably wouldn't appeal, either.

Possibly counter-intuitively, it is the nodes of low tension (and low rep) that I would like to have warnings for, while I think that threads of high tension are most likely to include trolls.

Caution: Contents may have been coded under pressure.

In Section Perl Monks Discussion