UNFOLLOW freaking me out.

[lkettle]

I've summarized 50 tests done from same INFILE varying only with regard to language proportions:

• stage1_unfollow: true (unfollow for relative chattiness)
• unfollow_tweet_rate: 10 (unfollow for absolute chattiness, being 10 tweets per simulated minute)

It appear from the data:

• 0 unfollows: 90% of the time
• 1000 unfollows: 10% chance

Another of the 1K unfollows happened in a test case of one Region.

4/5 of the 1000 unfollows occur when the two language groups are approximately 50-50 in 2 different regions.

So is this an issue? Why the all-or-nothing effect?

[lkettle]

These INFILES are referenced above:
B_50_10_40 0 Unfollows
G_45_10_45 1039 Unfollows
B_40_10_50 0 Unfollows

INFILE_B_50_10_40.zip
INFILE_G_45_10_45.zip
INFILE_B_40_10_50.zip

B_50_10_40 means:
Region 0: Eng 50 F+E 10 Fr 40
Region 1: Eng 40 F+E 10 Fr 50
equal numbers of English speakers and French speakers in world.
equal numbers of each Red & Blue ideology in world
equal numbers of Region 0 and Region 1 citizens

[lkettle]

No bilingualism:
L_50_0_50 both regions => 1047 Unfollows
R2_100_0_0 & 0_0_100 => 0 Unfollows
INFILE_L_50_0_50.zip
INFILE_R2_100_0_0.zip