floydf@attctc.Dallas.TX.US (Floyd Ferguson) (11/23/89)
(I tried posting this from work, but it did not seem to get out, so
if this is a repeat, please tolerate)
The problem with the s.a vote, and all votes, is that
1. they are not votes, but polls
which means they are stochastic processes, rather than legal
2. knowing the day to day vote tallies provides a partisan an inordinate
advantage, namely, the precise amount needed to win, and this information
becomes available _very_early_ in the process.
ARGUMENTATION:
Consider the pattern on incoming votes. I suspect these will be poisson
distributed, with a mean sometime between 5 and 10 days from the time the
call for vote is posted. (Someone who has run a vote with big turn-out
could check, maybe comp.mac.whatever, or comp.dsp, or mail me the dates
of arriving votes and I'll test the distribution.)
That means that somewhere arond 70% of the votes will have arrived before
the voting is half over! And if Richard (or Bob Webber, or whoever) knows
this, sufficient time remains to collect legitimate votes to "win" the
contest. The opposite side, not knowing this, sits in silence waiting for
time to be called, and find out how bad they lost. (But lose they will! No
vote run by a partisan ever need fail!)
CONCLUSIONS:
Make two minor changes in the current guidelines:
-- provide a mechanism for impartial vote counting. I think ncoast
has volunteered. During the discussion period, allow either of
the sides to request independent counting.
-- post the dates with the names in the final vote count.
Failing to do this will saddle any new voting scheme with today's problem:
i.e.: intrinsic knowledge regarding the polling provides inordinate advantage
if limited to only one side, which no amount of rules or complexity in voting
mechanism can overcome.
And implementing these two little changes may make major changes in the
guidelines superflous.
________________________________________
And until then, remember, WHOEVER RUNS THE VOTE ***W*I*N*S***
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
attctc!floydfwesommer@athena.mit.edu (Bill Sommerfeld) (11/26/89)
In article <10317@attctc.Dallas.TX.US> floydf@attctc.Dallas.TX.US (Floyd Ferguson) writes:
Consider the pattern on incoming votes. I suspect these will be poisson
distributed, with a mean sometime between 5 and 10 days from the time the
call for vote is posted.
Here's some Real Data:
I ran the rec.org.sca vote; it passed, roughly ~160-~20.
The machine which had the collection boxes, bloom-beacon.mit.edu, is
on the Internet, is well connected, and is in the UUCP maps.
The call for votes was posted on a Friday; there was a large initial
surge of 27 votes the first day, followed by another surge of 31 votes
on the following monday; the vote then tailed off to a couple a day
until I posted the first ack/second call for votes about two weeks
later; this appears to have caused another "hump" of ~20 votes spread
out over four days, *none of which affected the outcome of the vote*
The vote went over the 100-vote margin after only seven days; had the
CFV appeared on a monday, it might have occurred faster.
In the following table, only votes which had a Date: line parsable by
the C News "getdate" program were included, thus the counts will be a
little under the real vote total. The votes were counted based on the
"postmark", not the arrival time. "day" is the day of the vote; day
1=september 29. no votes arrived after the 28th day of the vote.
day tot. yes yes tot. no no
1 27 27 3 3
2 44 17 5 2
3 57 13 7 2
4 88 31 9 2
5 100 12 10 1
6 108 8 11 1
7 117 9 13 2
8 119 2 14 1
9 119 0 15 1
10 120 1 15 0
11 124 4 16 1
12 125 1 16 0
13 127 2 16 0
14 127 0 16 0
15 128 1 17 1
16 133 5 17 0
17 138 5 18 1
18 148 10 19 1
19 150 2 19 0
20 151 1 19 0
21 151 0 20 1
22 153 2 20 0
23 157 4 20 0
24 158 1 22 2
25 159 1 22 0
26 160 1 22 0
27 160 0 22 0
28 161 1 22 0
--
Henry Spencer is so much of a | Bill Sommerfeld at MIT/Project Athena
minimalist that I often forget | sommerfeld@mit.edu
he's there - anonymous |