[From Bill Powers (941020.1935 MDT)]
RE: varint1.pas VI-VI model.
I have come across an initialization that seems to generate pretty good
matching between r1/(r1+r2) and b1/(b1+b2). It doesn't quite match the
"programmed ratio" but it's close. I don't pretend to understand why we
get matching here. Here are the guts of the "initialize" procedure with
the number I used in it.
for i := 1 to maxkey do
begin
enabled[i] := true;
reinfsize[i] := 2.0;
numreinf[i] := 0;
numpecks[i] := 0;
end;
k[1] := 100;
k[2] := 200;
decay1 := 1 - 1/50000.0;
o1max := 100.0;
g1 := 0.2;
r1 := 100.0;
Note that the probability of enabling a key is done only every 10th
iteration in varint1.pas, which makes the above two probabilities 0.01
and 0.02 per sec, with one iteration being 0.1 sec -- I think.
Bill