My theories were correct! The two reintegrating processes both produced significantly better results. This is a good step towards closing the gap between hand-coded and random rules.
Reintegration, Constant, Random rules, 25 episodes:
6473.467
10504.867
12725.2
13897.669
16034.133
17547.467
19964.398
20061.664
22945.738
26300.332
31034.93
37206.73
41162.93
42182.27
43964.4
45580.4
45398.2
47703.28
47414.53
48040.535
48624.734
50309.95
52900.47
55861.996
53042.676
Reintegration, Decaying, Random rules, 25 episodes:
7208.8677
10385.999
13772.134
15568.865
18112.664
19186.865
18503.064
20184.537
20790.4
19688.266
21682.734
21267.0
22745.8
22324.803
24573.2
25340.732
24706.87
26738.598
28712.8
30328.598
32650.533
34717.46
33499.406
36348.332
39545.996
Reintegration, Delayed, Random rules, 25 episodes:
7570.1323
9847.801
13380.066
15245.935
17163.867
18554.064
20945.402
25564.07
33929.2
38915.87
42504.664
43389.664
43021.46
44017.195
43591.465
45236.8
44742.33
45766.336
46340.805
46072.0
46521.33
47241.01
47006.336
47252.53
46688.2
I will have a graph comparing the results up shortly.
GGGRRRAAAHHH! Fuck Macs! Useless things. All I want to do is save a graph to file and upload it… Goddamn, they are garbage.