DOES IID PRODUCE A VISIBLE CLUSTER? — NOI QUICK DRAW ANALYSIS
Speed3
Frame 1/60
STRUCTURED DRAW
Cluster-biased generation: Moore neighbors 2.5× more likely.
Absent numbers with deficit get elevated probability.
Mimics the confirmed balancing mechanism (B4, Z=104).
Core persistence
—
VS
CLUSTER PERSISTENCE
PURE IID DRAW
Genuine IID: 20 numbers chosen uniformly at random from 80.
No memory. No spatial preference. No balancing.
Same classification machinery applied identically.
Core persistence
—
What to observe:Structured: The cyan/blue anchor core
(A/CO) persists across frames — the same region of the
grid glows draw after draw. The cluster traverses the grid slowly,
expanding through diagonal corners around UN blockades.
|
IID: The "cluster" dissolves and reforms randomly
each draw. No region persists. The classification scheme finds
A/CO/GN numbers everywhere — but they appear in different
positions every draw. There is no traversal. There is no core.
|
Key metric: Core persistence
(fraction of current draw that was also in previous draw).
Structured: ~—
IID: ~—
— a — difference.