Network Effect

Porting scope (source: /opt/algorithms/functional/network_effect.py, 55 lines, sha256 bc0ecfa3…)
· Aligned: the computation at L30-42 (int(N), the α/β branching, V, Metcalfe, the integer division for max_conn, density, distance to the critical point, critical progress) and the 8 returned fields at L44-53; the default parameters are copied from L24-28 (α=1.5, β=2.3, c=0.001, n_critical=1000, avg=100.0); the rounding digits are copied too (V 4 digits / Metcalfe 2 / density 6 / progress 1), using pyRound() to match Python's round-half-to-even.
· Units: the source labels no units at all. V = c×N^α is a value index (about 0.001×1000^2.3 ≈ 7943 at N=1000), not money; avg_value_per_user only enters the Metcalfe formula, never V.
· Not aligned ①: for N ≤ 1 the source sets max_conn to 1 (the else 1 at L39), so at N=0 the density is connections÷1 — this page copies that as-is and does not correct it.
· Not aligned ②: for negative N, c×N^α comes out as a complex number in Python and round() raises a TypeError; this page rejects negative values outright and does not simulate that exception.
· Not aligned ③: once N(N−1)/2 exceeds 2^53 (N ≳ 9.49e7), JS double precision loses integer precision while Python uses arbitrary-precision integers; at that magnitude this page flags it in the result.
· Two same-origin versions with different formulas that this page does not adopt: promotion_spirit/spread_algorithms/algo_network_effect.py uses V = N×avg×(1+density), does not divide Metcalfe by 2, and adds 0.7/0.3 density bands; /opt/algorithms/network_effect_engine.py (549 lines, with the first stretch L30-58 visible) has three stages pre/critical/post, two coefficient tiers c1=0.001 and c2=0.0001, and density bands 0.01/0.1. Item 1.6 of the inventory points to the functional version, so this page aligns only with the functional version — the three versions produce different numbers, so do not mix them.
· Why this page is not linked to “Curve diagnosis / scale conversion”: what this page produces is a series of value indices (magnitude 1e0~1e4), whereas curve diagnosis classifies on a 0~10 score with a ±10 scale — linking straight across would classify every value as “peak”. The type looks similar but the conventions do not connect, so there is no link.
· The page's input layer (which the source does not have): leaving a field empty uses the source's default; entering a non-number raises an error.

[Evidence section · where every number can be checked]
(1) Source: the algorithm-port machine /opt/algorithms/functional/network_effect.py (55 lines, read-only): default parameters = lines 25-28 (α=1.5, β=2.3, c=0.001, n_critical=1000, avg=100.0); computation = lines 30-42; the 8 returned fields = lines 44-53.
(2) Three categories
· [Verifiable] int(N), the α/β branching (N < n_critical uses α), V = c×N^α, Metcalfe avg×N²÷2, max_conn = N(N−1)//2, density, distance to the critical point, critical progress and the rounding digits — all checkable line by line.
· [This page's display convention] Every row of the “scale comparison table” other than “← your input” is a page sample at another scale using the same formula, not source output (the source computes only the single N you entered).
· [No source found] Where the four defaults α=1.5, β=2.3, c=0.001, n_critical=1000 come from is neither commented nor sourced in the file header or the function body → no source found. Also: V carries no unit label at all, and it is not money.
(3) How to verify it yourself: enter N=1000, connections=999, avg_value_per_user=100 → since N ≥ n_critical it takes β: V = 0.001×1000^2.3 ≈ 7943.28; Metcalfe = 100×1000²÷2 = 50,000,000; density = 999 ÷ (1000×999÷2) = 0.002. The three numbers line up with the three formulas at lines 31-42.