关于「三十个站点构成一条马尔可夫链」,《诊断收费站、悬念缺口与解锁定价》给出的定量结论是什么?
结论:第四章把三十个站点解成一条马尔可夫链:同主题导流使购买提高到 1.4087 倍,每页四条链接最好。
数据口径:书中的用户、比率与参数均为示意模型;全部计算由书中程序生成;本书不是法律、财务或消费者保护方面的建议。 书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的假说。
对照数据表
下表数字由本章 Python 程序原样打印,可重新运行复现。 程序只用 Python 标准库与结论吻合的打印值: 1.4087
| policy | buys per 1000 arrivals | pages per visit | vs none |
| none | 30.000 | 1.0000 | 1.0000 |
| uniform | 35.544 | 1.4286 | 1.1848 |
| 80 | 40.637 | 1.4286 | 1.3546 |
| 95 | 42.260 | 1.4286 | 1.4087 |
查看完整验算程序
# Chapter 4. p07: Thirty sites as one Markov chain. Thirty diagnostic
# sites form 6 themes of 5. A visitor arrives at a site and keeps the
# theme of that site as her interest. On each page she buys (and leaves)
# with chance b_j r, where b_j runs from 0.010 to 0.030 and r is 1.5 on a
# site of her theme and 0.5 elsewhere; she follows a cross-link with
# chance 0.30, else leaves. Policies route a click to another site: none,
# uniform, 80% or 95% within the theme. We solve for purchases per 1000
# arrivals and pages per visit by iterating the chain to convergence.
import math
N, T = 30, 6
B = [0.010 + 0.020 * ((7 * j) % N) / (N - 1) for j in range(N)]
def theme(j):
return j // 5
def route(policy, j):
"""Probabilities of the next site after a click on site j."""
p = [0.0] * N
if policy == "none":
return p
same = [k for k in range(N) if theme(k) == theme(j) and k != j]
other = [k for k in range(N) if theme(k) != theme(j)]
w_same = {"uniform": len(same) / (N - 1), "80": 0.80, "95": 0.95}[policy]
for k in same:
p[k] = w_same / len(same)
for k in other:
p[k] = (1 - w_same) / len(other)
return p
def solve(policy, g):
buy = [B[j] * (1.5 if theme(j) == g else 0.5) for j in range(N)]
click = 0.30 if policy != "none" else 0.0
P = [route(policy, j) for j in range(N)]
v, n = [0.0] * N, [0.0] * N
for _ in range(80):
v = [buy[j] + click * math.fsum(P[j][k] * v[k] for k in range(N))
for j in range(N)]
n = [1 + click * math.fsum(P[j][k] * n[k] for k in range(N))
for j in range(N)]
return v, n
if __name__ == "__main__":
print("policy buys per 1000 arrivals pages per visit vs none")
base = None
for pol in ("none", "uniform", "80", "95"):
bs, ps = [], []
for i in range(N):
v, n = solve(pol, theme(i))
bs.append(v[i])
ps.append(n[i])
b = 1000 * math.fsum(bs) / N
base = base or b
print("{:8s} {:22.3f} {:15.4f} {:7.4f}".format(
pol, b, math.fsum(ps) / N, b / base))唯一数据源与引用
范雅福(Fan, Y.). (2026). 《算法诊断收费站、悬念缺口校准与多阶解锁经济学》第 4 章「三十个站点构成一条马尔可夫链」(版本 1.0.0). 5D Library. DOI 待发布
DOI 待发布 · 专著页正式发布后上线(本站暂无专著页) · 5DL-B-114 · v1.0.0 · CC-BY-NC-ND-4.0
BibTeX
@book{fan2026114,
author = {Fan, Yafu},
title = {Algorithmic Diagnostic Paywalls, Curiosity-Gap Calibration and Tiered Report Unlock Economics},
year = {2026},
publisher = {5D Library},
edition = {1.0.0},
note = {DOI pending}
}