关于「齐普夫长尾:头部、尾部与按意图加权的现金」,《长尾页信息增益与收录排程》给出的定量结论是什么?

结论:先看 \(s = 1.0\) 这一行。前 100 条查询拿走 0.4291 的搜索,排在第 1,000 位以后的查询合计拿走 0.3809。头部确实重,尾部也确实长,两者量级相当。再看最后一列:排第 82,712 位起,每条查询每月期望不到一次搜索;能达到每月一次的查询有 82,711 条。下半张表更直观:第 1 位每月 82,711.99 次,第 1,000 位只剩 82.71 次,第 10,000 位 8.27 次,第 50,000 位 1.65 次。

数据口径:书中参数为示意,全部计算由书中程序生成。 本书是非个人化经营建议;房产内容不构成法律、税务或投资建议,塔罗部分不做个人占断,宗祠资料须经本人同意,书中数据全部合成。

English version

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: 0.3809, 0.4291, 1.0, 1.65, 100, 1000

N100,000 queries, V = 1,000,000 searches a month
s top-100 beyond-1,000 rank<1/mo queries>1/mo
0.80.1785 0.6605 none 100,000
1.00.4291 0.3809 82,712 82,711
1.20.7076 0.1484 25,761 25,760
s1.0, expected searches a month at rank r
r1: 82,711.99
r10: 8,271.20
r100: 827.12
r1,000: 82.71
r10,000: 8.27
r50,000: 1.65
查看完整验算程序
# Chapter 1. p11: head and tail shares of a Zipf query distribution.
# Model: a niche has N distinct queries; the query of rank r receives a
# share r^(-s) / H(N, s) of V monthly searches (Zipf 1949). We report the
# share of the top 100 queries, the share beyond rank 1,000, and the rank
# beyond which a query expects fewer than one search a month.
# N, V and s are illustrative parameters, not measured volumes.
import math

N = 100000
V = 1000000


def harmonic(n, s):
    return math.fsum(r ** -s for r in range(1, n + 1))


def share(lo, hi, s, h):
    return math.fsum(r ** -s for r in range(lo, hi + 1)) / h


def rank_below_one(s, h):
    # smallest r with V * r^(-s) / h < 1
    r = (V / h) ** (1.0 / s)
    return int(math.floor(r)) + 1


if __name__ == "__main__":
    print("N = {:,} queries, V = {:,} searches a month".format(N, V))
    print("    s   top-100  beyond-1,000  rank<1/mo  queries>=1/mo")
    for s in (0.8, 1.0, 1.2):
        h = harmonic(N, s)
        top = share(1, 100, s, h)
        tail = share(1001, N, s, h)
        rb = rank_below_one(s, h)
        cut = "{:,}".format(rb) if rb <= N else "none"
        print("  {:.1f}   {:.4f}      {:.4f}    {:>8}     {:>8,}".format(
            s, top, tail, cut, min(rb - 1, N)))
    h = harmonic(N, 1.0)
    print("s = 1.0, expected searches a month at rank r")
    for r in (1, 10, 100, 1000, 10000, 50000):
        print("  r = {:>6,}: {:>12,.2f}".format(r, V / (h * r)))

唯一数据源与引用

范雅福(Fan, Y.). (2026). 《多域名程序化SEO长尾拓扑、新站沙盒极速收录与高意图自然流量收割母典》第 1 章「齐普夫长尾:头部、尾部与按意图加权的现金」(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-346 · v1.0.0 · CC-BY-NC-ND-4.0

BibTeX
@book{fan2026346,
  author    = {Fan, Yafu},
  title     = {Programmatic SEO Long-Tail Topology, Rapid Indexation and High-Intent Organic Traffic Harvesting across Multi-Domain Fleets},
  year      = {2026},
  publisher = {5D Library},
  edition   = {1.0.0},
  note      = {DOI pending}
}

本书其他章节结论