关于「一鱼五吃:改写公式、时间预算下的分配与对未知权重的稳健性」,《五大平台算法与一鱼五吃》给出的定量结论是什么?

结论:第六章给出一鱼五吃的改写公式:核心不变、表层按平台重写;一周 10 个小时的最优分配价值 71.83,是原样转贴的 3.32 倍,只比平均分配高 1.051 倍;平台参数大幅估错时,分散改写平均损失 0.126,主平台精做加转贴损失 0.352。

数据口径:书中把每条平台说法标为三类:带日期的官方披露、同行评审研究、本书的示意模型;书中的权重、概率与漏斗参数均为示意,不是任何平台的真实设定,全部计算由书中程序生成。 本书是一般性经营参考,不构成法律、税务或投资建议;不涉及假账号、机器人、互赞群、买互动、抓取个人资料或群发私讯;合作内容一律标明;达人酬劳只有单层直付,不涉及任何多层分润。

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: 1.051, 10, 3.32, 71.83

copy-paste to all five2.5 h value 21.61
equal 2 h each10.0 h value 68.37
all 10 h on RedNote10.0 h value 38.57
RedNote 8 h + 4 copies10.0 h value 53.33
optimal10.0 h value 71.83
optimal hours: RedNote 3.0, Threads 1.5, X 1.0, Instagram 3.0, Facebook 1.5
optimal / copy-paste3.32; / home + copies: 1.35; / equal split: 1.051
budget5.0 h: value 43.83; RedNote 2.0, Threads 0.5, X 0.5, Instagram 1.5, Facebook 0.5
platforms left as copy-paste: Facebook
budget 15.0 h: value90.35; RedNote 4.5, Threads 2.0, X 2.0, Instagram 4.0, Facebook 2.5
platforms left as copy-paste: none
查看完整验算程序
# Chapter 6. p61: one source, five adaptations under a 10-hour weekly budget.
# For platform j, h hours give A_j * (1 - exp(-h / B_j)) value units when h >= c_j,
# the hours a real rewrite needs; with less than c_j, only a 0.5-hour copy-paste is
# possible, worth phi_j times the adapted value at h = c_j. Value units stand for
# qualified actions (saves, replies, inbound DMs, group comments). A, B, c and phi are
# illustrative. Exact search: dynamic programming on a 0.5-hour grid.
import math

P = {"RedNote": (40.0, 3.0, 1.5, 0.35), "Threads": (18.0, 2.0, 0.5, 0.60),
     "X": (15.0, 2.0, 0.5, 0.60), "Instagram": (35.0, 3.0, 1.5, 0.40),
     "Facebook": (20.0, 2.5, 1.0, 0.50)}
BUDGET, STEP, COPY = 10.0, 0.5, 0.5


def val(j, h):
    a, b, c, phi = P[j]
    if h >= c:
        return a * (1 - math.exp(-h / b))
    if h >= COPY:
        return phi * a * (1 - math.exp(-c / b))
    return 0.0


def best_plan(budget):
    units = int(budget / STEP)
    table = {0: (0.0, {})}
    for j in P:
        new = {}
        for used, (v, plan) in table.items():
            for k in range(units - used + 1):
                h = k * STEP
                key = used + k
                cand = v + val(j, h)
                if key not in new or cand > new[key][0]:
                    new[key] = (cand, dict(plan, **{j: h}))
        table = new
    return max(table.values(), key=lambda x: x[0])


def total(plan):
    return math.fsum(val(j, h) for j, h in plan.items())


opt_v, opt = best_plan(BUDGET)
plans = {"copy-paste to all five": {j: COPY for j in P},
         "equal 2 h each": {j: 2.0 for j in P},
         "all 10 h on RedNote": {"RedNote": 10.0},
         "RedNote 8 h + 4 copies": dict({j: COPY for j in P}, RedNote=8.0),
         "optimal": opt}
for name, plan in plans.items():
    hours = math.fsum(plan.values())
    print("%-23s %4.1f h  value %6.2f" % (name, hours, total(plan)))
print("optimal hours: " + ", ".join("%s %.1f" % (j, opt[j]) for j in P))
print("optimal / copy-paste: %.2f; / home + copies: %.2f; / equal split: %.3f" % (
    opt_v / total(plans["copy-paste to all five"]), opt_v / total(plans["RedNote 8 h + 4 copies"]),
    opt_v / total(plans["equal 2 h each"])))
for b in (5.0, 15.0):
    v, plan = best_plan(b)
    copies = [j for j in P if 0 < plan[j] < P[j][2]]
    print("budget %4.1f h: value %6.2f; " % (b, v) + ", ".join("%s %.1f" % (j, plan[j]) for j in P))
    print("               platforms left as copy-paste: %s" % (", ".join(copies) or "none"))

唯一数据源与引用

范雅福(Fan, Y.). (2026). 《五大社交图谱算法异质性、跨平台同源异构分发与达人聚拢母典》第 6 章「一鱼五吃:改写公式、时间预算下的分配与对未知权重的稳健性」(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-359 · v1.0.0 · CC-BY-NC-ND-4.0

BibTeX
@book{fan2026359,
  author    = {Fan, Yafu},
  title     = {Algorithmic Heterogeneity across RedNote, Threads, X, Instagram and Facebook: Cross-Platform Adaptation and Creator Aggregation},
  year      = {2026},
  publisher = {5D Library},
  edition   = {1.0.0},
  note      = {DOI pending}
}

本书其他章节结论