关于「账号质量由谁决定:同意率、负面反馈与等级马尔可夫链」,《私域速率整形》给出的定量结论是什么?

结论:第二章写出四级账号等级的马尔可夫链:同意率从 0.90 降到 0.80,平均每日上限从 99054 跌到 24692;只改发送时间,期望负面反馈的变化是 0.000000。

数据口径:示意计划每分钟平均 10.0 条,攒批猛发仍被拒收 420 条,接上桶深 20 的令牌桶后为零。 书中参数为示意,平台数字注明文档与查阅日期;全部计算由书中程序生成,程序不读写文件、不联网;本书是非个人化经营建议,不是法律或合规意见,任何线上发送节奏的改动须由主事者人工复核放行。

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: 0.80, 0.90, 24692, 99054

consentfeedbackP(high)P(low)P(top)expected 24h limit
1.000.00100.9950.0001.000100000
0.950.00200.9000.0001.00099979
0.900.00290.6380.0070.98999054
0.800.00480.1570.1720.21724692
0.700.00670.0200.5810.000324
0.600.00860.0020.8760.000254
查看完整验算程序
# Chapter 2. p21: consent share drives the tier chain, through feedback.
# Model (illustrative; the platform does not publish its formula). Tier
# limits are Meta's published steps of unique recipients per moving 24 h
# (250, 2,000, 10,000, 100,000; "unlimited" is folded into the top).
# Each week the platform sees n = 2000 delivered messages; k of them
# draw negative feedback (block or report), k ~ Binomial(n, f). Quality
# is high if k <= 6 (rate <= 0.003), low if k > 12 (rate > 0.006),
# medium otherwise. High moves one tier up, low one tier down, medium
# stays. Per-message feedback rate: 0.001 from a consented contact,
# 0.02 from a contact who never opted in; f = s 0.001 + (1 - s) 0.02.
import math

LIMITS = (250, 2000, 10000, 100000)
N, HI_K, LO_K = 2000, 6, 12
P_C, P_U = 0.001, 0.02


def binom_cdf(k, n, p):
    terms = [math.comb(n, j) * p ** j * (1 - p) ** (n - j)
             for j in range(k + 1)]
    return math.fsum(terms)


def quality_probs(f):
    ph = binom_cdf(HI_K, N, f)
    pl = 1 - binom_cdf(LO_K, N, f)
    return ph, 1 - ph - pl, pl


def stationary(f):
    ph, pm, pl = quality_probs(f)
    m = len(LIMITS)
    tr = [[0.0] * m for _ in range(m)]
    for i in range(m):
        for j, p in ((min(i + 1, m - 1), ph), (i, pm),
                     (max(i - 1, 0), pl)):
            tr[i][j] = math.fsum([tr[i][j], p])
    pi = [1.0 / m] * m
    for _ in range(5000):
        pi = [math.fsum(pi[i] * tr[i][j] for i in range(m))
              for j in range(m)]
    return pi


if __name__ == "__main__":
    print("consent  feedback  P(high)  P(low)  P(top)  expected 24h limit")
    for s in (1.0, 0.95, 0.9, 0.8, 0.7, 0.6):
        f = s * P_C + (1 - s) * P_U
        ph, pm, pl = quality_probs(f)
        pi = stationary(f)
        lim = math.fsum(w * x for w, x in zip(pi, LIMITS))
        print("%7.2f  %8.4f  %7.3f  %6.3f  %6.3f  %18.0f"
              % (s, f, ph, pl, pi[-1], lim))

唯一数据源与引用

范雅福(Fan, Y.). (2026). 《私域通道防封号速率控制论、令牌桶限流与平台公示上限内的同意制触达吞吐母典》第 2 章「账号质量由谁决定:同意率、负面反馈与等级马尔可夫链」(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-271 · v1.0.0 · CC-BY-NC-ND-4.0

BibTeX
@book{fan2026271,
  author    = {Fan, Yafu},
  title     = {Rate Control for Private Messaging Channels without Account Bans: Token-Bucket Shaping, Platform-Published Limits and Consent-Based Reach Throughput},
  year      = {2026},
  publisher = {5D Library},
  edition   = {1.0.0},
  note      = {DOI pending}
}

本书其他章节结论