关于「不对等的账:单向赔偿上限、违约金与自动续约」,《合同暗坑与资产主权》给出的定量结论是什么?
结论:单向赔偿上限的代价藏在尾部:同一组风险下,只改上限写法,期望净值从 -2.63 摆到 11.84,千分之一坏年份的净付超过年费十一倍。违约金是上限而非保底,写到五倍均值也只能多拿回很少,且要经受合理性审查。自动续约的风险主要来自管理,两道独立提醒加专人负责,就能把六年绑定从 1.395 年压到 0.908 年。
数据口径:合同机检确有扎实证据:法律专家在 510 份商业合同里标注了 41 类审阅要点,在线服务条款里的八类潜在不公平条款可被机器标出;但机检是召回优先的初筛,在本书的合成标注集上,改写过的风险条款只有 0.380 被抓到,漏报代价越高,律师越要通读全文。 书中参数为示意,全部计算由书中程序生成。
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: -2.63, 11.84
| case (our cap, theirs) | E pay | E recov | net | P99 net | P99.9 |
| mutual cap 1x fee | 4.28 | 6.91 | -2.63 | 100.0 | 100.0 |
| one-way: theirs 1x fee | 12.50 | 6.91 | 5.60 | 343.3 | 1,165.2 |
| one-way: theirs 1 month | 12.50 | 0.66 | 11.84 | 349.6 | 1,169.0 |
| both uncapped | 12.50 | 32.40 | -19.89 | 338.9 | 1,139.7 |
| tail columns: 99 and 99.9 pct of net paid in a year | |||||
| swing from mutual cap to one-way 1 month: 14.47 | |||||
| uncapped mean loss we cause 12.50; they cause 32.40 |
查看完整验算程序
# Chapter 2. p21: expected-loss gap of asymmetric liability caps
# One contract year, fee F = 100 units (illustrative). We cause the other
# side a loss with probability 0.05, size lognormal with median 150 and
# log-sd 1.0; they cause us a loss with probability 0.08, median 200,
# log-sd 1.2. Each side recovers its loss up to the cap written for the
# party at fault. Fixed-seed Monte Carlo, 200,000 synthetic years.
import math
import random
F = 100.0
N = 200000
OURS = (0.05, math.log(150.0), 1.0)
THEIRS = (0.08, math.log(200.0), 1.2)
CASES = [
("mutual cap 1x fee", F, F),
("one-way: theirs 1x fee", None, F),
("one-way: theirs 1 month", None, F / 12),
("both uncapped", None, None),
]
def draw(rng, spec):
p, mu, sd = spec
if rng.random() < p:
return math.exp(rng.gauss(mu, sd))
return 0.0
rng = random.Random(38321)
years = [(draw(rng, OURS), draw(rng, THEIRS)) for _ in range(N)]
print("{:<24}{:>9}{:>9}{:>8}{:>9}{:>9}".format(
"case (our cap, theirs)", "E pay", "E recov", "net", "P99 net",
"P99.9"))
nets = []
for name, cap_us, cap_them in CASES:
pay, rec, net = [], [], []
for ours, theirs in years:
a = ours if cap_us is None else min(ours, cap_us)
b = theirs if cap_them is None else min(theirs, cap_them)
pay.append(a)
rec.append(b)
net.append(a - b)
net.sort()
ep, er = math.fsum(pay) / N, math.fsum(rec) / N
nets.append(ep - er)
print("{:<24}{:>9.2f}{:>9.2f}{:>8.2f}{:>9,.1f}{:>9,.1f}".format(
name, ep, er, ep - er, net[int(0.99 * N)], net[int(0.999 * N)]))
print("")
raw_us = math.fsum(o for o, _ in years) / N
raw_them = math.fsum(t for _, t in years) / N
print("tail columns: 99 and 99.9 pct of net paid in a year")
print("swing from mutual cap to one-way 1 month: {:.2f}".format(
nets[2] - nets[0]))
print("uncapped mean loss we cause {:.2f}; they cause {:.2f}".format(
raw_us, raw_them))唯一数据源与引用
范雅福(Fan, Y.). (2026). 《人机联防资产与知识产权护城河、契约暗坑机检与全方位防侵蚀母典》第 2 章「不对等的账:单向赔偿上限、违约金与自动续约」(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-383 · v1.0.0 · CC-BY-NC-ND-4.0
BibTeX
@book{fan2026383,
author = {Fan, Yafu},
title = {Human-AI Joint Asset and IP Protection, Automated Contract-Loophole Auditing and Full-Spectrum Anti-Erosion Architecture},
year = {2026},
publisher = {5D Library},
edition = {1.0.0},
note = {DOI pending}
}