关于「防跳单违约金,与检测功效」,《暗投授权与钩子归因》给出的定量结论是什么?

结论:第六章用二叉树为跳单期权定价,合理预估 20.82 已拦住示意模型中全部跳单。

数据口径:书中参数均为示意,全部计算由书中程序生成;本书是非个人化经营分析,不构成经营或法律建议。 第六章用二叉树为跳单期权定价,合理预估 20.82 已拦住示意模型中全部跳单。

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: 20.82

tree: u 1.1891, d 0.8410, risk-neutral q 0.4640
pre-estimate of agency loss per skip: 20.82
Loption valueskip chanceL / pre-estimate
0.003.2990.66920.00
3.001.7610.43530.14
6.000.8300.20020.29
9.000.3500.09250.43
12.000.1300.05630.58
15.000.0480.02150.72
20.820.0000.00001.00
service Sno-clause skippre-estimateL to stop all
180.669220.8220.65
121.000024.0023.01
61.000024.0025.51
01.000024.0028.27
查看完整验算程序
# Chapter 6. p61: the creator's option to skip the agency, on a tree.
# Model (illustrative): the creator's brand deals are worth D a year,
# D0 = 100 (thousand), volatility 0.6, risk-free rate 0.03. The agency
# keeps f = 0.20 but also brings deal flow and service worth S = 18
# a year to the creator. Skipping at month i nets (f D - S) times
# (12 - i) / 12 for the rest of the term and costs damages L.
# Cox, Ross and Rubinstein (1979) tree, 12 monthly steps, American
# exercise; probabilities are risk-neutral. The agency's loss if the
# creator skips at month i is the lost commission plus a re-matching
# cost of 4; its average over the skip nodes reached when L = 0 is
# the pre-estimate a liquidated-damages clause should track.
import math

D0, SIG, RF, F, STEPS, REMATCH = 100.0, 0.6, 0.03, 0.20, 12, 4.0
SERVICE = 18.0
DT = 1.0 / STEPS
U = math.exp(SIG * math.sqrt(DT))
DN = 1.0 / U
Q = (math.exp(RF * DT) - DN) / (U - DN)
DISC = math.exp(-RF * DT)


def deal(i, j):
    return D0 * U ** j * DN ** (i - j)


def saving(i, j, serv=SERVICE):
    return (F * deal(i, j) - serv) * (STEPS - i) / STEPS


def lost(i, j):
    return F * deal(i, j) * (STEPS - i) / STEPS + REMATCH


def solve(ld, serv=SERVICE):
    val = [0.0] * (STEPS + 1)
    ex = {}
    for i in range(STEPS, -1, -1):
        new = []
        for j in range(i + 1):
            cont = 0.0 if i == STEPS else DISC * (
                Q * val[j + 1] + (1 - Q) * val[j])
            now = saving(i, j, serv) - ld
            ex[(i, j)] = now > cont and now > 0
            new.append(max(cont, now, 0.0))
        val = new
    mass = {(0, 0): 1.0}
    hit, loss = [], []
    for i in range(STEPS + 1):
        for j in range(i + 1):
            m = mass.get((i, j), 0.0)
            if m == 0.0:
                continue
            if ex[(i, j)]:
                hit.append(m)
                loss.append(m * lost(i, j))
            elif i < STEPS:
                mass[(i + 1, j + 1)] = mass.get((i + 1, j + 1), 0.0) \
                    + m * Q
                mass[(i + 1, j)] = mass.get((i + 1, j), 0.0) \
                    + m * (1 - Q)
    p = math.fsum(hit)
    return val[0], p, (math.fsum(loss) / p if p > 0 else 0.0)


if __name__ == "__main__":
    print("tree: u {:.4f}, d {:.4f}, risk-neutral q {:.4f}".format(
        U, DN, Q))
    v0, p0, pre = solve(0.0)
    print("pre-estimate of agency loss per skip: {:.2f}".format(pre))
    print("     L   option value   skip chance   L / pre-estimate")
    for ld in (0.0, 3.0, 6.0, 9.0, 12.0, 15.0, round(pre, 2)):
        v, p, _ = solve(ld)
        print("  {:>5.2f}   {:>12.3f}   {:>11.4f}   {:>16.2f}".format(
            ld, v, p, ld / pre))
    print("service S   no-clause skip   pre-estimate   L to stop all")
    for serv in (18.0, 12.0, 6.0, 0.0):
        v, p, est = solve(0.0, serv)
        top = max(saving(i, j, serv) for i in range(STEPS + 1)
                  for j in range(i + 1))
        print("  {:>8.0f}   {:>14.4f}   {:>12.2f}   {:>13.2f}".format(
            serv, p, est, top))

唯一数据源与引用

范雅福(Fan, Y.). (2026). 《暗投白名单授权经济学、三秒钩子生存分析与单层星探归因博弈母典》第 6 章「防跳单违约金,与检测功效」(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-333 · v1.0.0 · CC-BY-NC-ND-4.0

BibTeX
@book{fan2026333,
  author    = {Fan, Yafu},
  title     = {Dark-Ads Whitelist Licensing Economics, Three-Second Hook Survival Analysis and Single-Tier Scout Attribution Games},
  year      = {2026},
  publisher = {5D Library},
  edition   = {1.0.0},
  note      = {DOI pending}
}

本书其他章节结论