关于「虚拟价值与迈尔森保留价:买家越少,底价越值钱」,《二价暗标与最优保留价》给出的定量结论是什么?
结论:第二章引入虚拟价值与最优保留价:均匀分布下保留价的相对增益从两位买家的 0.2500 降到五位的 0.0078,但在指数长尾下增益从一位买家的 5,666 元先升到两位的 6,843 元,这是对立项蓝图的修正;多一位真实买家在卖方自留价值为零时全部胜过最优保留价,在自留价值较高且有长尾时反转。
数据口径:书中的估值分布、保有价值、买家人数与价格均为示意;全部计算由书中程序生成;本书是非个人化经营建议,也不是法律意见,正式拍卖以上架前人工复核的当地法规为准;紧迫感只能来自真实库存与真实截止,不虚构竞价者,不托价。 第七章考辨《宋史》的买扑与苏辙「立定中价,不许添刬」之议(示意模型中按自身估计出价的违约概率达 0.9691)以及希罗多德记载的巴比伦婚配拍卖(收入是嫁资的 3.373 倍)。
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: 0.0078, 0.2500, 5666, 6843
| A. uniform [0,1], s | 0, r* = 1/2 (exact) |
| n no reserve with r* gain gain / no-reserve | |
| 1 | 0 1/4 1/4 inf |
| 2 | 1/3 5/12 1/12 0.2500 |
| 3 | 1/2 17/32 1/32 0.0625 |
| 4 | 3/5 49/80 1/80 0.0208 |
| 5 | 2/3 43/64 1/192 0.0078 |
| 6 | 5/7 321/448 1/448 0.0031 |
| 7 | 3/4 769/1024 1/1024 0.0013 |
| 8 | 7/9 1793/2304 1/2304 0.0006 |
| B. seller value s | 100,000 yuan; payoff with r = s vs r = r* |
| n law r = s r = r* gain gain / base | |
| 1 uniform | 100000 111250 11250 0.1125 |
| 2 uniform | 111250 119688 8438 0.0758 |
| 3 uniform | 121094 126104 5010 0.0414 |
| 5 uniform | 133398 134923 1524 0.0114 |
| 8 uniform | 142223 142481 258 0.0018 |
| 1 power | 100000 118519 18519 0.1852 |
| 2 power | 123484 129778 6293 0.0510 |
| 3 power | 134903 136914 2011 0.0149 |
| 5 power | 144647 144888 242 0.0017 |
| 8 power | 150275 150288 13 0.0001 |
| 1 exp | 100000 105666 5666 0.0567 |
| 2 exp | 103954 110797 6843 0.0658 |
| 3 exp | 109155 115461 6306 0.0578 |
| 5 exp | 119433 123608 4175 0.0350 |
| 8 exp | 131639 133527 1888 0.0143 |
核心公式
\[\varphi(v) = v - \frac{1 - F(v)}{f(v)}.\]
查看完整验算程序
# Chapter 2. p22: the fewer the buyers, the more the reserve is worth.
# Part A, exact: values U[0, 1], seller value 0, reserve r = 1/2.
# no reserve: R0(n) = (n-1)/(n+1)
# with r: R(n, r) = R0(n) + r^n - 2n r^(n+1)/(n+1)
# so the gain is 2^-n / (n+1), in exact fractions.
# Part B, numeric: bag values on [80,000, 160,000]-type laws of p21,
# seller value s = 100,000. Expected seller payoff (price if sold, s if
# kept) = s F(r)^n + n * int_r^top phi(v) F(v)^(n-1) f(v) dv, by
# composite Simpson with math.fsum. Baseline: reserve = s ("never below
# what it is worth to me"); policy: reserve = r* from p21.
import math
from fractions import Fraction as Fr
print("A. uniform [0,1], s = 0, r* = 1/2 (exact)")
print("n no reserve with r* gain gain / no-reserve")
for n in range(1, 9):
r0 = Fr(n - 1, n + 1)
g = Fr(1, 2 ** n * (n + 1))
rel = "inf" if r0 == 0 else "%.4f" % float(g / r0)
print("%d %-10s %-8s %-8s %s" % (n, r0, r0 + g, g, rel))
L, W, THETA, S = 80000.0, 80000.0, 30000.0, 100000.0
LAWS = {
"uniform": (lambda v: (v - L) / W, lambda v: 1.0 / W, L + W, 130000.0),
"power": (lambda v: ((v - L) / W) ** 2,
lambda v: 2.0 * (v - L) / W / W, L + W, 400000.0 / 3.0),
"exp": (lambda v: 1.0 - math.exp(-(v - L) / THETA),
lambda v: math.exp(-(v - L) / THETA) / THETA,
L + 20 * THETA, 130000.0)}
def payoff(law, n, r, s=S, k=2000):
F, f, top, _ = LAWS[law]
h = (top - r) / k
terms = []
for i in range(k + 1):
v = r + i * h
fv = f(v)
phi = v - (1.0 - F(v)) / fv if fv > 0 else 0.0
w = 1 if i in (0, k) else (4 if i % 2 else 2)
terms.append(w * phi * F(v) ** (n - 1) * fv)
return s * F(r) ** n + n * h / 3.0 * math.fsum(terms)
print("B. seller value s = 100,000 yuan; payoff with r = s vs r = r*")
print("n law r = s r = r* gain gain / base")
for law in ("uniform", "power", "exp"):
rstar = LAWS[law][3]
for n in (1, 2, 3, 5, 8):
base = payoff(law, n, S)
best = payoff(law, n, rstar)
print("%d %-7s %7.0f %7.0f %5.0f %.4f"
% (n, law, base, best, best - base, (best - base) / base))唯一数据源与引用
范雅福(Fan, Y.). (2026). 《维克里二价暗标、迈尔森最优保留价与孤品奢包限量席位激励相容拍卖母典》第 2 章「虚拟价值与迈尔森保留价:买家越少,底价越值钱」(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-237 · v1.0.0 · CC-BY-NC-ND-4.0
BibTeX
@book{fan2026237,
author = {Fan, Yafu},
title = {Vickrey Second-Price Sealed Bids and Myerson Optimal Reserve Prices: Incentive-Compatible Auctions for One-of-a-Kind Luxury Pieces and Limited Seats},
year = {2026},
publisher = {5D Library},
edition = {1.0.0},
note = {DOI pending}
}