关于「公理表与回合制引擎:十九条结论、零智能代理与一台会破产的沙盘」,《五方敌对代理兵棋沙盘》给出的定量结论是什么?
结论:第一章建立公理表与回合制引擎,并复现零智能交易者:预算约束一条规则把效率从 0.8929 抬到 0.9730。
数据口径:本书用一台回合制沙盘证伪它:五类敌对代理,即价格敏感买家、套利投机者、合规审查官、模仿竞争者与流动性冲击,单独出手时最坏只让示意方案的破产概率升到 0.0010;五类同时出手,破产概率是 0.2165,九成五区间从 0.1990 到 0.2351。 书中参数为示意,单位是无币种的点;全部计算由书中程序生成;本书是非个人化经营分析,不构成投资、法律或财务建议,也不构成任何收益承诺;套利代理只用于找出并堵上自家规则漏洞。
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: 0.8929, 0.9730
| maximum surplus 560.0 points, 7 efficient trades | |||
| ZI-C (budget rule) | efficiency 0.9730 | trades 8.17 | sessions<0 0 |
| ZI-U (no rule) | efficiency 0.8929 | trades 10.00 | sessions<0 0 |
查看完整验算程序
# Chapter 1. p12: zero-intelligence traders in a double auction
# After Gode and Sunder (1993): traders bid at random. ZI-C may not
# buy above value or sell below cost; ZI-U has no budget constraint.
# Values and costs are illustrative (in points). Efficiency = realised
# surplus / maximum surplus, averaged over seeded sessions.
import math
import random
VALUES = [200, 190, 180, 170, 160, 150, 140, 130, 120, 110]
COSTS = [60, 70, 80, 90, 100, 110, 120, 130, 140, 150]
TOP = 220
def max_surplus():
s = []
for v, c in zip(sorted(VALUES, reverse=True), sorted(COSTS)):
if v > c:
s.append(v - c)
return math.fsum(s)
def session(rng, constrained, steps=600):
buyers = list(range(len(VALUES)))
sellers = list(range(len(COSTS)))
bid, ask = None, None # (price, trader)
gains = []
for _ in range(steps):
if not buyers or not sellers:
break
if rng.random() < 0.5:
b = rng.choice(buyers)
hi = VALUES[b] if constrained else TOP
p = rng.uniform(1, hi)
if ask is not None and p >= ask[0]:
s = ask[1]
gains.append(VALUES[b] - COSTS[s])
buyers.remove(b)
sellers.remove(s)
bid, ask = None, None
elif bid is None or p > bid[0]:
bid = (p, b)
else:
s = rng.choice(sellers)
lo = COSTS[s] if constrained else 1
p = rng.uniform(lo, TOP)
if bid is not None and p <= bid[0]:
b = bid[1]
gains.append(VALUES[b] - COSTS[s])
buyers.remove(b)
sellers.remove(s)
bid, ask = None, None
elif ask is None or p < ask[0]:
ask = (p, s)
return math.fsum(gains), len(gains)
rng = random.Random(40612)
best = max_surplus()
print("maximum surplus {:.1f} points, {} efficient trades".format(
best, len([1 for v, c in zip(sorted(VALUES, reverse=True),
sorted(COSTS)) if v > c])))
for label, flag in (("ZI-C (budget rule)", True),
("ZI-U (no rule)", False)):
eff, trades, neg = [], [], 0
for _ in range(400):
g, n = session(rng, flag)
eff.append(g / best)
trades.append(n)
if g < 0:
neg += 1
print("{:<20} efficiency {:.4f} trades {:.2f} sessions<0 {}".format(
label, math.fsum(eff) / len(eff),
math.fsum(trades) / len(trades), neg))唯一数据源与引用
范雅福(Fan, Y.). (2026). 《四百卷公理驱动商业兵棋推演、五方敌对代理压力测试与零学费试错沙盘母典》第 1 章「公理表与回合制引擎:十九条结论、零智能代理与一台会破产的沙盘」(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-406 · v1.0.0 · CC-BY-NC-ND-4.0
BibTeX
@book{fan2026406,
author = {Fan, Yafu},
title = {Axiom-Driven Business Wargaming, Multi-Agent Adversarial Stress-Testing and Pre-Launch Zero-Tuition Sandboxes},
year = {2026},
publisher = {5D Library},
edition = {1.0.0},
note = {DOI pending}
}