关于「三明治夹击的上界:滑点容忍度就是写给对手的最大损失」,《做市损耗与三明治上界》给出的定量结论是什么?
结论:第五章只站在受害者一侧:三明治夹击的损失不超过滑点容忍度乘以报价,一阶模型给出对手无利可图的容忍度,例如池深一百万、费率 0.30 %、交易占池深百分之一时约为 0.00427。
数据口径:书中参数为示意,全部计算由书中程序生成。 本书是非个人化投资风险分析,不构成投资建议,不推荐任何代币、协议或交易;加密资产高风险,可能全部损失。
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。 与结论吻合的打印值: 0.00427, 0.30
| smallest profitable tolerance s_min (exact / first order) | |
| fee 0.05 % | |
| depth y0 a | 0.002 a=0.01 a=0.05 |
| 100,000 | 0.20374/0.20000 0.03131/0.03158 0.00578/0.00606 |
| 1,000,000 | 0.02001/0.02000 0.00313/0.00316 0.00058/0.00061 |
| 10,000,000 | 0.00200/0.00200 0.00031/0.00032 0.00006/0.00006 |
| fee 0.30 % | |
| depth y0 a | 0.002 a=0.01 a=0.05 |
| 100,000 none / none | 0.04293/0.04286 0.00611/0.00638 |
| 1,000,000 none / none | 0.00427/0.00429 0.00061/0.00064 |
| 10,000,000 none / none | 0.00043/0.00043 0.00006/0.00006 |
| fee 1.00 % | |
| depth y0 a | 0.002 a=0.01 a=0.05 |
| 100,000 none / none none / none | 0.00729/0.00750 |
| 1,000,000 none / none none / none | 0.00073/0.00075 |
| 10,000,000 none / none none / none | 0.00007/0.00007 |
核心公式
\[\pi \approx 2b\,y_0\,(a-g)-G .\]
查看完整验算程序
# Chapter 5. p52: the tolerance below which no sandwich pays.
# Model: as in p51 (constant-product pool, fee g on input, fixed cost G
# for the adversary's two transactions). For each trade size a (share of
# the Y reserve) and pool depth y0 we find by bisection the smallest
# tolerance s_min at which a worst-case adversary's best profit is
# positive. A victim whose tolerance is below s_min faces no profitable
# sandwich in this model. First-order closed form:
# s_min ~ G / (y0 (a - g)), and no attack at any s when a <= g.
# G = 30 in units of Y; depths and sizes illustrative.
GAS = 30.0
def buy_x(x, y, d, g):
out = x * d * (1.0 - g) / (y + d * (1.0 - g))
return out, x - out, y + d
def sell_x(x, y, e, g):
out = y * e * (1.0 - g) / (x + e * (1.0 - g))
return out, x + e, y - out
def best_profit(a_frac, s, g, y0):
a = a_frac * y0
quote, _, _ = buy_x(y0, y0, a, g)
floor = (1.0 - s) * quote
def run(b):
got, x1, y1 = buy_x(y0, y0, b, g)
out_v, x2, y2 = buy_x(x1, y1, a, g)
back, _, _ = sell_x(x2, y2, got, g)
return out_v, back - b - GAS
lo, hi = 0.0, y0
for _ in range(100):
mid = (lo + hi) / 2.0
if run(mid)[0] >= floor:
lo = mid
else:
hi = mid
return max(run(lo * i / 200.0)[1] for i in range(1, 201))
def s_min(a_frac, g, y0):
if best_profit(a_frac, 0.5, g, y0) <= 0.0:
return None
lo, hi = 0.0, 0.5
for _ in range(60):
mid = (lo + hi) / 2.0
if best_profit(a_frac, mid, g, y0) > 0.0:
hi = mid
else:
lo = mid
return hi
def fmt(v):
return " none " if v is None else "{:.5f}".format(v)
if __name__ == "__main__":
print("smallest profitable tolerance s_min (exact / first order)")
for g in (0.0005, 0.003, 0.01):
print("fee {:.2f} %".format(g * 100))
print(" depth y0 a=0.002 a=0.01 a=0.05")
for y0 in (100000.0, 1000000.0, 10000000.0):
cells = []
for a in (0.002, 0.01, 0.05):
ap = None if a <= g else GAS / (y0 * (a - g))
cells.append("{}/{}".format(fmt(s_min(a, g, y0)), fmt(ap)))
print(" {:>10,.0f} {}".format(y0, " ".join(cells)))唯一数据源与引用
范雅福(Fan, Y.). (2026). 《集中流动性做市商重平衡损耗(LVR)、链上三明治抢跑边界与去中心化金库风控母典》第 5 章「三明治夹击的上界:滑点容忍度就是写给对手的最大损失」(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-344 · v1.0.0 · CC-BY-NC-ND-4.0
BibTeX
@book{fan2026344,
author = {Fan, Yafu},
title = {Concentrated-Liquidity Loss-Versus-Rebalancing, On-Chain Sandwich Bounds and Decentralised Vault Risk Control},
year = {2026},
publisher = {5D Library},
edition = {1.0.0},
note = {DOI pending}
}