关于「六十条规则、四种情景:止盈比例与时间止损的模拟检验」,《期权提前止盈与最优停时》给出的定量结论是什么?

结论:第五章用六十条规则、四种情景检验从业者经验法则:对借方价差的买方,按年化夏普排第一的都是拿到期;对卖方,50% 止盈加 21 日止损在零成本时把夏普从 0.338 提到 0.558,成本 0.05 点即告翻转;即使差别真实,单个交易者也需要 73.1 年记录才能把它与零分开。

数据口径:书中价格一律以指数点表示,路径全部合成,参数为示意;全部计算由书中程序生成。 书中以标注框区分已发表的证据、可运行的 Python 计算(21 段程序)、原典与作者的假说。

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: 0.05, 0.338, 0.558, 21, 50

premium scenario, 3000 paths
costrulebuyer annSharpeseller annSharpe
0.00hold/0-1.116-0.338+0.631+0.338
0.0050%/0-1.367-0.443+0.855+0.433
0.0065%/0-1.367-0.429+0.878+0.437
0.0050%/21-1.321-0.451+0.972+0.558
0.00hold/21-1.247-0.425+0.705+0.425
0.05hold/0-1.304-0.406+0.496+0.270
0.0550%/0-1.701-0.575+0.441+0.230
0.0565%/0-1.657-0.540+0.538+0.275
0.0550%/21-2.177-0.763+0.314+0.183
0.05hold/21-2.033-0.713+0.225+0.138
0.10hold/0-1.483-0.474+0.365+0.202
0.1050%/0-2.018-0.709+0.041+0.022
0.1065%/0-1.932-0.654+0.207+0.109
0.1050%/21-2.988-1.076-0.323-0.191
0.10hold/21-2.777-1.000-0.240-0.150
0.20hold/0-1.813-0.610+0.116+0.066
0.2050%/0-2.605-0.990-0.725-0.411
0.2065%/0-2.440-0.889-0.425-0.235
0.2050%/21-4.488-1.701-1.541-0.941
0.20hold/21-4.156-1.575-1.129-0.724
查看完整验算程序
# Chapter 5. p53: costs, and the other side of the same spread. Paths
# of the "premium" scenario of p51 (index points, synthetic).
# Heston-type variance with jumps, one step a day: kappa 8, long-run
# variance 0.16^2, vol of vol 0.5, correlation -0.7; jumps 1.5 a year,
# log size mean -0.05 and sd 0.03, each jump adds 0.04 to variance.
# Marks use the expected average variance over the remaining life,
# times a premium factor m, plus jump variance. Scenarios: "bs" is the
# control, constant vol 0.16, no jumps, marks exact (a martingale);
# "sv" m 1.00; "premium" m 1.10; "edge" m 1.10 and a drift of -0.15 a
# year that the holder forecast correctly (zero drift otherwise).
# The buyer pays the debit; the seller of the same spread receives it
# as a credit, and takes profit at a share of that credit. A cost c is
# paid to open and again to close early (expiry is free). Each row is
# one rule at one cost, for the buyer and for the seller.
import bisect
import math
import random
import statistics

KAP, TH, XI, RHO = 8.0, 0.16 ** 2, 0.5, -0.7
LAM, MJ, SJ, VJ = 1.5, -0.05, 0.03, 0.04
DAYS, DT, NPATH = 45, 1 / 365, 3000
JV = LAM * (MJ * MJ + SJ * SJ)
KBAR = math.exp(MJ + 0.5 * SJ * SJ) - 1.0


def ncdf(x):
    return 0.5 * math.erfc(-x * 0.7071067811865476)


def spread(s, t, vol):
    if t <= 0.0:
        return min(max(100.0 - s, 0.0), 5.0)
    sd = vol * math.sqrt(t)
    ls = math.log(s)
    out = []
    for k in (100.0, 95.0):
        d1 = (ls - math.log(k) + 0.5 * sd * sd) / sd
        out.append(k * ncdf(sd - d1) - s * ncdf(-d1))
    return out[0] - out[1]


def iv(v, t, m):
    x = KAP * t
    return math.sqrt(m * (TH + (v - TH) * (1.0 - math.exp(-x)) / x) + JV)


def paths(m, mu, rng, bs=False):
    if bs:
        return bs_paths(rng)
    out = []
    for _ in range(NPATH):
        s, v = 100.0, TH
        marks = [spread(s, DAYS * DT, iv(v, DAYS * DT, m))]
        for j in range(1, DAYS + 1):
            z1 = rng.gauss(0.0, 1.0)
            z2 = RHO * z1 + math.sqrt(1 - RHO * RHO) * rng.gauss(0.0, 1.0)
            vp = max(v, 0.0)
            ls = (mu - LAM * KBAR - 0.5 * vp) * DT + math.sqrt(vp * DT) * z1
            v = v + KAP * (TH - vp) * DT + XI * math.sqrt(vp * DT) * z2
            if rng.random() < LAM * DT:
                ls += rng.gauss(MJ, SJ)
                v += VJ
            s *= math.exp(ls)
            t = (DAYS - j) * DT
            marks.append(spread(s, t, iv(max(v, 1e-6), t, m)) if t > 0
                         else spread(s, 0.0, 0.0))
        out.append(marks)
    return out


def bs_paths(rng):
    out, vol = [], math.sqrt(TH)
    for _ in range(NPATH):
        s = 100.0
        marks = [spread(s, DAYS * DT, vol)]
        for j in range(1, DAYS + 1):
            s *= math.exp(-0.5 * TH * DT
                          + vol * math.sqrt(DT) * rng.gauss(0.0, 1.0))
            marks.append(spread(s, (DAYS - j) * DT, vol))
        out.append(marks)
    return out


def run_rule(ps, tp, stop, cost, side):
    res = []
    for marks in ps:
        m0 = marks[0]
        if side == "buy":
            cap = m0 + cost
            share = [(x - m0) / (5.0 - m0) for x in marks]
        else:
            cap = 5.0 - m0 + cost
            share = [(m0 - x) / m0 for x in marks]
        pm, best = [], -9.0
        for x in share[1:DAYS]:
            best = max(best, x)
            pm.append(best)
        j = min(bisect.bisect_left(pm, tp) + 1,
                DAYS - stop if stop else DAYS)
        c_out = cost if j < DAYS else 0.0
        if side == "buy":
            pnl = marks[j] - m0 - cost - c_out
        else:
            pnl = m0 - marks[j] - cost - c_out
        res.append((pnl, cap, j))
    return res


def stats(res):
    ret = [r[0] / r[1] for r in res]
    q = 365 / statistics.fmean([r[2] for r in res])
    cap = math.fsum(r[1] * r[2] for r in res)
    return (math.fsum(r[0] for r in res) / cap * 365,
            statistics.fmean(ret) / statistics.pstdev(ret) * math.sqrt(q))


rng = random.Random(42653)
ps = paths(1.10, 0.0, rng)
RULES = ((9.0, 0), (0.50, 0), (0.65, 0), (0.50, 21), (9.0, 21))
print("premium scenario, %d paths" % NPATH)
print("cost  rule     buyer ann  Sharpe   seller ann  Sharpe")
for cost in (0.0, 0.05, 0.10, 0.20):
    for tp, stop in RULES:
        b = stats(run_rule(ps, tp, stop, cost, "buy"))
        s = stats(run_rule(ps, tp, stop, cost, "sell"))
        lab = ("hold" if tp > 1 else "%d%%" % round(tp * 100))
        print("%.2f  %-7s  %+.3f    %+.3f    %+.3f     %+.3f"
              % (cost, lab + "/%d" % stop, b[0], b[1], s[0], s[1]))

唯一数据源与引用

范雅福(Fan, Y.). (2026). 《期权提前止盈最优停时、恐慌脉冲波动率收割与持有至到期回吐陷阱母典》第 5 章「六十条规则、四种情景:止盈比例与时间止损的模拟检验」(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-426 · v1.0.0 · CC-BY-NC-ND-4.0

BibTeX
@book{fan2026426,
  author    = {Fan, Yafu},
  title     = {Optimal Early Profit-Taking in Options: Harvesting Volatility Spikes, Twenty-One-Day Decay Cliffs and the Expiration Hold Trap},
  year      = {2026},
  publisher = {5D Library},
  edition   = {1.0.0},
  note      = {DOI pending}
}

本书其他章节结论