5D Library › 期权隐含波动率曲面、偏度溢价与垂直价差结构的定量风控 · English

第三章 熊市看跌垂直价差:几何、成本与提前行权

封顶的收益与封顶的亏损

熊市看跌垂直价差(bear put spread)由两条腿组成:买入一个较高行权价 K1K_1 的看跌期权,同时卖出一个同一到期日、较低行权价 K2K_2 的看跌期权。开仓时付出净权利金 DD(「借方」),到期时的收益为

payoff(ST)=max⁡(K1−ST,0)−max⁡(K2−ST,0)=min⁡(max⁡(K1−ST,0),K1−K2).\text{payoff}(S_T) = \max(K_1 – S_T, 0) – \max(K_2 – S_T, 0) = \min\big(\max(K_1 – S_T, 0),\ K_1 – K_2\big).

于是,在欧式期权、持有到期、DD 包含全部交易成本的条件下:

  • 最大收益 =(K1−K2)−D= (K_1 – K_2) – D,在 ST≤K2S_T \le K_2 时取得;
  • 最大亏损 =D= D,在 ST≥K1S_T \ge K_1 时发生;
  • 盈亏平衡点 =K1−D= K_1 – D。

框架中的这三条公式核对无误,但三个条件都不能省略:提前平仓时,价差的价值由当时的市场价格决定,并要再付一次买卖价差;美式期权可能被提前指派;DD 若按中间价计算,会低估成本。单位说明:书中价格都按每一单位标的计,交易所的标准合约通常乘以 100。

与卖出看跌期权或贷方价差(例如牛市看跌价差)不同,借方价差的最大亏损在开仓时已经全额付清,所以只要两条腿都还在(空头腿没有被指派),它没有追加保证金的问题;贷方价差则需要按「行权价差 − 收到的权利金」缴纳保证金,并承担被指派的风险。

偏度与买卖价差的账

用第二章的偏度设定:行权价 100 的看跌期权隐含波动率 0.22,行权价 95 的为 0.245,剩余 30 天。下面的程序计算价差的成本、偏度的影响,以及买卖价差(每条腿的半价差 hh)对最大收益、最大亏损和盈亏平衡点的影响;最后假设标的的实际波动率为 0.18(低于隐含波动率),在不同的实际漂移下计算到期时落在盈亏平衡点以下的概率与期望盈亏。

推导

# Bear put spread geometry: buy the 100 put, sell the 95 put, 30 days, with a skewed
# implied-volatility quote (higher vol at the lower strike). Capped gain and loss,
# break-even, the cost of crossing bid-ask spreads, and the chance of finishing in profit
# under assumed real-world drifts. Prices per unit of underlying; ILLUSTRATIVE numbers.
import math

def ncdf(x):
    return 0.5 * (1.0 + math.erf(x / math.sqrt(2.0)))

def put(S, K, T, r, s):
    d1 = (math.log(S / K) + (r + 0.5 * s * s) * T) / (s * math.sqrt(T))
    return K * math.exp(-r * T) * ncdf(-d1 + s * math.sqrt(T)) - S * ncdf(-d1)

S, T, r, K1, K2, V1, V2 = 100.0, 30 / 365, 0.03, 100.0, 95.0, 0.22, 0.245

def e_put(K, mu, s=0.18):
    # E[(K - S_T)+] for a lognormal with drift mu (not discounted)
    d1 = (math.log(S / K) + (mu + 0.5 * s * s) * T) / (s * math.sqrt(T))
    return K * ncdf(-d1 + s * math.sqrt(T)) - S * math.exp(mu * T) * ncdf(-d1)

def p_below(x, mu, s=0.18):
    # P(S_T < x) for a lognormal with drift mu and volatility s
    return ncdf((math.log(x / S) - (mu - 0.5 * s * s) * T) / (s * math.sqrt(T)))

if __name__ == "__main__":
    p1, p2 = put(S, K1, T, r, V1), put(S, K2, T, r, V2)
    flat = put(S, K2, T, r, V1)
    print(f"long 100 put {p1:.3f} (vol {V1}), short 95 put {p2:.3f} (vol {V2});"
          f" mid debit {p1 - p2:.3f}")
    print(f"  without skew the 95 put would be {flat:.3f}: the skew lowers the debit by"
          f" {p2 - flat:.3f}")
    print("half-spread  debit  max gain  max loss  gain/loss  break-even  round trip")
    for h in (0.0, 0.02, 0.05, 0.10, 0.20):
        d = p1 - p2 + 2 * h
        g = K1 - K2 - d
        print(f"   {h:.2f}     {d:6.3f}   {g:6.3f}    {d:6.3f}    {g / d:5.2f}"
              f"      {K1 - d:6.2f}     {4 * h / g:5.1%}")
    d = p1 - p2 + 0.10
    print(f"realised vol 0.18, debit {d:.3f}: P(<= break-even) P(<= 95) E[P&L], expiry")
    for lab, mu in (("drift = r (no edge)", r), ("drift -20 %/yr", -0.20),
                    ("drift -60 %/yr", -0.60)):
        ev = math.exp(-r * T) * (e_put(K1, mu) - e_put(K2, mu)) - d
        print(f"  {lab:20s}{p_below(K1 - d, mu):15.3f}{p_below(K2, mu):9.3f}{ev:+11.3f}")
# long 100 put 2.391 (vol 0.22), short 95 put 0.873 (vol 0.245); mid debit 1.518
#   without skew the 95 put would be 0.672: the skew lowers the debit by 0.201
# half-spread  debit  max gain  max loss  gain/loss  break-even  round trip
#    0.00      1.518    3.482     1.518     2.29       98.48      0.0%
#    0.02      1.558    3.442     1.558     2.21       98.44      2.3%
#    0.05      1.618    3.382     1.618     2.09       98.38      5.9%
#    0.10      1.718    3.282     1.718     1.91       98.28     12.2%
#    0.20      1.918    3.082     1.918     1.61       98.08     26.0%
# realised vol 0.18, debit 1.618: P(<= break-even) P(<= 95) E[P&L], expiry
#   drift = r (no edge)           0.368    0.155     -0.070
#   drift -20 %/yr                0.511    0.258     +0.593
#   drift -60 %/yr                0.747    0.495     +1.776

按中间价,价差的成本为 1.518。如果没有偏度、两条腿都按 0.22 定价,卖出的 95 看跌期权只值 0.672;偏度使它值 0.873,于是开仓成本降低了 0.201。这就是第二章假说所说的「偏度越陡,借方看跌价差越便宜」。

买卖价差的影响比想象的大。开仓时付出两个半价差,借方从 1.518 升到 1.518+2h1.518 + 2h;半价差为 0.05 时,最大收益从 3.482 降到 3.382,收益与亏损之比从 2.29 降到 2.09;如果提前平仓,还要再付两个半价差,往返成本相当于最大收益的 5.9%;半价差为 0.20 的流动性差的合约,往返成本达到最大收益的 26%。穆拉维约夫与皮尔逊 2020 年发现,期权交易者通过选择成交时机,实际支付的有效价差可以不到常规估计的 40%,全体平均约比常规估计低四分之一 (Muravyev and Pearson 2020);所以「按报价的一半成交」既可能高估也可能低估成本,回测必须事先写明成交假设(第八章)。

最后三行说明,价差的「胜率」与期望盈亏取决于标的的真实漂移,而定价公式完全不知道这一点。在没有方向优势(漂移等于无风险利率)的情况下,到期落在盈亏平衡点以下的概率是 0.368,期望盈亏为 −0.070:隐含波动率高于实际波动率使买方吃亏,偏度又使卖出的一腿更贵,两者大致抵消,剩下的主要是买卖价差。只有当交易者确实有向下的信息(例如年化 −20% 的漂移),期望盈亏才明显为正(+0.593)。这些漂移都是假设,不是对任何市场的估计。

注

价差不能消除方向判断的错误,只能给错误封顶。 封顶亏损让最坏结果在开仓时就已知,这是借方价差最重要的风险性质;但它不会提高判断的准确率。如果交易者没有真实的方向优势,在成本之后期望收益为负。

美式期权:提前行权与提前指派

股票与多数交易型基金的期权是美式的,持有人可以在到期前任何时候行权。默顿 1973 年证明,标的不付股息时,提前行权美式看涨期权永远不是最优的 (Merton 1973);看跌期权则不同:深度实值时,立即行权拿到 K−SK – S 并把这笔钱存起来赚利息,可能比继续持有更好。下面的程序用 600 步的二叉树比较美式与欧式看跌期权,找出立即行权变得最优的临界价格,并看提前行权对价差的影响。

推导

# American versus European puts on a non-dividend-paying underlying (CRR tree, 600
# steps): the early-exercise premium, the critical price below which exercise is optimal,
# and what early assignment of the short leg does to a 100/95 bear put spread.
import math

def ncdf(x):
    return 0.5 * (1.0 + math.erf(x / math.sqrt(2.0)))

def euro_put(S, K, T, r, s):
    d1 = (math.log(S / K) + (r + 0.5 * s * s) * T) / (s * math.sqrt(T))
    return K * math.exp(-r * T) * ncdf(-d1 + s * math.sqrt(T)) - S * ncdf(-d1)

def amer_put(S, K, T, r, s, n=600):
    dt = T / n; u = math.exp(s * math.sqrt(dt)); d = 1 / u
    p = (math.exp(r * dt) - d) / (u - d); disc = math.exp(-r * dt)
    v = [max(K - S * u ** j * d ** (n - j), 0.0) for j in range(n + 1)]
    for i in range(n - 1, -1, -1):
        v = [max(disc * (p * v[j + 1] + (1 - p) * v[j]), K - S * u ** j * d ** (i - j))
             for j in range(i + 1)]
    return v[0]

def critical(K, T, r, s):
    lo, hi = 1.0, K                     # largest S at which exercising now is optimal
    for _ in range(40):
        mid = 0.5 * (lo + hi)
        ex = amer_put(mid, K, T, r, s, 300) <= K - mid + 1e-9
        lo, hi = (mid, hi) if ex else (lo, mid)
    return lo

if __name__ == "__main__":
    s = 0.22
    for r in (0.03, 0.08):
        print(f"r = {100 * r:.0f} %, put K = 100, vol {s}, 30 days: S  European American"
              "  premium")
        for S in (80, 90, 95, 100):
            e, a = euro_put(S, 100, 30 / 365, r, s), amer_put(S, 100, 30 / 365, r, s)
            print(f"{'':39s}{S:3d}   {e:7.3f}  {max(e, a):7.3f}   {a - e:6.3f}")
        cr = ", ".join(f"{critical(100, d / 365, r, s):.1f} ({d} d)" for d in (30, 7))
        print(f"  exercise now is optimal below S = {cr}")
    r = 0.03; T = 30 / 365
    print("100/95 spread, r = 3 %, 30 days:   S   European  American legs  difference")
    for S in (80, 90, 95, 100):
        e = euro_put(S, 100, T, r, 0.22) - euro_put(S, 95, T, r, 0.245)
        a = amer_put(S, 100, T, r, 0.22) - amer_put(S, 95, T, r, 0.245)
        print(f"{'':34s}{S:3d}   {e:7.3f}   {a:7.3f}         {a - e:+.3f}")
    print("payoff if the short 95 put is assigned at S = 85 and the long 100 put is"
          " exercised")
    print(f"  at once: (100 - 85) - (95 - 85) = {(100 - 85) - (95 - 85)} = spread width;"
          " max loss stays the debit")
# r = 3 %, put K = 100, vol 0.22, 30 days: S  European American  premium
#                                         80    19.754   20.000    0.246
#                                         90     9.883   10.023    0.140
#                                         95     5.526    5.578    0.052
#                                        100     2.391    2.407    0.015
#   exercise now is optimal below S = 88.6 (30 d), 93.4 (7 d)
# r = 8 %, put K = 100, vol 0.22, 30 days: S  European American  premium
#                                         80    19.345   20.000    0.655
#                                         90     9.495   10.000    0.505
#                                         95     5.207    5.392    0.186
#                                        100     2.193    2.246    0.052
#   exercise now is optimal below S = 91.1 (30 d), 94.6 (7 d)
# 100/95 spread, r = 3 %, 30 days:   S   European  American legs  difference
#                                    80     4.972     5.000         +0.028
#                                    90     4.243     4.336         +0.093
#                                    95     2.983     3.021         +0.039
#                                   100     1.518     1.530         +0.012
# payoff if the short 95 put is assigned at S = 85 and the long 100 put is exercised
#   at once: (100 - 85) - (95 - 85) = 5 = spread width; max loss stays the debit

利率为 3%、剩余 30 天时,行权价 100 的美式看跌期权在标的为 80 时的提前行权溢价为 0.246(美式价格就是内在价值 20),在平值时只有 0.015;立即行权在标的低于 88.6 时最优,剩余 7 天时临界价格升到 93.4。利率为 8% 时,溢价与临界价格都更高(平值时 0.052,临界价格 91.1 与 94.6),因为提前拿到行权价的利息更值钱。

对价差而言,两条腿都是美式时,价差的价值比欧式略高(标的为 90 时高 0.093):多头腿的提前行权权利属于自己,空头腿的属于对手,前者行权价更高、更实值,价值更大,所以净效果为正。关键问题是空头腿被提前指派:如果标的跌到 85、95 看跌期权被指派,交易者以 95 买入标的;只要他同时行使(或卖出)100 看跌期权,就以 100 卖出,合计收到行权价差 5,最大亏损仍是借方 DD。风险在于两件事没有同时发生:

  • 钉住风险。 到期日收盘价恰好在空头行权价附近时,交易者不知道自己会不会被指派;如果被指派而多头腿已经作废,下一个交易日开盘就持有一个没有保护的标的头寸。
  • 盘后与跳空。 指派在收盘后通知,若次日跳空,原本封顶的亏损就可能被突破。

这两点都在定价模型之外。实务上的应对是在到期前主动平仓或展期,这正是第五章模型中「到期也要付平仓成本」的来源。

本章小结

熊市看跌垂直价差的最大收益、最大亏损与盈亏平衡点在欧式、持有到期、成本全部计入的条件下成立;借方价差的最大亏损已全额付清,两条腿都在时没有追加保证金的问题。偏度降低了开仓成本,买卖价差则显著压缩收益空间;期望盈亏取决于交易者是否有真实的方向优势,定价公式本身不提供这种优势。美式期权的空头腿可能被提前指派;保留多头腿时亏损仍封顶,但钉住风险与跳空在模型之外。

参考文献

Merton, Robert C. 1973. “Theory of Rational Option Pricing.” The Bell Journal of Economics and Management Science 4 (1): 141–83. https://doi.org/10.2307/3003143.
Muravyev, Dmitriy, and Neil D. Pearson. 2020. “Options Trading Costs Are Lower Than You Think.” The Review of Financial Studies 33 (11): 4973–5014. https://doi.org/10.1093/rfs/hhaa010.

Fan, Y. (2026). Implied Volatility Surfaces, Skew Premia and Quantitative Risk Control of Vertical Spreads: Microstructure Pricing and Time-Stop Dynamics under Non-Gaussian Tails [期权隐含波动率曲面、偏度溢价与垂直价差结构的定量风控] (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.23796765 · Licensed under Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International (CC BY-NC-ND 4.0).