国信证券学习系列（8）

我为什么要用国信，就是这个原因，可以做期权，期货，股票，etf，可转债的回测。满足了我所有的需要，我要做指数增强。通常的做法是股票和期货。但实际上，股票和期权做组合，成本更低。

（1）认沽期权和认购期权的隐含波动率大于50etf波动率时，
以标的资产当前的价格作为行权价买入相同数量、相同标的资产和相同行权日的认购和认沽期权开仓
（2）认沽期权和认购期权的隐含波动率小于50etf波动率且有市场时，平仓
获取数据：

import datetime
from math import (log, pow, sqrt, exp)from scipy import stats
import numpy as np
import talibcdf = stats.norm.cdf
STEP_CHANGE = 0.001
STEP_UP = 1 + STEP_CHANGE
STEP_DOWN = 1 - STEP_CHANGE
STEP_DIFF = STEP_CHANGE * 2
DX_TARGET = 0.00001
'''
变量说明
s：标的物股票价格
k：行权价
r：无风险利率
t：剩余到期时间（年）
v：隐含波动率
cp：期权类型，+1/-1对应call/put
price：期权价格
'''

初始参数：

def init(ContextInfo):ContextInfo.contract_unit = 10000#随着时间推移该对合约可能不存了，根据需求手动修改更新ContextInfo.optOne="10001713" ContextInfo.optTwo="10001714"#上面合约的到期年月，根据需求手动修改更新ContextInfo.aa= get_week_of_month(2019, 3)ContextInfo.holdings = {ContextInfo.optOne: 0, ContextInfo.optTwo: 0}ContextInfo.accountID = '8861103625'ContextInfo.s_volatility = []

计算隐含波动率：

def calculateImpv(price, s, k, r, t, cp):"""计算隐含波动率"""# 检查期权价格必须为正数if price <= 0:return 0# 检查期权价格是否满足最小价值（即到期行权价值）meet = Falseif cp == 1 and (price > (s - k) * exp(-r * t)):meet = Trueelif cp == -1 and (price > k * exp(-r * t) - s):meet = True# 若不满足最小价值，则直接返回0if not meet:return 0# 采用Newton Raphson方法计算隐含波动率v = 0.29 # 初始波动率猜测for i in range(50):# 计算当前猜测波动率对应的期权价格和vega值p = calculatePrice(s, k, r, t, v, cp)vega = calculateOriginalVega(s, k, r, t, v, cp)# 如果vega过小接近0，则直接返回if not vega:break# 计算误差dx = (price - p) / vega# 检查误差是否满足要求，若满足则跳出循环if abs(dx) < DX_TARGET:break# 计算新一轮猜测的波动率v += dx# 检查波动率计算结果非负if v <= 0:return 0# 保留4位小数v = round(v, 4)return v

计算vega：

def calculateOriginalVega(s, k, r, t, v, cp):"""计算原始vega值"""    price1 = calculatePrice(s, k, r, t, v*STEP_UP, cp)price2 = calculatePrice(s, k, r, t, v*STEP_DOWN, cp)vega = (price1 - price2) / (v * STEP_DIFF)return vega

计算期权价格：

def calculatePrice(s, k, r, t, v, cp):"""计算期权价格"""# 如果波动率为0，则直接返回期权空间价值if v <= 0:return max(0, cp * (s - k))d1 = (log(s / k) + (r + 0.5 * pow(v, 2)) * t) / (v * sqrt(t))d2 = d1 - v * sqrt(t)price = cp * (s * cdf(cp * d1) - k * cdf(cp * d2) * exp(-r * t))return price

获取日期：

def get_week_of_month(year, month):"""获取指定的某天是某个月中的第几周周一作为一周的开始"""begin = datetime.datetime(year, month, 1).weekday()delta = datetime.timedelta(days=21+begin)exercise_date = datetime.datetime(year, month, 1) + deltareturn  exercise_date.strftime('%Y-%m-%d')

策略代码：

#coding:gbk
"""
跨式期权策略
回测模型示例（非实盘交易策略）（1）认沽期权和认购期权的隐含波动率大于50etf波动率时，
以标的资产当前的价格作为行权价买入相同数量、相同标的资产和相同行权日的认购和认沽期权开仓
（2）认沽期权和认购期权的隐含波动率小于50etf波动率且有市场时，平仓
"""
import datetime
from math import (log, pow, sqrt, exp)from scipy import stats
import numpy as np
import talibcdf = stats.norm.cdf
STEP_CHANGE = 0.001
STEP_UP = 1 + STEP_CHANGE
STEP_DOWN = 1 - STEP_CHANGE
STEP_DIFF = STEP_CHANGE * 2
DX_TARGET = 0.00001
'''
变量说明
s：标的物股票价格
k：行权价
r：无风险利率
t：剩余到期时间（年）
v：隐含波动率
cp：期权类型，+1/-1对应call/put
price：期权价格
'''def init(ContextInfo):ContextInfo.contract_unit = 10000#随着时间推移该对合约可能不存了，根据需求手动修改更新ContextInfo.optOne="10001713" ContextInfo.optTwo="10001714"#上面合约的到期年月，根据需求手动修改更新ContextInfo.aa= get_week_of_month(2019, 3)ContextInfo.holdings = {ContextInfo.optOne: 0, ContextInfo.optTwo: 0}ContextInfo.accountID = '8861103625'ContextInfo.s_volatility = []
def handlebar(ContextInfo):#当前K线的对应的下标从0开始index = ContextInfo.barpos#当前K线对应的时间：毫秒realtime = ContextInfo.get_bar_timetag(index)current_time = timetag_to_datetime(realtime,'%Y-%m-%d')#当前周期period = ContextInfo.period#取当前K线图对应的合约当前K线的当前主图复权方式下的收盘价price_call = ContextInfo.get_market_data(['close'],period=period, stock_code = [ContextInfo.optOne+".SHO"]) price_put = ContextInfo.get_market_data(['close'],period=period, stock_code = [ContextInfo.optTwo+".SHO"])s = ContextInfo.get_market_data(["close"],stock_code = ["510050.SH"],period=period)#print price_call,price_putContextInfo.s_volatility.append(s)s_volatility = np.std(np.array(ContextInfo.s_volatility))ContextInfo.paint("s_volatility", s_volatility*10, -1, 0)k = 2.55r = ContextInfo.get_risk_free_rate(index)/100t = ((datetime.datetime.strptime(ContextInfo.aa, "%Y-%m-%d") - datetime.datetime.strptime(current_time, "%Y-%m-%d")).days) / 365.0cp_call = 1cp_put = -1implied_volatility_call = calculateImpv(price_call, s, k, r, t, cp_call)implied_volatility_put = calculateImpv(price_put, s, k, r, t, cp_put)#print "implied_volatility_call", implied_volatility_call#print "implied_volatility_put", implied_volatility_putContextInfo.paint("implied_volatility", (implied_volatility_put + implied_volatility_call), -1, 0)if (implied_volatility_put + implied_volatility_call) > s_volatility*10 :passorder(50,1101,ContextInfo.accountID ,ContextInfo.optOne,5,-1,1,ContextInfo)passorder(50,1101,ContextInfo.accountID ,ContextInfo.optTwo,5,-1,1,ContextInfo)ContextInfo.holdings[ContextInfo.optOne] = 1ContextInfo.holdings[ContextInfo.optTwo] = 1elif ContextInfo.holdings[ContextInfo.optOne] == 1 and ContextInfo.holdings[ContextInfo.optTwo] == 1 and (implied_volatility_put + implied_volatility_call) < s_volatility*10 :passorder(53,1101,ContextInfo.accountID ,ContextInfo.optOne,5,-1,1,ContextInfo)passorder(53,1101,ContextInfo.accountID ,ContextInfo.optTwo,5,-1,1,ContextInfo)ContextInfo.holdings[ContextInfo.optOne] = 0ContextInfo.holdings[ContextInfo.optTwo] = 0def calculateImpv(price, s, k, r, t, cp):"""计算隐含波动率"""# 检查期权价格必须为正数if price <= 0:return 0# 检查期权价格是否满足最小价值（即到期行权价值）meet = Falseif cp == 1 and (price > (s - k) * exp(-r * t)):meet = Trueelif cp == -1 and (price > k * exp(-r * t) - s):meet = True# 若不满足最小价值，则直接返回0if not meet:return 0# 采用Newton Raphson方法计算隐含波动率v = 0.29 # 初始波动率猜测for i in range(50):# 计算当前猜测波动率对应的期权价格和vega值p = calculatePrice(s, k, r, t, v, cp)vega = calculateOriginalVega(s, k, r, t, v, cp)# 如果vega过小接近0，则直接返回if not vega:break# 计算误差dx = (price - p) / vega# 检查误差是否满足要求，若满足则跳出循环if abs(dx) < DX_TARGET:break# 计算新一轮猜测的波动率v += dx# 检查波动率计算结果非负if v <= 0:return 0# 保留4位小数v = round(v, 4)return vdef calculateOriginalVega(s, k, r, t, v, cp):"""计算原始vega值"""    price1 = calculatePrice(s, k, r, t, v*STEP_UP, cp)price2 = calculatePrice(s, k, r, t, v*STEP_DOWN, cp)vega = (price1 - price2) / (v * STEP_DIFF)return vegadef calculatePrice(s, k, r, t, v, cp):"""计算期权价格"""# 如果波动率为0，则直接返回期权空间价值if v <= 0:return max(0, cp * (s - k))d1 = (log(s / k) + (r + 0.5 * pow(v, 2)) * t) / (v * sqrt(t))d2 = d1 - v * sqrt(t)price = cp * (s * cdf(cp * d1) - k * cdf(cp * d2) * exp(-r * t))return pricedef get_week_of_month(year, month):"""获取指定的某天是某个月中的第几周周一作为一周的开始"""begin = datetime.datetime(year, month, 1).weekday()delta = datetime.timedelta(days=21+begin)exercise_date = datetime.datetime(year, month, 1) + deltareturn  exercise_date.strftime('%Y-%m-%d')