［策略］基于胜率的趋势交易策略
策略说明
简单构建了一个基于胜率的趋势交易策略。认为过去一段时间（N天）内胜率较高、信息比率较高的股票会在紧随其后的几天有较好的表现
1）先根据胜率要求筛选出过去N天胜率高的股票作为预选股票（benchmark可以是定义的确定阈值，或者是某个指数相应的收益率），用aprior算法进行快速筛选。第i只股票胜率的计算方式如下:
winRate(i) = sum([sign(ret(i,t)-ret(bm,t))==1]/N)|t~(t-N,t)
＊ret(i,t): i股票在第t天的收益率；
＊ret（bm,t）: benchmark在第t天的收益率；
2）从筛选出的股票中选择过去N天信息比率（收益率／波动率）高的部分股票构建备选投资组合； 3）依据被选投资组合做买入操作，使用可用资金的50%～70%； 4）设定股票止损位在收益下跌至0.95，止损时将仓位调整至原仓位的40%～60%； 5）调仓频率为5天，股票池为沪深300。
import numpy as np
from CAL.PyCAL import *
################################################################################
# Back Test Functions
################################################################################
def initialize(account): # 初始化虚拟账户状态
pass
####init the univese of the choosen stock
def universeInit():
stockComponent = DataAPI.MktTickRTSnapshotIndexGet(securityID=u"000300.XSHG",field=u"lastPrice,shortNM")
stockCount = len(stockComponent)
stockTicker = stockComponent['ticker']
stockExchgID = stockComponent['exchangeCD']
stockID = []
for index in range(stockCount):
stockID.append(stockTicker[index] + '.' + stockExchgID[index])
return stockID
####deal with the trading signals
def handle_data(account): # 每个交易日的买入卖出指令
####Presettings
histLength = 10
stockDataThres = 0
####Dictionary of the return Rate
closePrice = account.get_attribute_history('closePrice',histLength)
retRate = {}
for index in account.universe:
retRate[index] = ((closePrice[index][1:] - closePrice[index][:-1])/closePrice[index][:-1]).tolist()
###ret list of the benchmark
calendar = Calendar('China.SSE')
startDate = calendar.advanceDate(account.current_date,'-'+str(histLength)+'B').toDateTime()
benchmark = DataAPI.MktIdxdGet(ticker = "000300",
field = "closeIndex",
beginDate = startDate,
endDate = account.current_date,pandas = '1')
bmClose = benchmark['closeIndex'].tolist()
bmRet = []
for index in range(len(bmClose)-1):
bmRet.append((bmClose[1:][index]-bmClose[:-1][index])/bmClose[:-1][index])
####List of transactions
transactions = []
for index in range(histLength-1):
tmpt = []
for stock in account.universe:
if retRate[stock][index] > stockDataThres:
# if retRate[stock][index] > bmRet[index]:
tmpt.append(stock)
transactions.append(tmpt)
####List of hot stocks
hotStock = []
hotStockDict,hotStockList = apriori(transactions,0.95)
for index in hotStockList:
for stock in index:
if stock not in hotStock:
hotStock.append(stock)
####List of the portfolio
retRate = {}
fluctRate = {}
sharpRate = {}
for index in hotStock:
retRate[index] = ((closePrice[index][-1] - closePrice[index][0])/closePrice[index][0])
fluctRate[index] = np.std(closePrice[index])
sharpRate[index] = retRate[index]/fluctRate[index]
portfolio = [index[0] for index in sorted(sharpRate.items(),key = lambda sharpRate:sharpRate[1])[-len(sharpRate)/2:]]
####Stop loss at -0.05
validSecHist = account.get_attribute_history('closePrice',2)
for index in account.valid_secpos:
if (validSecHist[index][-1] - validSecHist[index][0])/validSecHist[index][0] < -0.05:
order_to(index,0.45*account.valid_secpos[index])
####Buy portfolio
for index in portfolio:
amount = 0.65*account.cash/len(hotStock)/account.referencePrice[index]
order(index,amount)
return
########################################################################################
# Aprior algorithm
########################################################################################
def elementsDet(datasets):
if type(datasets) == list:
elements = {}
for index in datasets:
for index1 in index:
if elements.has_key(index1) == False:
elements[index1] = 1
else:
elements[index1] += 1
return elements
if type(datasets) == dict:
elements = {}
for index in datasets:
if type(index) == tuple:
index = list(index)
for index1 in index:
if elements.has_key(index1) == False:
elements[index1] = 0
else:
elements[index] = 0
return elements
pass
def checkAssociation(subset,objset):
for index in subset:
if index not in objset:
return False
return True
pass
def support(subset,datasets):
count = 0
for transaction in datasets:
if checkAssociation(subset,transaction) == True:
count += 1
return 1.0*count/len(datasets)
pass
def apriori(datasets,minsup):
candidateIterator = []
electIterator = []
length = len(datasets)
##init part
#the candidate
elements = elementsDet(datasets)
candidate = {}
for index in elements:
candidate[index] = 1.0*elements[index]/length
candidateIterator.append(candidate)
#the elect
elect = {}
for index in candidate:
if candidate[index] > minsup:
elect[index] = candidate[index]
electIterator.append(elect)
##the update part
itera = 1
while(len(electIterator[-1]) != 0):
candidateOld = candidateIterator[-1]
electOld = electIterator[-1]
elementsOld = elementsDet(electOld)
# print elementsOld
candidate = {}
##the candidate
for index in electOld:
for index1 in elementsOld:
if type(index) != list and type(index) != tuple:
if index1 != index:
tmp = []
tmp.append(index)
tmp.append(index1)
tmp.sort()
if candidate.has_key(tuple(tmp)) == False:
candidate[tuple(tmp)] = 0
if type(index) == tuple:
tmp = list(index)
if tmp.count(index1) == False:
tmp1 = tmp
tmp1.append(index1)
tmp1.sort()
if candidate.has_key(tuple(tmp1)) == False:
candidate[tuple(tmp1)] = 0
candidateIterator.append(candidate)
##the elect
elect = {}
for index in candidate:
candidate[index] = support(index,datasets)
for index in candidate:
if candidate[index] > minsup:
elect[index] = candidate[index]
electIterator.append(elect)
# print 'iteartion ' + str(itera) + ' is finished!'
itera += 1
##the elected frequency sets dictionary: the value is the key's support
electedDict = {}
for index in electIterator:
for index1 in index:
electedDict[index1] = index[index1]
##the elected frequency sets lists
electedList = []
for index in electIterator:
tmp = []
for index1 in index:
if type(index1) == tuple:
tmp1 = []
for ele in index1:
tmp1.append(ele)
tmp.append(tmp1)
else:
tmp.append([str(index1)])
tmp.sort()
for index1 in tmp:
electedList.append(index1)
return electedDict,electedList
################################################################################
# Back Test Presetting
################################################################################
start = '2011-01-01' # 回测起始时间
end = '2015-11-01' # 回测结束时间
benchmark = 'HS300' # 策略参考标准
universe = set_universe('HS300')
# universe = universeInit() # 证券池，支持股票和基金
capital_base = 100000 # 起始资金
freq = 'd' # 策略类型，'d'表示日间策略使用日线回测，'m'表示日内策略使用分钟线回测
refresh_rate = 5 # 调仓频率，表示执行handle_data的时间间隔，若freq = 'd'时间间隔的单位为交易日，若freq = 'm'时间间隔为分钟
策略表现
- 策略能产生一定的alpha；
- 策略表现与起点强相关，sharpRatio不稳定；
- 策略表现会受到自身参数设定的影响，例如胜率选择周期、筛选阈值、调仓频率、建仓头寸、止损仓位等，需要依据表现对其进行优化；
- 策略在2011年4月至12月、2015年6月到11月有相对好的表现，可见其相对较适用于趋势下跌的市场环境。
问题探讨
因子选股模型的流程应该是怎样的？
小编认为构建因子选股的模型需要有如下过程：
- 大类配置：根据宏观判断市场，进行市场判断（根据不同市场选择不同因子）、资产配置（不同风险性证券的配比选择➡️不同热度的行业配比选择）和策略选择（市场中性、单边做多等）；
- 选股－alpha端：对选股因子进行有效性分析，包括单因子的预测性、因子间相关性，构建多因子模型使得选股有尽可能高的alpha；
- 选股－风险端：对alpha端的多因子模型进行风险评估，根据风险因子优化模型，使模型尽可能达到有效边界；
- 择时－买卖时点：对根据因子模型选出的股票进行择时分析，进一步筛选投资组合中的股票及判断作何操作；
因子选股中比较basic的问题，欢迎社区的小伙伴们发表看法、评论和拍醒~