DilphinDB使用案例11：动量交易策略

• 动量策略

  动量策略是最著名的长短期股票策略之一。

  自从Jegadeesh和Titman(1993)首次提出这个概念以来，它广泛出现在学术研究和销售方面的著作中。

  投资者在动量策略中相信，个股中，过去的赢家将超越过去的输家。

  最常用的动量因素是股票出去最近一个月，在过去12个月的收益。

  在学术出版物中，动量策略通常是一个月调整一次且持有期也是一个月。

  在本例中， 我们每天重新平衡我们的投资组合的1/21，并持有新份额21天。为简单起见，我们不考虑交易成本。

• 加载数据并整理

• 加载股票数据

  # 载入股票数据
  CH = ploadText("D:/DolphinDB/Data/CHstock1990_2018.csv")
  

  

• 整理数据并计算动量信号

  # 计算动量信号
  def loadPriceData(inData){
         CHstocks = select TS_CODE, TRADE_date, abs(open) as PRC, vol as VOL,close/open-1 as RET from inData where weekday(trade_date) between 1:5, isValid(open), isValid(VOL) order by ts_code, trade_date
         CHstocks = select ts_code, trade_date, PRC, VOL, RET, cumprod(1+RET) as cumretIndex from CHstocks context by ts_code
         return select ts_code, trade_date, PRC, VOL, RET, move(cumretIndex,21)\move(cumretIndex,252)-1 as signal from CHstocks context by ts_code
  }
  
  priceData = loadPriceData(CH)
  

  

  从SQL语句中可以看到，数据库字段不区分大小写。

  一般的数据库SQL语言是不区分大小写的，但是DolphinDB区分大小写，sql关键字只能用小写。这里区分大小写指的是sql关键字。

• 释放内存占用

  # 释放内存
  undef(`CH, VAR)
  # undef函数释放对象需要是字符串标量或向量，所以在CH前加上`
  

• 为动量策略生成投资组合

• 筛选个股

  # 筛选个股
  def genTradables(indata){
  	return select trade_date, ts_code, signal from indata where PRC>5, VOL>0, isValid(signal) order by trade_date
  }
  tradables = genTradables(priceData)
  

  

  筛选条件是，动量信号值无缺失、当日是正交易量、动量信号无缺失、股价大于5元。

• 计算每日收益

  # WtScheme=1表示等权重 ，WtScheme=2表示值权重
  def formPortfolio(startDate, endDate, tradables, holdingDays, groups, WtScheme){
  	ports = select trade_date, ts_code, rank(signal,,groups) as rank, count(ts_code) as symCount, 0.0 as wt from tradables where trade_date between startDate:endDate context by trade_date having count(ts_code) >= 100
  	if (WtScheme == 1){
  		update ports set wt = -1.0\count(ts_code)\holdingDays where rank=0 context by trade_date
  		update ports set wt = 1.0\count(ts_code)\holdingDays where rank=groups-1 context by trade_date
  	}
  	else if (WtScheme == 2){
  		# 缺数据未计算
  	}
  	return select ts_code, trade_date as tranche, wt from ports where wt != 0 order by ts_code, trade_date
  }
  startDate = 1996.01.01
  endDate = 2018.01.01
  holdingDays = 21
  groups =10
  ports = formPortfolio(startDate, endDate, tradables, holdingDays, groups, 1)
  dailyRtn = select trade_date, ts_code, RET as dailyRet from priceData where trade_date between startDate:endDate
  

  根据动量信号，制定10组可交易股票。只保留2个最极端的群体（赢家和输家）。假设我们总是想在21天内，每天多头1美元和空头$1，所以我们每天在赢家组多头$1/21，在输家组每天空头$1/21。在每组中，我们可以使用等权重或值权重， 来计算投资组合形成日期上每个股票的权重。

• 21天内每支股票利润走势

  # 计算每支个股21天损益
  def calcStockPnL(ports, dailyRtn, holdingDays, endDate, lastDays){
         ages = table(1..holdingDays as age)
         dates = sort distinct ports.tranche
         dictDateIndex = dict(dates, 1..dates.size())
         dictIndexDate = dict(1..dates.size(), dates)
         pos = select dictIndexDate[dictDateIndex[tranche]+age] as trade_date, ts_code, tranche, age, take(0.0,size age) as ret, wt as expr, take(0.0,size age) as pnl from cj(ports,ages) where isValid(dictIndexDate[dictDateIndex[tranche]+age]) and dictIndexDate[dictDateIndex[tranche]+age]<=min(lastDays[ts_code], endDate)
         update pos set ret = RET from ej(pos, dailyRtn,`trade_date`ts_code)
         update pos set expr = expr*cumprod(1+ret) from pos context by ts_code, tranche
         update pos set pnl = expr*ret/(1+ret)
         return pos
  }
  lastDaysTable = select max(trade_date) as date from priceData group by ts_code
  lastDays = dict(lastDaysTable.ts_code, lastDaysTable.date)
  undef(`priceData, VAR)
  stockPnL = calcStockPnL(ports, dailyRtn, holdingDays, endDate, lastDays)
  

• 计算投资组合的利润/损失，并绘图

  # 计算总体收益并绘图
  portPnL = select sum(pnl) as pnl from stockPnL group by trade_date
  portPnL = select * from portPnL order by trade_date
  plot(cumsum(portPnL.pnl) as cumulativeReturn,portPnL.trade_date, "Cumulative Returns of the Momentum Strategy")
  

最后两部分还没完全搞懂，主要是示例中所用美股数据字段与A股数据不同，后续重新构建一个数据集再把这块补上