Strategy Library

Short Term Reversal with Futures

Introduction

The short-term reversal strategy buys stocks which are past losers and sells stocks which are past winners. It is commonly used in the equity market. This algorithm will explore the reversal effect in the futures market. Research also suggests that trading volume contains information about future market movements. The algorithm will be constructed with both the volume and return reversal effect.

Method

The daily continuous futures data with the adjusted price is from Quandl. The investment universe consists of 24 types of US futures contracts: 4 currencies, five financial indices, eight agricultural products, and seven commodities. The contracts are all front month contracts.

class Futures(PythonData):
    # import the futures custom data from dropbox
    def GetSource(self, config, date, isLiveMode):
        source = "your_custom_source_url"
        return SubscriptionDataSource(source, SubscriptionTransportMedium.RemoteFile)

    def Reader(self, config, line, date, isLiveMode):
        futures = Futures()
        futures.Symbol = config.Symbol
        data = line.split(',')
        if data[0] == "date": return None
        futures.Time = datetime.strptime(data[0], "%m/%d/%y")
        futures.Value = float(data[4])
        futures["settle"] = float(data[4]) # add the settle price
        futures["volume"] = float(data[5]) # add the volume
        futures["open_interest"] = float(data[6]) # add the open interest
        return futures

The algorithm uses a weekly time frame(Wednesday-Wednesday interval). Therefore the self.Schedule.On() will fire the trading every Wednesday. To compare the weekly change of volume and open interest, we create the deque list with the length to be two to save the history data frame of the most recent week and the week before.

  def Initialize(self):
      self.SetStartDate(2017, 7, 10)
      self.SetEndDate(2018, 8, 1)
      self.SetCash(10000000)

      self.tickers = ["CME_SF1_EF", "CME_MP1_FF",  "CME_CD1_EF", "CME_ED8_FF",               # Currencies
                      "CME_NQ1_EF", "CME_MD1_EF", "CME_ES1_EF", "CME_YM1_EF", "CME_NK1_EF",  # Financial indices
                      "CME_C1_EF", "CME_SM1_EF",  "ICE_CC1_EF",  "CME_LC1_EF",               # Agricultural product
                      "CME_KW1_EF", "CME_S1_EF", "ICE_KC1_EF", "CME_LC1_EF",
                      "CME_HG1_FF",  "CME_GC1_EF", "SHFE_AL1_EF",  "SHFE_CU1_EF",            # Commodities
                      "CME_CL1_EF", "ICE_T1_FR", "CME_HO1_EF"]
      self.length = len(self.tickers)
      for ticker in self.tickers:
          self.AddData(Futures, ticker)
      # create the deque list to save the weekly history dataframe
      self.window = deque(maxlen=2)
      # rebalance the portfolio every week on Wednesday
      self.Schedule.On(self.DateRules.Every(DayOfWeek.Wednesday, DayOfWeek.Wednesday), self.TimeRules.At(0, 0), self.WeeklyTrade)

In the weekly rebalance function, weekly history data frame is saved. The contract is defined as the high(low) volume contract if the contract's volume changes between the period from t-1 to t and period from t-2 to t-1 is above(below) the median volume change of all contracts, In addition, all contracts are also assigned to either high-open interest (top 50% of changes in open interest) or low-open interest groups (bottom 50% of changes in open interest) based on lagged changes in open interest between the period from t-1 to t and period from t-2 to t-1. We created the method CalculateChange to calculate the increment value and rank the contracts based on the weekly history data frame.

  def WeeklyTrade(self):
      hist = self.History(self.tickers, self.Time-timedelta(days=7), self.Time,Resolution.Daily)
      self.window.append(hist)
      if len(self.window) == self.window.maxlen:
          hist_t2 = self.window[0] # the weekly history dataframe two week ago
          hist_t1 = self.window[1] # the weekly history dataframe a week ago
          top_vol, bottom_vol = self.CalculateChange("volume", hist_t2, hist_t1)
          top_OI, bottom_OI = self.CalculateChange("open_interest", hist_t2, hist_t1)
  def CalculateChange(self, column_name, hist_t2, hist_t1):
      # calculate the weekly change and sort by the colume value
      value_t2 = hist_t2[column_name].unstack(level=0).sum(axis=0)
      value_t1 = hist_t1[column_name].unstack(level=0).sum(axis=0)
      delta = (value_t1 - value_t2).sort_values(ascending=True)
      top = list(delta[:int(self.length*0.5)].index)
      bottom = list(delta[-int(self.length*0.5):].index)
      return top, bottom

We take the intersection of the top volume group and bottom open interest group. The trading candidates are selected from this group. Next futures in the intersection are sorted based on the return in the previous week. The algorithm goes long on futures from the high-volume, low-open interest group with the lowest return and shorts the contract with the highest return.

# the intersection of top volume group and bottom open interest group
trade_group = list(set(top_vol) & set(bottom_OI))
returns = {}
for ticker in trade_group:
    res = (hist_t1.loc[ticker]["settle"][-1] - hist_t2.loc[ticker]["settle"][-1]) / hist_t2.loc[ticker]["settle"][-1]
    returns[ticker] = res

sortedByReturn = sorted(returns, key = lambda x: returns[x])

if len(sortedByReturn) >= 2:
    if self.Portfolio.Invested:
        self.Liquidate()
    self.SetHoldings(sortedByReturn[-1], -0.3)
    self.SetHoldings(sortedByReturn[0], 0.3)

Algorithm

You can also see our Documentation and Videos. You can also get in touch with us via Chat.

Did you find this page Helpful ?