Portfolio Value Vecomes NAN

mcandar

I am trying to implement a strategy that longs selected n stocks from a universe. In a time interval, I have more than 200 stocks with the format below. However, in this time interval, some of the stocks are listed after the beginning date and also some are unlisted before the end date so there are nan's. (Delisted column shows if that stock is going to be delisted after 20 days.)

                open      high       low     close  delisted
2005-01-03  0.211401  0.258859  0.200831  0.215716       0.0
2005-01-04  0.211401  0.258859  0.200831  0.215716       0.0
2005-01-05  0.227417  0.278469  0.216046  0.232058       0.0
2005-01-06  0.224214  0.274547  0.213003  0.228789       0.0
2005-01-07  0.233823  0.286313  0.222132  0.238595       0.0
2005-01-10  0.240229  0.294158  0.228217  0.245131       0.0
.........

I am trying to calculate a metric only for not nan stocks currently and the ones that won't be nan during incoming 20 days.

    def my_metric(self):
        t = self.datas[0].datetime.date(0)
        cons = list(self.p.cons.loc[t].values)
        asset_datas = [d for d in self.datas if d._name in cons]

        mom = []
        for d in asset_datas:
            dlst = d.delisted.get(size=22)
            dlst = np.asarray(dlst)
            if np.any(dlst > 0): # if it is going to be unlisted in 20 days
                self.log(f'Skipping delisted stock {d._name}')
                continue

            rt = d.close[0]
            rt_1 = d.close[-self.p.t1*self.p.per]
            rt_12 = d.close[-self.p.t12*self.p.per]

            short_term = (rt - rt_1)/rt_1
            long_term = (rt - rt_12)/rt_12

            result = long_term - short_term
            if np.isnan(result):
                continue

            mom.append((result,d))
        return sorted(mom,reverse=True) # sort descending

I monitor important values inside next()

    def next(self):
        if len(self) < self.p.t12*self.p.per:
            return
        
        month = self.datas[0].datetime.date(0).month

        if month != self.last_month:
            self.rebalance()
            self.last_month = month

        self.posdata = [d for d, pos in self.getpositions().items() if pos]
        
        totalwealth = self.broker.getvalue()
        cash = self.broker.getcash()
        invested = totalwealth - cash

        self.log("Total Wealth: %.2f, Invested: %.2f, Cash-On-Hand: %.2f" %(totalwealth,
                                                                    invested,
                                                                    cash))
        if np.isnan(totalwealth):
            exit()

        self.log(f"IN HAND ---> {[(d._name,round(d.open[0],2),round(d.high[0],2),round(d.low[0],2),round(d.close[0],2),int(d.delisted[0])) for d in self.posdata]}")

It goes well up to a point at which the portfolio value turns out to be nan.

2020-05-20, Total Wealth: 1620177.67, Invested: 1533365.52, Cash-On-Hand: 86812.15
2020-05-20, IN HAND ---> [('KAREL', 15.93, 19.51, 15.14, 16.26, 0), ('GUBRF', 18.46, 22.61, 17.54, 18.84, 0), ('HEKTS', 13.67, 16.74, 12.99, 13.95, 0), ('RYGYO', 4.73, 5.8, 4.5, 4.83, 0), ('GUSGR', 3.24, 3.97, 3.08, 3.31, 0)]
2020-05-21, Total Wealth: 1639291.91, Invested: 1552479.76, Cash-On-Hand: 86812.15
2020-05-21, IN HAND ---> [('KAREL', 15.56, 19.06, 14.78, 15.88, 0), ('GUBRF', 19.99, 24.48, 18.99, 20.4, 0), ('HEKTS', 13.58, 16.63, 12.9, 13.86, 0), ('RYGYO', 4.81, 5.89, 4.57, 4.91, 0), ('GUSGR', 3.19, 3.91, 3.04, 3.26, 0)]
2020-05-22, Total Wealth: 1657420.90, Invested: 1570608.75, Cash-On-Hand: 86812.15
2020-05-22, IN HAND ---> [('KAREL', 15.67, 19.19, 14.89, 15.99, 0), ('GUBRF', 19.46, 23.83, 18.49, 19.86, 0), ('HEKTS', 14.7, 18.0, 13.96, 15.0, 0), ('RYGYO', 4.81, 5.89, 4.57, 4.91, 0), ('GUSGR', 3.21, 3.94, 3.05, 3.28, 0)]
2020-05-27, Total Wealth: 1686695.71, Invested: 1599883.56, Cash-On-Hand: 86812.15
2020-05-27, IN HAND ---> [('KAREL', 16.32, 19.98, 15.5, 16.65, 0), ('GUBRF', 19.57, 23.96, 18.59, 19.97, 0), ('HEKTS', 14.9, 18.24, 14.15, 15.2, 0), ('RYGYO', 4.9, 6.0, 4.66, 5.0, 0), ('GUSGR', 3.27, 4.01, 3.11, 3.34, 0)]
2020-05-28, Total Wealth: 1669631.77, Invested: 1582819.62, Cash-On-Hand: 86812.15
2020-05-28, IN HAND ---> [('KAREL', 15.11, 18.5, 14.36, 15.42, 0), ('GUBRF', 20.99, 25.7, 19.94, 21.42, 0), ('HEKTS', 14.51, 17.77, 13.79, 14.81, 0), ('RYGYO', 4.76, 5.83, 4.52, 4.86, 0), ('GUSGR', 3.22, 3.95, 3.06, 3.29, 0)]
2020-05-29, Total Wealth: 1692349.23, Invested: 1605537.08, Cash-On-Hand: 86812.15
2020-05-29, IN HAND ---> [('KAREL', 14.95, 18.31, 14.21, 15.26, 0), ('GUBRF', 22.13, 27.1, 21.02, 22.58, 0), ('HEKTS', 14.5, 17.76, 13.78, 14.8, 0), ('RYGYO', 4.72, 5.78, 4.49, 4.82, 0), ('GUSGR', 3.29, 4.03, 3.13, 3.36, 0)]
2020-06-01, Leave KAREL
2020-06-01, Rebal GUBRF
2020-06-01, Rebal HEKTS
2020-06-01, Rebal RYGYO
2020-06-01, Rebal GUSGR
2020-06-01, Enter DEVA
2020-06-01, Total Wealth: nan, Invested: nan, Cash-On-Hand: 86812.15

I rebalance once in a month and check values everyday. In the logs I don't see any nan value for the stocks I have in hand, but the portfolio value is nan. What is that I am missing here?

mcandar

@mcandar Any clues how to avoid this? What is the reason that causes NaN values?

ab_trader

I would recommend the follwoing tools -

or another debugging tools (if available) in your IDE. Can be quicker, than waiting for an answer here.