Build Error: 47 Error Runtime exception: ValueError: cannot convert float NaN to integer

from pandas import rolling_std def initialize(context): context.secs = [ symbol('XLY'), # XLY Consumer Discrectionary SPDR Fund symbol('XLF'), # XLF Financial SPDR Fund symbol('XLK'), # XLK Technology SPDR Fund symbol('XLE'), # XLE Energy SPDR Fund symbol('XLV'), # XLV Health Care SPRD Fund symbol('XLI'), # XLI Industrial SPDR Fund symbol('XLP'), # XLP Consumer Staples SPDR Fund symbol('XLB'), # XLB Materials SPDR Fund symbol('XLU') ] # XLU Utilities SPRD Fund context.dev_wins = [15,20,25,30,35] context.reg_wins = [260] context.scaling = 0.0225 context.maxdw = max(context.dev_wins) def handle_data(context, data): returndata = history(bar_count=781, frequency='1d', field='price').pct_change().ix[1:] desiredpositions = {} for security in context.secs: rmse = {} for rw in context.reg_wins: rmse[rw] = rolling_std(returndata[security],rw) begin = -65 signals = {(rw,dw):0.0 for rw in context.reg_wins for dw in context.dev_wins} for dt in returndata[security].index[begin:]: for rw in context.reg_wins: resids = returndata[security].ix[:dt].ix[-rw:].mean() - returndata[security].ix[:dt].ix[-context.maxdw:] resids.index = list(range(context.maxdw)) stderr = rmse[rw][dt] for dw in context.dev_wins: signals[(rw,dw)] = resids.ix[-dw:].sum()/stderr desiredpositions[security] = sum(signals.values())/len(signals) * context.scaling log.info(desiredpositions.values()) for security in context.secs: order_target_percent(security, desiredpositions[security]) #This is the line that provokes the error # order_target_percent(symbol('XLF'), 1.0)

Hi Angelo,

What's happening is that when you run it for all the securities, there are some that may not have data reaching back the full 781 days. What history() does when it doesn't have data, is fill the missing dates with NaNs which is why you're currently getting the NaN error.

What I've done is test whether or not 'sum(signals.values())/len(signals) * context.scaling' is a NaN and convert it to 0 if it is

from pandas import rolling_std  
import numpy as np

def initialize(context):

    context.secs = [ symbol('XLY'),  # XLY Consumer Discrectionary SPDR Fund  
                     symbol('XLF'),  # XLF Financial SPDR Fund  
                     symbol('XLK'),  # XLK Technology SPDR Fund  
                     symbol('XLE'),  # XLE Energy SPDR Fund  
                     symbol('XLV'),  # XLV Health Care SPRD Fund  
                     symbol('XLI'),  # XLI Industrial SPDR Fund  
                     symbol('XLP'),  # XLP Consumer Staples SPDR Fund  
                     symbol('XLB'),  # XLB Materials SPDR Fund  
                     symbol('XLU') ] # XLU Utilities SPRD Fund

    context.dev_wins = [15,20,25,30,35]  
    context.reg_wins = [260]  
    context.scaling = 0.0225  
    context.maxdw = max(context.dev_wins)  
def handle_data(context, data):  
    returndata = history(bar_count=781, frequency='1d', field='price').pct_change().ix[1:]  
    desiredpositions = {}

    for security in context.secs:  
        if security in data:  
            rmse = {}  
            for rw in context.reg_wins:  
                rmse[rw] = rolling_std(returndata[security],rw)

            begin = -65

            signals = {(rw,dw):0.0 for rw in context.reg_wins for dw in context.dev_wins}

            for dt in returndata[security].index[begin:]:  
                for rw in context.reg_wins:  
                    resids = returndata[security].ix[:dt].ix[-rw:].mean() - returndata[security].ix[:dt].ix[-context.maxdw:]  
                    resids.index = list(range(context.maxdw))  
                    stderr = rmse[rw][dt]  
                    for dw in context.dev_wins:  
                        signals[(rw,dw)] = resids.ix[-dw:].sum()/stderr  
                desired = sum(signals.values())/len(signals) * context.scaling  
               #: Convert to 0 instead  
                if np.isfinite(desired) != True:  
                    desired = 0  
                desiredpositions[security] = desired

    log.info(desiredpositions.values())  
    for security in context.secs:  
        order_target_percent(security, desiredpositions[security]) #This is the line that provokes the error

    # order_target_percent(symbol('XLF'), 1.0)

I hope that helps and let me know if you have any questions on that

Seong