You said you're calling history() in your initialize() function, but that would usually generate a TypeError rather than a KeyError. If you post a stripped-down backtest with the code that generates the error, it would be easier to diagnose the problem.

In Python, and in object oriented programming in general, we view objects as having state and behavior. In this case the object is a Pandas Series and the behaviour is the ability to calculate pct_change on itself. Here are the Docs for .pct_change() What you were trying to do was pass the object as a parameter for a method designed for itself if that makes any sense. For example, to get the pct_change of a single security from history you can use...

percent_changes = histories[security].pct_change()  

where histories is a data frame object containing the historical information and security is a Quantopian Security object as the key.

Does that help a little?

Also as @Michael said, using the history() function in initialize() will give you a TypeError.

You were both correct. Calling it on initialize () was a mistake. I changed the code the way @James suggested and now it works. The explanation was spot on. Thanks guys!

I am getting the same error, but I am not running history in initialize, and I am calculating percent change manually.

KeyError: Security(48073, symbol='CRC', security_name='CALIFORNIA RESOURCES CORP', exchange='NEW YORK STOCK EXCHANGE',
start_date=Timestamp('2014-11-13 00:00:00+0000', tz='UTC'),
end_date=Timestamp('2015-06-23 00:00:00+0000', tz='UTC'),
first_traded=None) There was a runtime error on line 230.

def initialize(context):

    # Select a group of stocks based on the (price * average_volume).  
    set_universe(universe.DollarVolumeUniverse(  
            floor_percentile=90.0, ceiling_percentile=94.9)

    # Submit new trades at market close  
    schedule_function(HandleEntry, date_rules.every_day(),  
                          time_rules.market_close(minutes=10))

def HandleEntry(context, data):

    # Make a list of all available stocks with data  
    eligible_stocks = [stock for stock in data if 'price' and 'volume' in data[stock]]

    # Get the last X days of prices for every stock in our universe.  
    prices = history(50, '1d', 'price').dropna(axis=1)

    # Calculate the 1-day return for all stocks remove any NaNs if there are missing prices  
    returns = ((prices.iloc[-1] - prices.iloc[-2]) / prices.iloc[-2]) * 100  
    returns = returns.dropna()

    MinPctChangeLong = 3.0  
    MinPctChangeShort = -3.0

    # Remove stocks below minimum change requirements  
    for stock in eligible_stocks:  
        if MinPctChangeShort < returns[stock] < MinPctChangeLong:  # <---- This is the line with the error  
            eligible_stocks.remove(stock)  

Any ideas what is happening?

Well if the stock is in eligible_stocks and got dropped from returns because it had a NaN value that would be one possible reason.

James - Good idea! Do you think this will solve that condition? (Updated with new code)

    # Calculate the 1-day return for all stocks remove any NaNs if there are missing prices  
    returns = ((prices.iloc[-1] - prices.iloc[-2]) / prices.iloc[-2]) * 100  
    returns = returns.dropna()

    # Only include stocks that have returns  
    eligible_stocks = [stock for stock in eligible_stocks if stock in returns]  

Honestly Tristan I would get rid of the eligible_stocks variable altogether. Basically you spend a line of code at the top iterating through all the securities and filtering them, and then you iterate over most of them again later to execute your order logic. I would do it with just one iteration like so:

for security in data:  
    if security in returns:  
        # execute logic  

There maybe a more Pythonic way to check the two collections with set / intersect... Hope that works!