Hi Ricardo,

Here is sample code for your strategy. I didn't include the condition when close > high of yesterday because this would be forward-bias. In the middle of the day today you cannot know what will be the stock's close price.

This strategy will exit after a 3% loss or a 5% profit. Meanwhile, it will keep buying 100 shares of the stock.

For your reference, there are also other ordering methods available, such as stop-order, limit order, and stop-limit orders.

Cheers,
Alisa

Hey Ricardo,

I made a few additions to Alisa's algorithm. I assume by close > high of yesterday then buy at next open 100 shares you mean that if today's close is greater than the high of yesterday buy 100 shares of NFLX at the open tomorrow. I implemented functions to check if the current time is at the open/close of the day. Let us know if you have any more questions!

Ryan

Thank you both but the codes above do not reflect my conditions. By the way, close > high of yesterday, means that the close of today is higher then the high of yesterday and it is not forward looking in any sense but I should have made that clear I guess. If I assume correctly that in highPrice.ix[0, context.nflx] > highPrice.ix[2, context.nflx] the former is the high of today and the latter the high of 2 days ago, then this is not related to my conditions.

Hello Ricardo,

highPrice = history(6,'1d', 'high')  

produces a DataFrame of results with a column for each security. The value of '6' periods specifies that 6 rows of results are returned i.e. for 5 days plus the current minute. Then

highPrice.ix[0, security]  

denotes the first/oldest value i.e. 5 days ago and

highPrice.ix[2, security]  

denotes the value for 3 days ago.

P.

Thank you for your answers but I wonder why is this here done opposite to the way the majority of other platforms do it with 0 representing the most recent value. I find it very confusing trying to convert my algos to this notation. Has the developer looked at how this is done the last 25 years by most trading platforms or it was deliberately done this way for purpose of differentiation? I mean essentially it does not matter but it would make life easier to maintain some consistency with other platforms.

Hello Ricardo,

I would have thought numbering from 0 (ignoring R and FORTRAN) for the first/oldest element of a time series was the norm? The only time I can recall 0 indexing the most recent element in a time series is MQL i.e. "Series array elements are indexed in the reverse order, i.e., from the last one to the first one. The current bar which is the last in the array is indexed as 0." See: http://docs.mql4.com/predefined/close

Where else have you seen the 0th element as the most recent?

P.

Tradestation, Amibroker, Ninja Trader, Meta Trader, Metastock and in a few more I now do not recall the names. An index of 0 for the most recent element is natural because if the lookback period changes, when for example when one wants to add a rule referencing elements beyond the current lookback period, the code need not change. The way the elements are indexed in Quantopian means that the indexing must completely change. Maybe this was done for some other good reason I cannot think of at the moment but it is not the way everyone else does it (or most of the platforms I have seen) and to me it appears like a limitation unless there is a good justification for it.

Hey Ricardo,

If you would prefer to have have the indices reversed, feel free to use the function I wrote here. I demonstrate it using price and volume historical data.

from pandas import DataFrame as df  
import numpy as np

def initialize(context):  
    context.securities = [sid(24), sid(2673), sid(5061)]

def handle_data(context, data):  
    prices = history(11, '1d', 'price').ix[:-1]  
    new_prices = reverse(context, prices)  
    print('New data frame of prices:' + '\n' +  '%s' % new_prices)  
    volumes = history(11, '1d', 'volume').ix[:-1]  
    new_volume = reverse(context, volumes)  
    print('New data frame of volume:' + '\n' +  '%s' % new_volume)

def reverse(context, prices):  
    symbols = []  
    for sec in context.securities:  
        symbols.append(sec.symbol)  
    df3 = df(np.zeros((len(prices), 3)), columns = symbols, index = prices.index.values[::-1])  
    df3.ix[:,symbols] = prices  
    for i in range(0,len(context.securities)):  
        df3.ix[:,i] = np.array(prices)[:,i][::-1]  
    return df3