Algorithmic Trading Video Series #4 | Taking profit

Thanks for all the great educational content, Louis!

I send all my other quasi-quants / wannabe-quant friends to your videos.  Always great stuff.  

Thank you very much for the award and positive feedback!

Thanks for sharing your knowledge in these videos. Are you going to continue uploading more videos?

Yes, I will continue uploading videos. If everything goes well, this series could easily have 20+ videos. But since future videos are longer and include more coding and more advanced topics, they take a little longer to make.

Great! I will be attentive to the next videos of your YouTube channel.

Greetings!

Thank you. I was wanting to include something similar to this parameter actually earlier this week, and was thinking cleanest way of doing it. Thanks for doing my heavy lifting. 

Great work. Appreciate @ Louis 😊
Please keep adding more good content.

Yeap, Great Work and continue with your content! :)

Hi Louis, 
Thanks for the education and great content. I am curious if it is possible to delay execution to a specific time of day whilst still maintaining the rebalancing code: ”self.nextEntryTime = self.Time” and “self.nextEntryTime <= self.Time”? 
Ie if the algo rebalances on close but would prefer to trade at 9am or market on open the following day.
Thanks for your help! 

Thanks for the comment. Yes, you can specify a specific time to rebalance, however, note that this algorithm uses daily data, so to accomplish that, you would have to switch to a lower resolution. 

After that, an easy way to accomplish this, would be to used scheduled events and schedule a rebalancing function at your desired time.

I hope this helps.

you may come across below error:

Runtime Error: AttributeError : 'NoneType' object has no attribute 'Close'
at OnData
price = data[self.spy].Close

Solution:

    def OnData(self, data):
        if not self.spy in data:
            return
        if self.CurrentSlice.Bars.ContainsKey(self.spy):
            # price = data.Bars[self.spy].Close
            price = data[self.spy].Close
            
            # price = self.Securities[self.spy].Close
            
            if not self.Portfolio.Invested:
                if self.nextEntryTime <= self.Time:
                    self.SetHoldings(self.spy, 1)
                    # self.MarketOrder(self.spy, int(self.Portfolio.Cash / price) )
                    self.Log("BUY SPY @" + str(price))
                    self.entryPrice = price
            
            elif self.entryPrice * 1.1 < price or self.entryPrice * 0.90 > price:
                self.Liquidate()
                self.Log("SELL SPY @" + str(price))
                self.nextEntryTime = self.Time + self.period
                    

Hi Louis,

many thanks for consistently sharing great content - much appreciated!

Having looked at most of your tutorials, I noticed that sometimes (like in this example) you are using DataNormalizationMode.Raw.

I always use Adjusted to make sure to account for splits and dividends in historic data, so I've been wondering under what circumstances Raw would be needed.

I can see why it makes sense to look at an unadjusted price in your Take Profit example as you need to determine your profit. But would Adjusted even have the potential to change the price when it comes through as a data slice in OnData?

I would have thought that Adjusted only changes historical data to account for splits and dividends so that the history is consistent with the situation coming through in OnData. But maybe that's a misunderstanding?

I'm currently working on an ML algorithm that uses features derived off a few years of daily price data. Could Adjusted introduce the risk of look ahead bias here? i.e. can you see a reason why I should not be using Adjusted for this purpose?

All the best

Lars

Hi Louis, 

a fellow trader, and computer science enthusiast here. Love the series. I finished (except for the Tiingo lesson) the Bootcamp on QuantConnect and just started with your course. I love the way how you edited the video with clearly illustrated examples. Makes it for visual learners a lot easier to grasp the concept. I just finished lesson 4 Handling data but came across something weird when running the backtest. 

There is a difference in the backtest result between my equity curve, max. profit, PSR, etc, and the one in your video. I used the exact code, but somehow it yield a different result. It makes me worried about the reliability of the backtest engine itself. Do you know where this is coming from? I was thinking maybe it has to do with the LEAN engine version of the current master branch. Hope you can shine some light. 

Hope you can shine some light. 😊

Van Quoc Nguyen You are using 5% take profit - he is testing 10%

algo_dude Thanks for pointing this out.😊

If in backtest error message comes up “ 'QCAlgorithm' is not defined”,

add in the first line the following code:
 

from AlgorithmImports import *