QuantConnect Community

Without seeing the code I am guessing you're calling self.History(), you need to access the algorithm parameter which is passed into the selection model. i.e. algorithm.History. The self refers to the current class; if you're in the history API is accessible on the algorithm object.

Hi Anthony FJ Garner ,

On the 'AttributeError' issue, we can see that it was fixed after you changed self.History for algorithm.History (line 30).

fine is a list of FineFundamental objects where each item has the AdjustedPrice attribute. If you loop through fine or sortedByDEratio, you should be able to access it:

Hi Anthony FJ Garner ,

Sorry about the misinformation.

Please find in the attached backtest a working version of what you are trying to accomplish.

If you have further questions, do not hesitate to ask. :-)

Thanks for the exampleAlexandre Catarino - this is helpful for me as well.

Alexandre

I have been going through your kindly provided code and believe I have spotted an error.  The following lines are an unnecessary duplication:

for symbol, mom in averages.items():
c = self.coarse.pop(symbol, None)
mom.Update(c.EndTime, c.AdjustedPrice)

In fact, by adding an extra and duplicate line to averages, an incorrect momentum is calculated and passed on to the next stage of the Framework.

The dataframe of prices (created from the call to History) for each stock already includes the lates price and is called each time that the Course/Fine function is called.

Incidentally, as regards Momentum, there is no need to calculate a momentum figure for each day included in history - all we are interested is the latest momentum. It saves a lot of preocessing power and time NOT to have to loop through each line of history. While it does no actual harm of course. Nonethless, perhaps the example code should stand "as is" since for some indicators, it is necessary to calculate for each day - a moving average for instance.

Sorry to be long winded but I thought this might be useful for anyone following.

Thank you, Anthony FJ Garner .

In fact, more generally, we would need to check the last (current) indicator value to see its EndTime property before we try to update:

Alexandre

Under what circumstances would:

 c.EndTime 

bear a later date than the close in:

df = history[symbol].dropna() ?

The close in df is what is used to update the indicator and it seems odd if the last row of df woul bear a different date to that in c.EndTime?

Hi Anthony FJ Garner ,

In some applications, we don't need to make historical data requests for all symbols since the difference between the previous request and the current one is just one data point. So, instead of making an expensive operation that includes the data in CoarseFundamental object, we just use it.

In your algorithm particular case, because of the yearly selection, I think it is better to just request all the data.