QuantConnect Community

Hi Miguel,

The reason for this is because this PCM requires Insight magnitude, thus the default Alpha Models in this algorithm will need to be copied and tweaked to include magnitude as an argument. I've added magnitude as an argument to the custom Alpha this can be seen in the attached backtest.

Furthermore, this backtest shows that when using Alphas added through AddAlpha and SetAlpha, only the Alpha Model added from SetAlpha will be used, thus I'd suggest using AddAlpha with multiple Alpha models.I'd also suggest looking at the EMA Universe on how to avoid using History calls daily which can greatly slow down the backtest.

If you are still having issues, please update with a simplified backtest that isolates the issue.

Best,
Shile Wen

Hi Shile Wen,

Thank you for your help! 

For your history design suggestion, would it be similar to the first code snippet below? 

Additionally, I've run into an error where the algorithm doesn't seem to trade? I updated the resolution to minute resolution and added magnitudes to the alphas. I don't have too much experience creating alphas. Are these the correct modifications to the standard alpha models? It doesn't seem like the magnitudes are able to be manually set? 

Thank you in advance,

Miguel

	if cf.Symbol not in self.holdsec:

        self.holdsec[cf.Symbol] = SymbolData(cf.Symbol)

    

    # Updates the SymbolData object with current EOD price

    upsec = self.holdsec[cf.Symbol]

    upsec.update(cf.EndTime, cf.AdjustedPrice)

   

Hi Miguel,

Yes, that code snippet would be similiar.

The algorithm above doesn't trade because the insights need to be given a `magnitude`. To add a magnitude, we just need to replace

insight = Insight.Price(sd.Security.Symbol, self.insightPeriod, direction)

with

insight = Insight.Price(sd.Security.Symbol, self.insightPeriod, direction, magnitude=0.2)

Additionally, there is currently a bug in the python implementation of the BlackLittermanOptimizationPortfolioConstructionModel so we need to comment out the line that reads

from Portfolio.BlackLittermanOptimizationPortfolioConstructionModel import BlackLittermanOptimizationPortfolioConstructionModel

By commenting out the line, the C# version of the model is used by default. We've created a GitHub Issue to have the problem with the python version of the model resolved. Subscribe to our progress here.

See the attached backtest for reference.

Best,
Derek Melchin