Deep dive analysis and fee model

Hi Dat En, 

I can help answer your first question. You can try creating a data dictionary to store your values and develop a form of analysis within your algorithm. In terms of plotting, consider plotting using hour bars instead of minute, it requires much less memory. From my experience, a scatter plot with buy and sell prices coupled with a line chart of my indicator was sufficient. I've also used the tagging feature to tag orders with relevant information to make sure my algorithm was working correctly. 

There are also other ways to conduct analysis using the research notebooks. If you are interested in statistically modeling your alphas, take a look at Quantopians Alphalens and the lecture regarding factor analysis. Please let me know if this helps or if you have any questions :) 

Cheers

Thanks Jovad for the suggestions. However in my case I still need to plot at higher resolution and be able to zoom in to periods of interest. I will have look at Alphalens.

Hi Dat,

1. There are no logging limits if LEAN is run locally
2. To see the Fxcm fee model details, I will link to the implementation here. According to this link, this is the fee structure (as of on Oct 6th, 2015):

• Forex:
• $0.04 per side per 1k lot for EURUSD, GBPUSD, USDJPY, USDCHF, AUDUSD, EURJPY, GBPJPY
• $0.06 per side per 1k lot for other instruments
• CFD: no commissions

To set custom fee models, we first need to define a custom fee model class. Here is one example pulled from the documentation:

class CustomFeeModel:
def GetOrderFee(self, parameters):
fee = max(1, parameters.Security.Price
* parameters.Order.AbsoluteQuantity
* 0.00001)
return OrderFee(CashAmount(fee, 'USD'))

Then, we need to set our security fee models (here we use EURUSD) to this model using the following:

Please see the attached backtest for reference.

Best,
Shile Wen