BLUESHIFT Backtesting engine is not reliable

Hi all!



I would like to point out that I think the blueshift backtesting system is not good.

I have attached a picture of the basic RSI strategy:

https://drive.google.com/file/d/1p-ebeQvSItqB15U5c5iXsot90x_AIYFD/view?usp=sharing



This is the strategy:

https://github.com/QuantInsti/blueshift-demo-strategies/blob/master/forex/fx_demo_rsi.py



The backtest gives amazing results.

I did not touch the basic parameters.

Leverage: 30, Order quantity: 1000, Asset: EURUSD, Freq: 1 minute, Starting capital: 1000 USD



I have got ROI 1500% for 1 month!

How?

Drawdown% is almost 0%!

How?

Sharpe is 13.11!



When I deployed the strategy to the FXCM demo account, it made $ -70 in a matter of hours.

Paper trading and backtesting have nothing to do with each other at the performance level.



I think the backtesting engine is NOT reliable!

 

Well, as mentioned to you before (check email from blueshift-support@quantinsti.com), the default cost modelling for FX is not great. We will move to a new bid-ask data based default modeling in a newer version. For now you need to manually set the correct cost based on what you actually see in paper trading account. Note the pip-cost that is actually charged and plug that in. I am pretty certain the resulting strategy will not have a Sharpe of 13.