I’ve been live trading for a couple of years and in that time i’ve tried numerous strategies until late last year when i finally developed my own system. Due to my statistics and mathematics backgroung and job experience, I developed a prediction model that eseentailly uses the last 3 years data to predict how the market will move (considering trending and ranging factors as well).
I’ve backtested the model on 30M, 1HR and 4HR timeframes as well and the results are very encouranging as it has managed to correctly predict the right decision on more than 50% of the time(ranges between 50% to 60%). The prediction model is designed in R programming language and it seems to work best on GBPJPY. After R gives the prediction, i execute the buy or sell order in 4 hour timelines with proper money management in my platform(using Ctrader at the moment).
Ive been trading it on the 4HR chart with a stop loss of 20pips and take profit figure of 45pips and based on my sucess ratio of 55% per month, the results are very encouraging. I chose those 2 figures(stop loss and take profit) based on average pip movt of the GBPJPY for the last 2 years on the 4HR bars (details in attached excel sheet).
I need someone who can help me backtest it on 5 year to 10 year data for 15 mins timeframe as my machine isnt capable of doing it due to processing speed requierements.
I’ve attached the R script for back testing and the one for live trading as well as a statisctical analysis sheet that displays the back tested data.
Any details on how to fine tune the model are welcome.
Any very fast machine can work. Preferably a virtual machine that has access to cloud computing i think can crack the code. I backtest 3 years data with 1500 records on my 7the Gen i7 and it takes roughly 6 hours to complete and this is only for the 1hr. The 15mins data will have roughly 50,000 records for 5 years if you only use 12 hours in a day. That will just fry my laptop probably
ok, … hehe
can you QUANTIFY… "MIN. FAST MACHINE"
can you give me the Mainboard Specs, Memory Size and Frequency, and Processor speed that you deem to be
MINIMUM FAST MACHINE
also
again…
MIN FAST VIRTUAL MACHINE SETTINGS
Not all VM’s are created equal are they, not even by default
Now… here is an answer i can work with
ok… it would also be useful to know how much memory you have and what frequency the memory is running at.
also it seems like TIME is an issue for you here.
so let me ask this
IN WHAT TIME FRAME WOULD YOU NEED THE COMPUTER TO COMPLETE THE TASK IN ?
@_bob
No No No… i think this is a justified comment.
i mean if the guy is talking about an i7 CPU and it’s still not good enough for him,
either … his computer is not optimized , or has data collissions and runs slow, or the memory frequency is low but high end cpu
or… HE IS LOOKING FOR SOMEONE WITH AN i9 CPU hehe… Good luck with that .
now the other way around it is to go for a server board with a Multi CPU Board
but i still don’t know what min specs he’s after, and until i know that, how can anyone help him.
Ideally, backtesting using indicators is really very simple and fast but this strategy is running phase reconstruction in R which involves reconstrucing the data into new matrices and then running non linear time series on the data. Hence the reason the machine is taking forever to complete what looks like a simple baxktesting task.
As per mimimum requirements needed, i think an I7,7000U processor with 16GB ram can crush this code in a few hours or less.
