I would try vectorbt, but... I really like to write strategies by my own =} and the reason I would try vectorbt is to see examples how they implemented all this with numba... One work-based history: I had script that loops over 4800 pairs of futures and trades its spread forward walking and after 1.5 days only 130 pairs were done... Okey, I decided to speed it up and wasted 3 days to rewrite code, get rid of every single pandas object, even writing some numpy functions equivalents (like np.all(ndarray, axis=1) because additional parameters like axis are not implemented) and even type errors like stupid one when I typed that whenever I don't have money for one/both tickers - then real deposits for both tickers is np.arr([0, 0]) but it should be float type like np.arr([0., 0.]) Or declare dtype... 3 days of hell, but numba gave me 130 pairs in 10 minutes! And it's with compilation time ~40 seconds. So I would like to see how they solved problems I got in my experience.
Great video as always thanks so much! Would you know of any backtesting package that deals with Bid Ask Data as separate data source (handy for entry and exit that takes into account Bid-Ask spread) ?
Are you able to do Walk Forward Optimization with either framework?
VectorBT looks more advanced ! We need more tutorials/examples for crypto data , Do you know how to use pro version?
Great channel bro. Thanks for the vids!
I would try vectorbt, but... I really like to write strategies by my own =} and the reason I would try vectorbt is to see examples how they implemented all this with numba...
One work-based history: I had script that loops over 4800 pairs of futures and trades its spread forward walking and after 1.5 days only 130 pairs were done... Okey, I decided to speed it up and wasted 3 days to rewrite code, get rid of every single pandas object, even writing some numpy functions equivalents (like np.all(ndarray, axis=1) because additional parameters like axis are not implemented) and even type errors like stupid one when I typed that whenever I don't have money for one/both tickers - then real deposits for both tickers is np.arr([0, 0]) but it should be float type like np.arr([0., 0.]) Or declare dtype... 3 days of hell, but numba gave me 130 pairs in 10 minutes! And it's with compilation time ~40 seconds.
So I would like to see how they solved problems I got in my experience.
Great video as always thanks so much!
Would you know of any backtesting package that deals with Bid Ask Data as separate data source (handy for entry and exit that takes into account Bid-Ask spread) ?
Vectorbt PRO will do this for you if you ask nicely
Have you used Blankly, also a new Framework?
I hadn't noticed that one! Will try it out!
you were spot on
Thanks for your good job.
Can we get reasons if python backtesting is better than MQL5 and pinescript?
Pinescript has forward bias
Look ahead bias
MQL5 is best if you are going to run entirely on mt5 .
Sir, which backtesting framework allows to put trailing stoploss ?
It's possible with both of them. Also possible with backtesting.py
@@ChadThackray forgot to add - sir :) to replay
how to pay with ETH?
Anderson Barbara Anderson Daniel Martin Sharon