search

nkaz001 / algotrading-example

algorithmic trading backtest and optimization examples using order book imbalances. (bitcoin, cryptocurrency, bitmex, binance futures, market making)
283 stars 95 forks source link

data source #2

Closed jeanfabrice94121 closed 3 years ago

jeanfabrice94121 commented 3 years ago

Hi,

I was just trying to replicate your scripts crypto-scratch whose make use of data from tardis. I was able to download the sample data from this data provider but it appears to be not enough data to perform the entire analysis (in-out of sample periods).

I was wondering which data you are reading with the following command (e.g. xbtusd-normalization-market-making-backtest-hedge.ipynb ) :

df = pd.read_pickle('data2')

imbalace or trades? That you saved in a pickle object in the notebook : tardis-process.ipynb

Did you subscribe to tardis for additional data? what subscription plan would do the job (retrieving enough data to run your notebooks).

Any advice/recommendation is welcome. Thank you in advance. BTW nice job! Jean-Fabrice

nkaz001 commented 3 years ago

That's normalized orderbook imbalances. Please see process_L2_normalization in tardis-process.ipynb.

In order to build the data, you need a full L2 order book(the most granular one) stream and a trade stream.

When I bought the data from tardis first there was a cheaper option, but now it seems that Solo Individual Exchange billed yearly is the cheapest option to access the full historical data they provided. I think you might need to talk with tardis for more detail.

Also, I got additional data for the early period from another source(an individual).

As the data cost gets increased, collecting the data by yourself would become a good option again.

I think data is one of the most difficult barriers for algotrading, but still getting crypto data is much cheaper and easier than others such as Nikkei, HSI, ES, and etc.

jeanfabrice94121 commented 3 years ago

Thanks a lot for your answer. Regards, JF