Seconds crypto data in FinRL_MS

[artiko88]

Hi!

First of, very excited with the whole FinRL community and movement. Quite a lot of work and all open source. Thank you guys!

I was thinking what should be changed in FinRL Market Simulator in order to work with orderbook data that is updated every second? I am designing a market making strategy that sends orders every second and uses information from orderbook. Currently i am backtesting it in hftbacktest and i was using regulrar q-learning from insinde of hftbacktest which is not very effective in terms of my q-learning algo learns. So i was thinking abut using FinRL and Fin_RL_Market_Simulator seems like the closest thing.

I can derive necessary data as in the stocks.csv example, but the granularity will be seconds, not minutes. From what i briefly observed in env.py, i will have to change a bit the preprocessing step so that it doesn't delete indecies that are outside of trading hours since crypto assets are traded 24 hours a day.

Am i safe to assume that after these modifications i can use tFinRL as an environment in the FinRL three layer model?

My main task is to test the algo that i have in my head in a market simulator with an RL algo. This means supplying it with features i think are important and make it choose among actions that i think it should make, like place ask/bid orders simultaneously at different volumes and different prices. Please, advise, which steps should be taken when FinRL is modified (a bit) to work with seconds crypto data?