I’m trying to come up with non-price based datasets to be input into a genetic trading algorithm. I’m really interested in testing data for randomness, and representing information through graphs/Markov models.
The first dataset I think might be interesting is the flow of volume between pairs. Specifically I’m treating a trade where the taker exchanges 1 BTC to USD as -1 for the BTC node and +the equivalent for the USD node. That way over a period of time both the long USD and short USD for each crypto-USD pair can be recorded, as well as the node total.
Does this actually seem like it is a reasonable way to think about things, or do you see some glaring problem?
Submitted October 03, 2020 at 12:45PM by QuixoticValentine