Long time lurker first time poster.
Been working with deep orderbook and trades analysis on crypto tokens (BTC & ETH). I am currently utilising EMA'S with a 5h decay as I feel OB and trade data is more relevant to short term price movements.
I have found that orderbook imbalance slope tends to have a decent correlation to price movement and trade spikes particularly aggressive (market order) trade spikes tend to indicate significant moves but I am struggling to capitalise on this algorithmically due to the noisy nature of the data I am processing.
Questions for this community:
1) Does anyone here have any suggestions for advanced data processing of noisy websocket feeds? I have tried Kalman filtering but it is still too noisy
2) Is orderbook and trade analysis a genuine edge that most people ignore because it is too difficult to extract the edge? If so I am patient and willing to do the grind necessary to extract this edge
3) Is orderbook and trades processing strictly limited to short term edge or is there long term potential and implementing a longer term EMA decay would fix my noise issue? If so simple problems have simple solutions.
Thanks in advance, any insight is greatly appreciated!
Posted by Guarado