op-analytics
op-analytics copied to clipboard
Check Lead/Lag values for Net Flows to Resolve Bad Data
Currently, net flows can record hugely bad information (i.e. 58T in flows), due to bad token-level data, either in usd price, or number of tokens.
Build an intermediate model with some logic that says "if the lead and lag (up to 2 away) is > 20x different than the current value, null this row due to bad data"
Example below, we know the middle row is incorrect: