Extreme outliers in aggregated candle data


#1

Thanks for this fix for the aggregated candles! The plottet charts look way better now!
However, for some currencies, there still exist some extreme wicks. Some examples:

PIVX:
https://api.nomics.com/v1/candles?interval=1d&currency=PIVX&key=2018-09-demo-dont-deploy-b69315e440beb145

Zcash:
https://api.nomics.com/v1/candles?interval=1d&currency=ZEC&key=2018-09-demo-dont-deploy-b69315e440beb145

ETH:


https://api.nomics.com/v1/candles?interval=1d&currency=ETH&key=2018-09-demo-dont-deploy-b69315e440beb145

Would be great if you could look into that - thanks in advance!


#3

Thanks!

Yes, this is due to extreme outliers, generally on markets with thin books where trades can be matched in extreme ways to make the price jump. When aggregated, these still affect a volume weighted average. We’re working on an outlier detection system that will prevent those kinds of trades from being used in our candles.