Slow/Smooth Exponential Moving Average [SEMA}

2020년 6월 12일

An almost impossibly smooth moving average that is convenient for any application where a moving average is needed and where noise reduction is more important than responsiveness to very recent data. You get to choose what the S stands for.

Built by taking a weighted difference between two EMA's in such a way that the contribution of the latest value is cancelled out, while staying a proper moving average with all weights greater than zero, and with the first moment picked to be the same for a given length as the sma. Still has a somewhat ema-like behaviour in the sense that old values are gradually forgotten over time.

2020년 6월 16일

릴리즈 노트

Made slight mathematical improvement: since the last valued is canceled out of the moving average it is possible to tune the formula to shift the moving average forward by one candle. This makes it slightly more responsive to recent movements without sacrificing noise reduction.

The old behavior can be regained by simply delaying the new version.