---
title: "Comment by OneLifeVibe on \"RSI in Python: Wilder’s Formula and Live Candle Updates\" | Hashnode"
description: "The worked five-period example is what makes this land — seeing the seed as a plain average of the first five gains and losses, then watching 75.00 drop to 64.8"
url: "https://hashnode.com/posts/rsi-in-python-wilder-s-formula-and-live-candle-updates/6af32998f910c8b38f0b389e/comment/6af34afa3b8023106a83078b"
---

[Discussion](https://hashnode.com/posts/rsi-in-python-wilder-s-formula-and-live-candle-updates/6af32998f910c8b38f0b389e)

Comment [RSI in Python: Wilder’s Formula and Live Candle Updates](https://hashnode.com/posts/rsi-in-python-wilder-s-formula-and-live-candle-updates/6af32998f910c8b38f0b389e)

[OneLifeVibe](https://hashnode.com/@onelifevibe)

[Nov 10](https://hashnode.com/posts/rsi-in-python-wilder-s-formula-and-live-candle-updates/6af32998f910c8b38f0b389e/comment/6af34afa3b8023106a83078b)

The worked five-period example is what makes this land — seeing the seed as a plain average of the first five gains and losses, then watching 75.00 drop to 64.8649 on a single close of 103, shows exactly how much history Wilder's smoothing carries. The preview versus update split is the part I would have wanted spelled out when first wiring up a live feed: treating each tick as a candidate against the same committed state, and only committing the final close, is a clean way to stop a forming candle from being counted several times. I also appreciated the pandas note — that ewm gets the decay right but still seeds from the first observation, which explains a lot of those small, stubborn mismatches between two charts showing the same RSI(14).
