Investigating Intermittent Differences Between RSI Calculations

Summary

The document describes a troubleshooting question about occasional differences between a custom 14-period RSI and a charting platform’s RSI. The calculations usually agree closely after their initial values converge, but sometimes diverge noticeably for a few minutes before matching again. The author reports using one-minute closing prices sourced from an exchange websocket and says the RSI smoothing method and period match the platform’s stated settings.

The text raises plausible areas to investigate, including whether both systems receive identical price data and construct minute closes in the same way. It does not include an answer, identify the cause, or provide evidence that the discrepancy is due to data quality, timing, or implementation. The initial-value difference is acknowledged, but because the divergence recurs after a long period of agreement, the document remains an open diagnostic question rather than a confirmed explanation or RSI trading method.

Key ideas

• Different initial RSI values can arise from different starting price histories.
• The reported RSI values usually converge but sometimes diverge briefly after extended agreement.
• The custom calculation uses one-minute closes received through an exchange websocket.
• The document poses possible data and timing differences but does not establish a cause.

Tags

# Why do I randomly see a significant difference in my RSI and that of Trading View


# Why do I randomly see a significant difference in my RSI and that of Trading View












Now I realise my own RSI values will never be exactly the same as those from Trading View (TV), as I don’t have the same data for the first 14 periods. However I would expect the two to converge, and they do. Most of the time the values only differ by about 0.1. This leads me to believe that there’s no issue with my algorithm.

Unfortunately sometimes, and quite randomly, the two will vary by up to 5. It will last few minutes and then they match up again.

For reference, my interval is 1 minute, with the price at the end of every second being interpreted as the closing price for the current (minute) interval. I’m using Periods=14 and Exponential Moving Average with alpha = 1/Periods in my algorithm. I’ve verified that TV does the same. I’m obtaining price data via my exchange’s websocket.

Question: Under what conditions could there be such a random divergence in RSI if they’ve remained the same for a long time (more than 2 hours)? For example, could it be a case of TV and me receiving different price data and could this cause such a sudden significant difference?

Shown in full with attribution under the source's licence. Licence: CC BY-SA 4.0 (Stack Exchange)

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.