Why RSI Values Change with Data Frequency

Summary

The document asks why a 14-period RSI calculated from daily closes differs from an hourly RSI over the same calendar span, even when the hourly period is scaled to 98 bars. It compares roughly 230 daily observations with about 1,500 hourly observations for AAPL from February to October 2019 and reports that the hourly RSI has a narrower range and otherwise looks substantially different.

The central lesson is that RSI is calculated from changes between consecutive observations, so changing the sampling frequency changes the price movements entering the calculation. Matching the number of calendar days covered by the lookback does not make the underlying inputs equivalent. The document poses, but does not answer, whether a formula adjustment could make the two indicators resemble one another. It offers no test of the charts or proposed correction, so it is a question about indicator behavior rather than a validated method for converting RSI across frequencies.

Key ideas

• RSI calculated from hourly closes can differ substantially from RSI calculated from daily closes over the same calendar period.
• Scaling the hourly lookback by the number of hours does not ensure that the resulting indicator has a similar range or shape.
• RSI depends on the sequence of price changes at the chosen sampling frequency.
• The document raises the question of adjusting RSI across frequencies but provides no answer or tested correction.

Tags

# Hourly RSI calculation to match Daily RSI


# Hourly RSI calculation to match Daily RSI












I implemented the RSI calculation based on the ta4j implementation that you can find here: https://github.com/ta4j/ta4j/blob/master/ta4j-core/src/main/java/org/ta4j/core/indicators/RSIIndicator.java.

If I use hourly data instead of daily data, for the same period, I end up having very different indicators.

For example:

Calculating the RSI using daily close values for 1 year ( around 230 points ) Vs Calculating the same RSI on the same period but using Hourly data ( around 1500 points )

It will give me very different RSI values, especially the range between max and min RSI drops significantly when using more granularity.

I would like to know if this is expected, and if yes, what correction to the RSI indicator I should use to have both RSI indicator looking similar independently of the granularity used.

For example, SMA indicators look similar if I use daily data or minute data.

As an example, here are the AAPL stock RSI 14 days with the daily close from Feb->Oct 2019. https://www.tradingview.com/chart/AAPL/wPE6FWmX-RSI-chart-for-Stack-Exchange-question/

And the same AAPL with hourly close, with an RSI 14*7=98 period in the same period.

https://www.tradingview.com/chart/AAPL/Y5KrNSfP-The-RSI-based-on-hourly-ticks/

As we can see, these two RSI look very different.

What modification to the original RSI formula needs to be made for these two look similar?

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.