Estimating Intrabar RSI Signals from OHLC Data

Summary

The question asks whether research on signals such as RSI should use tick data, since the indicator can fluctuate within a candle and closing-price calculations may miss when a threshold was reached. Tick data can make a long historical sample much larger, so the practical issue is how to study intrabar behavior without retaining every trade or quote.

The answer suggests using OHLC data to calculate RSI under price paths based on the candle’s high and low, treating those values as boundary conditions for a distribution of possible indicator readings. That distribution can then support testing. This is a proposed approximation, not evidence from a reported comparison showing that candle and tick results agree. The document does not specify how to model the within-candle path, estimate the distribution, or account for sequencing and market microstructure, so the approach’s accuracy depends on the research question and assumptions.

Key ideas

• Tick-level RSI can vary within a candle, while close-only calculations may miss intrabar threshold events.
• OHLC highs and lows can be used to estimate boundary conditions for a distribution of possible RSI values.
• The proposed distribution can be used for testing without storing the full tick history.
• The answer does not provide empirical validation or specify the assumptions needed to construct the distribution.

Tags

# Is there a noticeable difference in making scatter plots and regression models with tick-data or with candle data?


# Is there a noticeable difference in making scatter plots and regression models with tick-data or with candle data?












I am asking this question because I want to research some variables. An example is the RSI where the current RSI is updated every tick. This means that the value of the RSI is fluctuating a lot inside an individual candle. Lets say I want to research what happens after the RSI reaches a value. But if I only take the close prices and not the tick-prices I don't get exactly what I want but a value that is close to the real value (assumption).

A reason for me not to work with tick-data is that is makes the data set very big. Imagine that I want to analyse 4 years of price data using tick-data. That is an immense amount of rows.

Is there someone who tested the close price and tick-data and found a major difference?

## Answer by babelproofreader (score 1, accepted)

https://quant.stackexchange.com/a/57026

You don't necessarily need to use tick data to accomplish what you want. If you have OHLC data you can just calculate RSI values using the extremes of the H and L values to get the boundary conditions of a density distribution and then use this distribution to do your testing.

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.