Practical Checks for Assessing OHLC Data Quality

Summary

The document considers how to assess the quality of open, high, low, and close price data used in strategy backtests. It lists checks for outlier or negative prices, internally inconsistent bars such as a high below a low, and missing minute observations. The response says there is probably no universal standard for grading such data, so these checks are a useful starting point rather than a complete scoring framework.

A practical comparison method is to obtain data from multiple sources and compare a sample across a selected period or instrument set. Sampling can reveal the kinds of discrepancies present without requiring a full comparison of every observation. The answer also points to a reference on cleaning high-frequency foreign-exchange tick data. It does not define thresholds for outliers, prescribe a single quality score, or explain how to adjudicate conflicting vendors, so the checks and comparisons need to be tailored to the data and intended use.

Key ideas

• OHLC validation can check for outliers, negative prices, impossible high-low relationships, and missing observations.
• The response does not identify a universal standard for grading OHLC data quality.
• Comparing a sample from multiple data sources can help reveal the types of errors present.
• The sample need not cover every period or instrument to provide an initial diagnostic.
• Thresholds and resolution of vendor disagreements are left to the user.

Tags

# Is there a standard / methodology to determine and grade the quality of OHLC data?


# Is there a standard / methodology to determine and grade the quality of OHLC data?












Inputs are most important to any decision making. For strategy backtesting, the OHLC data is one of the most inputs.

So to ensure the correctness and integrity of OHLC data, we have checks on the following OHLC issues:

- Outlier Price

- Negative Price

- Corrupted Price (High < Low, etc)

- Missing minute ticks in 1-min timeframe

However, there is a lacking of a standard or methodology to grade the data quality. So is there any standard / methodology to determine and grade the quality of the OHLC data?

## Answer by NPE (score 4)

https://quant.stackexchange.com/a/1417

I think your list is pretty good, and I doubt there are any "standards" in this space.

What I tend to do is source the data from multiple places, and compare. This doesn't have to be for the entire period/universe; a sample is often sufficient to figure out what types of problems I am dealing with.

Lastly, there is a chapter in "An Introduction to High-Frequency Finance" devoted to data cleaning (they look at FX tick data).

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.