Thought I'd be productive and give everyone a heads up for those looking for utilising eodhistoricaldata as their historical data source for minute intervals.
I decided to trial their paid ‘All World Extended’ service for Asia Pacific stocks as they offer data from 2021-05-17 onwards. The results were... interesting. They are an official vendor for ASX EOD data so I gave the benefit of the doubt that their minute level data would have a reasonable level of integrity behind it. Outcome - they are from different sources and currently unreliable as of 11 April 2023.
To do the analysis I exported the compressed minute level data [Link] (Intraday Historical Data API — EOD) from Amibroker into excel and compared the resulting end of day data to their straight EOD api Link Here. This should produce the same result as EOD is ultimately compiled from the intraday results. I then also compared this to the free Yahoo historical EOD data.
For the year 2022 the results are as follows:
Yahoo vs EODHistorical EOD data matched 100% of the time, with some minor discrepancies ~2% of cases for the high and low where yahoo rounds values to 2 decimal places.
The outcome for the compressed intraday data vs EOD end of day showed a completely different story. The open values matched only 8%, the High 73%, the low 82% and close 100%. Digging into the data, it quickly became apparent that EODhistorical intraday lacked the opening minutes of the day (first 5-10 minutes), thus causing the erroneous results in the the subsequent high and lows. Since the close bar was always present the close price matched with true accuracy.
When contacting EOD for a response, the general tone is that their intraday API for non-US stocks is "under development" and not documented i.e. we will say the data is available, but not document that it will be incomplete or in "testing mode" Furthermore, because you can't test the non-US API call you will have no way of knowing this since it isn't documented, made known it's in pilor mode or even available to screen... If that's not false advertising... they said it best:
There's no real need to technically limit the access to this type of data, if it's available, even if incomplete.
Ultimately, they refuse to issue a full refund at this stage and are claiming to have this 'issue' fixed. At this stage I can't recommend them as a data source, nor can I support their handling of the issues raised. There is clear deception whether knowingly or unknowingly and given their own datasets don't align, there is minimal validation for data integrity on their end. Use with extreme caution.