One of the main problems that I have encountered is that Brokers do not provide very much data to backtest with on the MT4 trading platform. I decided to attempt to compare historic data for GBPUSD from other data providers to my Broker’s data.
After the data has been adjusted by 6 hours to get it into the same time as the Broker data, we can still see from the following graph that small differences exist between the two sets of data.
It is also important to make adjustments for Daylight Savings, which can have a significant impact on the data. This can be seen in the following Graph.
Finally, it is also important to check the records to ensure they are within the Brokers normal trading hours. Data outside of this trading range can be seen in the following image.
Once all of these adjustments have been made to the data and it is compared to the initial Broker Data over a time period of many years, the following result was obtained:
To be thorough, I also decided to run a similar procedure on another data source provider. Unfortunately that provider was missing lots of Sunday trading information and the following result was obtained:
With respect to the previous results, I believe the following points are valid conclusions:
- Forex Data is not perfect and there can be differences between different forex providers. I suspect the reason for this is the variation in the Brokers spread, but there may also be other reasons. This makes me worry about the path dependency of a strategy.
- It may be worth excluding the Christmas and New Year period when backtesting because data can be missing.
- It may also be worth excluding the Brexit time period from any backtest as this is an extremely unusual event and the data may be extremely inaccurate.