Back testing news! Why has no one done it?

New comment
 

So I've been designing this EA, but I can not seem to get ANYONE who can get backtested News data that I can use in a EA. 

The goal here would be to somehow get a .txt file of all the news data and proiorty levels back to 2012 or so, to current day. Then use that to alter trading in a backtest on MT4/5. So far I have found NOBODY who has or can make this. Anyone have any ideas? 

If you DO know how to make this, please let me know! Thanks! Will definitely pay you

Testing trading strategies on real ticks
Testing trading strategies on real ticks
  • www.mql5.com
The article provides the results of testing a simple trading strategy in three modes: "1 minute OHLC", "Every tick" and "Every tick based on real ticks" using actual historical data.
 

Please don't use ALLCAPS for sentences or titles. It is considered shouting and rude.


The reason for the difficulty of testing news in back tests, is that it is not easy to find a reliable source of free historical data of economic news events. I believe FXStreet allows viewing older economic events and downloading it, but I have never used it myself. I am unsure of this, so do some research.

You would also need a daylight savings database so as to adjust the time-zone and DST to align with the calender events.

In all, it is not easy to run such a back-test.

 
The only time I saw someone do this was with Python and the goal was to exclude news events from their backtest results but even for this it was so much work that I'm not sure they're still using this method. Even if you download tick data with the correct timezone (adjusted to your DST), you'd have to create a database of all news events and manually adjust the database according to the same DST. And even if you managed to do this, you're never going to be able to know how much slippage you'll experience and it can vary a lot during news events so your backtest is almost 100% sure to be unreliable. On top of that, backtests aren't usually reliable until you get around to 5,000+ uncorrelated trades (10k+ is much better) so if you're only trading news I don't think you'll ever be able to backtest on a large enough sample size to extract any statistical significance from it, even if trading multiple symbols. What makes this even worse is the fact that trading news means you'd be trading intraday and typically profitable intraday algos are extremely difficult to create, if not maybe impossible (some people might have different opinions on this though).
New comment