Historical Data consistency

FTMO Trader Scouting


New Member
Hey everyone,

I'm reading Kevin Davey's book Building Winning Algorithmic Trading Systems from the recommended books list and I have a question based on the following quote from the author:

Let’s say you are trading gold, and you want to use 20 years (excellent choice!) of data, with 60‐minute bars. Twenty years ago, the pit was the only data source, so you have to use that. In a daily pit session, there were probably six to eight bars roughly (since pit trading hours over the years changed, the number of 60‐minute bars per day will change, too). For your strategy, let’s say you use a 14‐period moving average. This will typically represent two trading days. Now, fast‐forward to today’s electronic markets. Today’s markets trade for roughly 23 hours per day. If you still use a 14‐period moving average, that will only equate to a half trading day, instead of the previous two trading days. Do you think that can radically influence your historical tests? It sure can!

I'm wondering how you deal with this. I'm currently using NinjaTrader 8 for my entire stack, so I use Continuum as my data provider. I'm sure every provider could be different, but I'm really asking about how this issue can be dealt with in general. For example, how would you go about investigating a data provider to see if this problem exists and how would you fix it? How would you even know where to check if the problem exists?

I feel out of my depth here so I'm eager to learn. Thanks a lot for any knowledge you can share!
FTMO Trader Scouting