As we all know it well, GDX is a relatively volatile instrument. What we have noticed while performing the RT back-tests is that the intraday volatility of the GDX Money flow is very important. The average intraday volatility is about 0.5. This is represented by the pink arrow in the figure below.
The GDX MF sometimes indeed fluctuates several times during the same trading day around a Buy/Sell level, even when taking into account wider porosity levels. This means that the GDX RT system is victim to a high number of whipsaws.
Because of this high intraday volatility, when an intraday signal is issued, we cannot be confident that this signal will stay valid by the end of the day. On the other hand, if we wait 16:00, we can only act the following day and back-tests have shown that this would considerably hurt profitability.
Therefore, we tested the GDX RT system, but we took only the trades when the signal was still valid 20 minutes before the close.
You can see below the comparative results between trades taken at the open the day after a signal is issued or trades taken only when a signal is valid 20 minutes before the close.
When separating by signal type, we can see that the standard "Buy" trades (which occur when we cross above the 0 level) produce good returns when the signal is traded 20 minutes before the close instead than at the next open. This means that many algos are detecting changes and that the next day's opening gap already factors in what was detected on the previous night.
Conclusions:
You should act on a trade only if you are confident that it will stay valid until the close. In order to help you, we are going to add in the coming days a "confidence level" field that will show the probability that the signal will stay by the close of the day.
Note: Since the "Buy Oversold" signals are the most profitable, we hope to be able to come with a special alert system that will, independently from the model, detect intraday reversals of deep oversold situations.
Reply With Quote
Originally Posted by DJones
