When I saw your posts this morning I had a question about whether you are applying the EMA10 on the HA Close or on the standard bar close. Rather than ask the question I decided to repeat the spreadsheet exercise I did the other day and plotted the EMA10 vs the standard bar close and vs the HA bar close, for that 4 month sample of NU daily data that I used to plot the SMA5 & SMMA5 graphs - because I wanted to see the relative effect of each of the variations (I believed the SMM5 would have less whipsaws than the EMA10 whether using the EMA10 on standard or HA bars).
The results are as follows:
I have plotted the EMA10 vs the standard daily close (1st graph) and HA daily close (2nd graph):
I then tabulated the results in terms of pips return for each variation in that table I posted for the SMA5 & SMMA5 - see below:
So this table gives a direct comparison of the 4 methods (albeit the sample is small, only 4 months of daily data, it is randomly selected).
What we can see is that in terms of pips return, the order from best to worst is:
1. SMMA5 using HA High & Low
2. EMA10 using HA Close
3. SMA5 using HA High & Low
4. EMA10 using Standard Close
Note there is a big gap between 2nd & 3rd best, so EMA10 using HA Close is almost as good as using SMMA5 using HA High & Low, and both are much better than the 3rd & 4th methods.
Note also the number of trades each produces - the SMMA5 using HA High & Low produces sgnificantly less trades than the other methods, meaning it appears to keep you in the trade longer than the other methods, i.e. less prone to whipsaws.
Speaking of whipsaws, note in the bottom 2 tables the dates highlighted in red, which are trades entered on consecutive days - the EMA10 using Standard Close produces 3 consecutive day whipsaws (in one case 2 whipsaws in 3 consecutive days) so imo is the least desirable method.
Unfortunately the SMMA5 using HA High & Low is the most difficult to visualise (I haven't been able to find settings in any of my charting packages to display it), but based on the limted sample looks the most promsing in terms of results it can produce - I better get on to that revision of the EA for that method...
