nonetheless wrote:

@dnystrom I dont think you would want an actual crossover in both time frames as that would be a fairly rare occurrence and, as you say, result in v few trades. Better would be to use some kind of trend indicator in the daily TF, then insert the part of the code i posted above (MACross finder) in the shorter TF and let it find the best crossover to work in conjunction with your daily filter.

It’s possible of course to load it in both TFs and optimize them separately, but I wouldn’t expect it to work very well. (or maybe it would – try it and let us know!)

Thank you, yes i will try it later today.
I just need help it two things. How do i use Jans code with just the two averages that i want and its not do all the crossings each backtest? So I easily can change the averages on each timeframe, default and upper time frame?
And how can I code the two averages on the upper time frame so when that condition fills in, then look for the trade in the default time frame?

Hi dnystrom,

Very interesting idea to try this crossing strategy in Multiple Time Frames.

I am not sure if it much limits the trades using an optimum in both a long and short time frame, because there are many short time frames available for crossing when the long time frame crosses.

(Theoretically) it would be possible to do so, but it can not be tested at once (70 x 70 x 70 x 70 = 24.010.000 variations, way too much), and you would have then 4 x 70 averages, which is a long code.

Image you do this for proper testing on more then 100.000 bars (of the short time frame), I hope we do not get the RED FLAG from ProRealTime for overusing the server !

The trading concept would be: (if I understand the idea of dnystrom correctly)

if AvType1 (long time frame) crosses above AvType2 (long time frame)  and AvType3 (short time frame) crosses above AvType4 (short time frame), go long

Vice versa  if AvType1 (long time frame) crosses under AvType2 (long time frame)  and AvType3 (short time frame) crosses under AvType4 (short time frame), go short

I will work out a strategy like this and test it with a small range of variables  and let you know.  (Testing until the RED Flag will be given . . ), I can share the code.

You may develop it yourself,  copy first range of averages (0-69) to Notepath,  find/replace MATpe to MAType3 /4, find/replace Series to Series3 /4,  find/replace Period to Period3 /4,  find/replace AFR to AFR3 /4, and add back the new averages, started those with the smaller time frame, start the original 2 ranges with the higher timeframe, and extend the smoothed averages and the crossings in the bottom lines of the code, and see and test what it delivers), Do not forget to add also AFR3 and AFR4 to the variables to be tested above the code.

Thanks for the input.
Exactly the way I was intended to test Jan. But i was thinking of first do a test on lets say a 2 or 4H chart then a separately test on the 15 min chart, and then run a test with thoose 4 avg that gives the best output. But your way seems so much better =)

I will make some test later today with different kind of indicators for bull/bear on the upper TF then the optimisation on the lower.
I let you know how it goes.

Jan wrote:

Testing until the RED Flag will be given

I know it was humour! 🙂

PRT will have their own way of throttling / limiting our backtesting … you can tell, it’s like we get a second or so every minute sometimes?

When I’ve been backtesting max combo’s 10,000 over 100k bars it can seem like it has stopped and then every minute or so 10 or so more results pop into the table … even at the start when the table is empty then 10 to 20 results appear after a minute or so?

Anybody else seen / had this?

Isn’t it possible to copy out the code to a new “trading system” then search and replace, and then copy in that and so on for the 4 different numbers of avg?

I haven’t tested this but my instinct is that for 2 TFs, it might be better to use

in the longer TF, so that it’s more of a trend indicator than a crossover.

I agree with you, I have tested it and with a double change crossing (longer and shorter timeframes) very few trades are generated, so better use higher timeframes as a trend.