Trading Strategies Based On Digital Filters - page 18
You are missing trading opportunities:
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
Registration
Log in
You agree to website policy and terms of use
If you do not have an account, please register
Robertinno,
Which tool are you using for the spectral analysis? It looks different from DFM.
Simba,
For a summery of what I've learned so far, P1 and D1 are not the periods. What are they exactly? When we do a spectral analysis, I was thinking that the peaks correspond to the period of the most significant cycles.
I was thinking that the coefficients used to compute SATL, for example, are the n coefficients for a function of degree n approximating the cycle we want to filter.
And about RSTL, how should we compute it then? Was it right what was posted? The standard RSTL looks really different.
Robertinno,
Which tool are you using for the spectral analysis? It looks different from DFM.
Simba,
For a summery of what I've learned so far, P1 and D1 are not the periods. What are they exactly? When we do a spectral analysis, I was thinking that the peaks correspond to the period of the most significant cycles.
I was thinking that the coefficients used to compute SATL, for example, are the n coefficients for a function of degree n approximating the cycle we want to filter.
And about RSTL, how should we compute it then? Was it right what was posted? The standard RSTL looks really different.Dvarrin,I have tried to explain about P1 ,D1 to you several times,with no success,so,probably another member with better communication capabilities ,or better understanding of digital filters than me
,can help you,I can only repeat what I said in 2 previous posts...no need to do it,you can reread them.
The RSTL looks different from the standard because it is based on a specific SATL..every RSTL is f(SATL)
>>Which tool are you using for the spectral analysis? It looks different from DFM.
Finware analyzer
Dvarrin,I have tried to explain about P1 ,D1 to you several times,with no success,so,probably another member with better communication capabilities ,or better understanding of digital filters than me
Hi Simba,
I'm sorry, that's only that I'm not sure if I understood everything correctly. P1 and D1 are used to filter some frequencies and let the others pass through. But does it means that when we do a spectral analysis and we get a peak at 115 as in your example, then we have to put it as P1 and the number of coefficients used in the indicator is the cycle's period for the currency?
And for RSTL, I've tried to take one half more coefficients than for SATL, but the result is very different from the standard RSTL I could find and you did not tell me what is the correct RSTL and how to create it yet :-( Like I did, RST is only offset compared to SATL, but the standard RSTL is not at all like that :-(:-(
Dvarrin,
Here are the steps that i have been using with my currency pairs. This is basically a culmination of what i have learned/read, including tips from Simba. I use this same process for every pair, and have created a strategy that incorporates the indicators and seems to be very successful thus far( of course....much more testing is needed).
1. Markets are fractal in nature. The analysis you do with a 4 H pair should be very close to that of a 15 min pair. Now whether this is full true or not i don't know. This is just how i do it.
2. Download 4H data into the analyzer. KEY - Try to not use more than 250-300 records. The least amount is the best. If you do an analysis with 2000+ records, no "true" peaks will come out, and i'm afraid you will have a nightmare on your hands. Again, this is just how i do it.
3. After running the analysis, just guess and check ( try different combos) of SATL's and FATL's like Simba walked through with you. Visually seeing how the lines interact with price is the best way to go about deciding which to go with.
4. Key - At least this is key for me. I do not want SATL's or FATL's that have more than 20 coefficients. I want quicker indicators that don't bog your system down. I have 2 Gig of RAM and a duel processor, but you would be surprised how choppy a 100+ coefficient indicator can make your charts.
5. So lets say you make an SATL that has 14 coefficients. 14 is your period. From what i learned from Simba, the # of coefficients is your period. P and D do not refer to periods. If you access the help file inside the DF software, you will see little graphs of how signals are processed in Low pass, high pass, band pass, and band stop filters. You can see what P and D mean from those charts. So, with 14 coefficients, i want an RSTL that is roughly 1.5 times the period of SATL, give or take 10%. If you have 40 coefficients, you will have to be delaying your indicator by 18+ or so, and the validity of the signals will no longer render any reliable information to you. This is very important, especially when you are trying to trade these.
6. Once you have RSTL, STLM is simply SATL - RSTL. I've explained to you ( please refer to that post) how i just use a STLM template and replace the coefficients that are generated by the DF software with them. Don't worry about comparing what you are doing to the "standards" in terms of numbers. Just know how many periods you are looking for, and how to get there. :-)
I hope this helps.....i'm not the authority on this by any means, but i feel like my "understanding" (above) makes my process repetitive, which makes my analysis on any currency pair consistent.
Simba....would you mind if we move on and start hashing out RBCI like we did with the histogram indicators....and maybe trading strategies ( i have one i'd like to share). Not looking for a handout :-) I know you won't give them anyways lol.
Best,
cl
Dvarrin,
Here are the steps that i have been using with my currency pairs. This is basically a culmination of what i have learned/read, including tips from Simba. I use this same process for every pair, and have created a strategy that incorporates the indicators and seems to be very successful thus far( of course....much more testing is needed).
1. Markets are fractal in nature. The analysis you do with a 4 H pair should be very close to that of a 15 min pair. Now whether this is full true or not i don't know. This is just how i do it.
2. Download 4H data into the analyzer. KEY - Try to not use more than 250-300 records. The least amount is the best. If you do an analysis with 2000+ records, no "true" peaks will come out, and i'm afraid you will have a nightmare on your hands. Again, this is just how i do it.
3. After running the analysis, just guess and check ( try different combos) of SATL's and FATL's like Simba walked through with you. Visually seeing how the lines interact with price is the best way to go about deciding which to go with.
4. Key - At least this is key for me. I do not want SATL's or FATL's that have more than 20 coefficients. I want quicker indicators that don't bog your system down. I have 2 Gig of RAM and a duel processor, but you would be surprised how choppy a 100+ coefficient indicator can make your charts.
5. So lets say you make an SATL that has 14 coefficients. 14 is your period. From what i learned from Simba, the # of coefficients is your period. P and D do not refer to periods. If you access the help file inside the DF software, you will see little graphs of how signals are processed in Low pass, high pass, band pass, and band stop filters. You can see what P and D mean from those charts. So, with 14 coefficients, i want an RSTL that is roughly 1.5 times the period of SATL, give or take 10%. If you have 40 coefficients, you will have to be delaying your indicator by 18+ or so, and the validity of the signals will no longer render any reliable information to you. This is very important, especially when you are trying to trade these.
6. Once you have RSTL, STLM is simply SATL - RSTL. I've explained to you ( please refer to that post) how i just use a STLM template and replace the coefficients that are generated by the DF software with them. Don't worry about comparing what you are doing to the "standards" in terms of numbers. Just know how many periods you are looking for, and how to get there. :-)
I hope this helps.....i'm not the authority on this by any means, but i feel like my "understanding" (above) makes my process repetitive, which makes my analysis on any currency pair consistent.
Simba....would you mind if we move on and start hashing out RBCI like we did with the histogram indicators....and maybe trading strategies ( i have one i'd like to share). Not looking for a handout :-) I know you won't give them anyways lol.
Best,
cldvarrin:1-I believe that clahn04 explained it much better than I could.His steps are exactly how I do it,and how I tried to explain by asking questions and giving tips,probably less clearly than him
, a few posts ago when I posted the spectrum analyzer pic,for the last 211 periods(you can choose anything from 200 to 250 periods),with the biggest peak at 115.
2-When then I calculated a custom SATL based on P1=115 and D1=14..what I did was:FIRST: to take all cycles contained between those periods,excluding anything else below or above,since everything else was not relevant.SECOND:By setting P1=115,I set it to the highest amplitude peak,so,the filter(please read the info dvarrin suggested you to read)accepts this cycle in full,and then it filters down to D1=14,accepting the other cycles ONLY PARTIALLY,so it gives full preeminence to the strongest cycle and partial importance(visualize a decreasing slope from 115 to 14) to all other selected cycles..you are creating a composite,and the end result is a SATL with 16 coefficients applied to periods from actual to minus 15,both included,so,the end result is that with a SATL based on 16 periods,theoretically,you are modelling a cyclic composite from 14 to 115 periods..in practice,you check visually and decide if you like it or not..if you like it,to do the rstl,you just change the delay of the SATL from 0 to 8(in practice you can change to 7/9 too,allow 10% room,and check which one you like more).THIRD:I suggest that you try to do a new SATL with P1=115 and D1=nearest peak(I don`t remember the pic,I think it was 80 or something),then another from P1=115 and D1=the third next peak to 115..then compare all satls,including the one I posted..basically,if I am not mistaken ,the one from 115 to 80(or whatever),will have the most coefficients,will be smoother..and much less responsive..then ,you chose the one you want..and,from it,you calculate the rstl by just delaying the chosen satl by half its coefficients.
clahn04:Well,this is exactly how I do it,regarding trading strategies and RBCI,I would suggest to wait for dvarrin,if he is setup,we can proceed with the strategies,and we can even create a simple EA to test them,though,we will have to take care with testing,since digital filters,specially stlm2 and rbci,with several coefficients and double buffers,as you noted,are very heavy on memory use...BTW,no problem to give you the strategies,they are very dependent on the actual composition of the SATL,for example for the 16 periods satl and 22/24 periods rstl,the best strategy would be to use the slope of the rstl(I tested it..I made the EA ,but have it at my other pc, 250 km from where I am now,no problem,will redo)..and I think that it is even better if I give you a way to test all possible profitable strategies,and then decide based on facts
Sounds great. Dvarrin, please let us know if that response helped. I want you to be up to speed and comfortable before we start talking about other things.
cl
SIMBA:
Do you prepare quotations to analysis? do you analysis for all period?
SIMBA:
Do you prepare quotations to analysis? do you analysis for all period?
Robertinho,
1-No,when working with digital filters I don`t prepare quotations for analysis.I tried once to apply a standard SATL to a very short Fatl(smoothed real time representations of price),and the only thing I achieved was the cpu jumping at 98 degrees Celsius,just before the system stopped working.
2-I do analysis for D1,H4,H1,M30,M15..anything above and below ,I consider too different from my trading style.Additionally,I have found too that markets are fractal,so,usually a very good filter at H4 works exceptionally well for all below H4 timeframes too..and,sometimes,a very good filter at m30 works exceptionally well for H4,H1 and M15 too.
3-One important thing I do when doing digital filters is trying to find 2 harmonic cycles in 2 harmonic timeframes,i.e:the representation of the same cycle in both timeframes..M30 you have a dominant period of 32..then at M15 you have a dominant period of 64..this cycle is going to be,first:very useful at most timeframes,second:very long living(high bartels).
Robertinho,
1-No,when working with digital filters I don`t prepare quotations for analysis.I tried once to apply a standard SATL to a very short Fatl(smoothed real time representations of price),and the only thing I achieved was the cpu jumping at 98 degrees Celsius,just before the system stopped working.
2-I do analysis for D1,H4,H1,M30,M15..anything above and below ,I consider too different from my trading style.Additionally,I have found too that markets are fractal,so,usually a very good filter at H4 works exceptionally well for all below H4 timeframes too..and,sometimes,a very good filter at m30 works exceptionally well for H4,H1 and M15 too.
3-One important thing I do when doing digital filters is trying to find 2 harmonic cycles in 2 harmonic timeframes,i.e:the representation of the same cycle in both timeframes..M30 you have a dominant period of 32..then at M15 you have a dominant period of 64..this cycle is going to be,first:very useful at most timeframes,second:very long living(high bartels).1.
>>No,when working with digital filters I don`t prepare quotations for analysis.
attachment eur d1 1 pic.- 2000 bar 2pic - 1000 bar
>>I tried once to apply a standard SATL to a very short Fatl(smoothed real time representations of price),and the only thing I achieved was the cpu jumping at 98 degrees Celsius,just before the system stopped working.
Do you use fatl with dll ???
>>3-One important thing I do when doing digital filters is trying to find 2 harmonic cycles in 2 harmonic timeframes,i.e:the representation of the same cycle in both timeframes..M30 you have a dominant period of 32..then at M15 you have a dominant period of 64..this cycle is going to be,first:very useful at most timeframes,second:very long living(high bartels).
Did you try use for analysis non-standard timeframe? D3, H9 etc. ? do you analysis alpari quotations or other broker?