Hi all,
I was wondering whether has anybody had any experience testing the Donchian or Modified Turtle system on weekly data?
I have been searching the forum for quite some time and I can't seem to find a guide on testing systems on the weekly time frame?
Could somebody point a newbie such as myself in the right direction? Thanks so much....
Donchian or Modified Turtle System on Weekly Time Frame
Hi James,
If "testing Donchian system on weekly data"
you wish to convert 20 days to 4 weeks, 50 days to 10 weeks, that can be accomplished by changing either your data feed setting or using weekly H, L, C setting with strategy code. Your results will be somewhat different, but do not expect major change.
If what you really have in mind is to test a 20 week or a 50 week price channel, you have a completely different set of issues. The number of entry opportunities for a 20 WEEK price channel system are far far fewer then for a 20 day system. From a methodology perspective, this becomes a problem. You will not have sufficient number of trades to make valid assignment of the model's performance. Even if you trade it on faith, several years, yes years, of out of sample draw down might NOT imply model failure, but simply indeterminacy stemming from lack of test data.
As a novice, better use of your time would be to investigate trading models which can generate a minimum of 10 round turns per market per year. Ten round turns is not much, 500 trades for 50 market in ten years--barely adequate to make an assessment.
L
If "testing Donchian system on weekly data"
you wish to convert 20 days to 4 weeks, 50 days to 10 weeks, that can be accomplished by changing either your data feed setting or using weekly H, L, C setting with strategy code. Your results will be somewhat different, but do not expect major change.
If what you really have in mind is to test a 20 week or a 50 week price channel, you have a completely different set of issues. The number of entry opportunities for a 20 WEEK price channel system are far far fewer then for a 20 day system. From a methodology perspective, this becomes a problem. You will not have sufficient number of trades to make valid assignment of the model's performance. Even if you trade it on faith, several years, yes years, of out of sample draw down might NOT imply model failure, but simply indeterminacy stemming from lack of test data.
As a novice, better use of your time would be to investigate trading models which can generate a minimum of 10 round turns per market per year. Ten round turns is not much, 500 trades for 50 market in ten years--barely adequate to make an assessment.
L
- generate a minimum of 10 round turns per market per year. Ten round turns is not much, 500 trades for 50 market in ten years--barely adequate to make an assessment.
- 50 markets
- 10 years
- 10 trades/market/year
Solution: 50 x 10 x 10 = 5000 trades
