Pathfinder swing TS
03/04/2017 at 11:39 AM #2712603/04/2017 at 1:13 PM #2713303/04/2017 at 2:00 PM #2713403/04/2017 at 2:04 PM #2713503/04/2017 at 4:34 PM #2715303/04/2017 at 5:15 PM #27157
I just want to be clear what the .itf is that you provide above? March2 = 0 and all variables in your Mar2 version above are the same as Gold Mar1 V1 but position size in yours above is 20, whereas in Mar1 V1 position size is 5.
Also the Mar1V1 file was provided by Kasper (ElsbourgTrading) so I expected Kasper would be providing the update for Mar2?
Maybe I am wrong and any of us can do ‘re-optimisations? Somehow though I see ‘anybody do anything’ as a recipe for confusion as we need to know who to get back to if we spot issues etc. Kaspers name is still in the file you provide above.
Please forgive and correct me if I’m wrong?
GraHal03/05/2017 at 8:19 AM #27176
Don’t know what you are looking at, but the information Bertrand provided is different from the Mar1 version from Kasper.
A couple of examples:
March1 March2 (Bertrand) ThirdMA 6 10 LongMA 7 39 SL 5 5 TP 5 3,75 MaxLong 9 8 WithoutP 9 14
My positionsize for March1 is 20. I can not remember that i’ve adjusted it. The minimum amount of contracts for Gold mini is 10.
Maybe that is different in the UK. I do not know. I can show you a picture if needed.
My experience is with PRT when you have a lot of BT running is that the right BT is not always loaded well in the system. When names
are pretty much a like, it looks like PRT sometime has problems with it. Sometimes i have to load three times the template to get it right.
Or delete the code and import it again and save the template again to get it right.
I’ve seen that you have running 50+ BT in PoOrder. I can not imagine that i am the only one with this problems.
By the way, i was not able to improve Bertrand’s results.
Lets wait and see when Kasper does his thing.
Patrick03/05/2017 at 10:26 AM #27183
Patrick, thank you, always good to have a 2nd pair of eyes … I’ll repeat my comparison exercise and report back.
It looks like I reduced the min position size from 20 to 5 on my DFB Spreadbet.
I must say, at first glance there are some big differences, esp that Long MA from 7 to 39; and that with only, maybe 2 or 3 weeks more history included in the optimisation (from when Kasper last optimised?).
Also there has only been 3 days trading history in the Mar1 period, so isn’t 4 March a bit early to re-optimise anyway? Unless I’ve got it all wrong how the re-optimisation of the Swings should be carried out?
We certainly are getting top the nitty gritty with these Swing Systems, but this is good as it leads to better understanding all round.
GraHal03/05/2017 at 1:12 PM #27213
Since English isn’t my first language I might have misunderstood you.
As I understand it, the optimization for the 2nd half of March doesn’t have anything to do with the 1st half, since the optimization is done on all the 2nd halves of March (in this case).
Therefore the optimization can be done in April or October or February and does not need to take the 1st half of March into consideration.
Anyone, please correct me if I’m wrong.
Best regards, David03/05/2017 at 1:13 PM #27214
Maybe i’m not understanding totally what you mean, but the history goes back for this example (Gold) to January 2nd. 1976 for Mar1 and for Mar2 as well.
The way i see it is that Mar1 and Mar2 are not comparable at all. Two total different situations because in Mar1 you have Mar2 at zero and in Mar2 it is the other
way around. From my point of view you can not compare these two situations. This means in Mar1 you have the history of all first half of the month from 02-01-1976
until now and in the optimisation of Mar2 you have the history of the 2nd. part of March from 02-01-1976 until now with Mar1 at zero. So that the outcome is completely different
is understandable. I may be wrong or misunderstood the problem. Please let me know if you think of it otherwise.
Patrick03/05/2017 at 1:14 PM #2721503/05/2017 at 3:59 PM #27242
Thanks guys, so from what you say I conclude …
Mar2 could have been optimised anytime after the Mar2 period in 2016 and be ready to go for Mar2 in 2017?
Also the only periods that are included / optimised for Mar2 2017 are all the 2 week periods (16 March – 31 March) from 2 Jan 1976 until 31 Mar 2016? So that’s 40 years x 2 weeks = 80 weeks of data?
Apologies for my questions, but I feel like I am understanding the ‘swing optimisation process’ better now?
As there is only ‘1 time period’ that can be selected in the Optimisation engine, how do you arrange for the Opto engine to give you 40 x 2 weeks worth of data?
My last question above makes me think 40 years x 2 weeks = 80 weeks of data backtested (that I proposed) can’t be right??
Aha, is it that the periods with a 0 (only Mar2 = 1) are not backtested? Me thinking aloud here, so bear with me please 🙂 More like … all 40 years x 12 months are backtested, but because no trades are done in the ‘0 periods’ then variables remain unchanged (no trades = no optimisation) for 0 periods??
GraHal03/05/2017 at 5:21 PM #27252
Yes indead. It could have been tested at any moment after March ’16 for March ’17.
Second question is also right. It only includes the periode where you have a figure for Mar2. The rest of the periods are put to zero, so do not count for Mar2. You optimize after you have done all other variables
the months, in this case Mar2. I give it letter “a”, but you can name it what you want, and you define “a” with 0 to 3 with step 1. In the calculation you have roughly 40 x 2 weeks of data. The history provides you the data.
This means you have 24 periods in a year. When you optimize Mar2, you set every other 23 periods at zero. Mar2 you do like described above or as like the link i attached. Your last thinking out loud describes it….
If a chart has 2 years of data, you can only optimize it with that data. That’s why you need as much as data as available because it provides you a better average. For example one year data is no relyable BT because
too less data. But data from 1976 till now gives you more relyable information because that is 40 x 2 weeks of data.
You optimize it according to Reiners instructions:
Reed it a couple of times. It needs some practice 🙂
No apologies necessary. It is difficult enough.
Patrick03/05/2017 at 9:34 PM #2726803/06/2017 at 1:39 AM #27273