Oil 15 minutes meanreverting strategy

Viewing 15 posts - 16 through 30 (of 49 total)

← 1 2 3 4 →

Author

Posts
10/06/2017 at 9:13 AM #48394 quote

Yngve
Participant

Senior

I have had alot of issues when running this code. The system have stoped several times, and because of that I missed winning trades. When I have restarted the system again the system have taken a different direction than the system is actually in. Have anyone other than me experiancede the same?

10/09/2017 at 8:25 PM #48811 quote

ozz87
Participant

Senior

Tried to optimize this one for AU. Would say average results. Will not use it myself, since it requires abit to much to run for me (min 10 contracts on Swedish IG).

Andyswede thanked this post

Meanrevert-AU-15-min.itf meanrevert_au.png meanrevert_au_wf.png

10/09/2017 at 9:47 PM #48819 quote

Andyswede
Participant

Master

@ozz87 It´s possible to buy 1 contract Avista Gold mini 10 oz with Swedish IG

10/09/2017 at 9:56 PM #48821 quote

ozz87
Participant

Senior

Andyswede: Hm. When I check contract information it says minimum 10.

asdasd.png

10/09/2017 at 9:57 PM #48824 quote

ozz87
Participant

Senior

@andyswede: Ah, I saw now there is another contract. Cool 🙂

10/09/2017 at 10:36 PM #48828 quote

Despair
Blocked

Master

A little bit longer backtest.

ozz87 thanked this post

Bildschirmfoto-2017-10-09-um-23.35.13.png

10/09/2017 at 10:42 PM #48832 quote

Abz
Participant

Veteran

Despair, is crude oil also available for 200000 candles in premium?

10/09/2017 at 11:07 PM #48833 quote

Despair
Blocked

Master

No it isn’t. There is infuriating little history on oil. That’s all we get for 15min.

Bildschirmfoto-2017-10-10-um-00.05.49.png

10/10/2017 at 6:19 AM #48847 quote

Yngve
Participant

Senior

I have been running the alog for a about a month now and the results are not good. The algo stops alot and I miss winning trades because of it.

Las time it stoped I got the following message

“This trading system was stopped due to a division by zero during the evaluation of the last candlestick. Please add protection to your code to prevent division by zero”

I have not been able to resolve the issue yet, but I will keep looking. Let me know if any of you gus have any suggestions to how to fix it.

10/10/2017 at 7:29 AM #48854 quote

jebus89
Participant

Master

@Yngve, I’m getting the same error, and yes, the past month has not showed great results. It might ofc be from overfitting or it might just be a random slope in the algos profits. Guess we gotta keep this bad boy on demo until errors go down and profits go up 🙂

10/10/2017 at 8:39 AM #48869 quote

stockdemon
Participant

Junior

There’s a division by zero error which I’ve solved by using the following code:

if abs(Dhigh(0)-Dlow(0)) = 0 then
fullness = (Dclose(0)-Dopen(0))/0.000001
else
fullness = (Dclose(0)-Dopen(0))/abs(Dhigh(0)-Dlow(0))
endif

You could also probably just skip the evaluation when you’ve got zero division. I think the above works as well.

10/10/2017 at 10:07 AM #48883 quote

Francesco78
Participant

Master

As I posted many times, this is the solution

fullness = abs((Dclose(0)-Dopen(0))/(Dhigh(0)-Dlow(0)+0.00001)

10/11/2017 at 9:12 AM #49006 quote

Yngve
Participant

Senior

@Francesco78 Thank you for the reply. I tried to change the line of code your referring to but got even worse results. So I changed it to

fullness = (Dclose(0)-Dopen(0))/abs(Dhigh(0)-Dlow(0)+0.00001)

What do you think about that?

@jebus89 Yes I’ll be running it Demo from now on for awhile

10/11/2017 at 7:23 PM #49060 quote

stockdemon
Participant

Junior

Francesco78 wrote:

As I posted many times, this is the solution fullness = abs((Dclose(0)-Dopen(0))/(Dhigh(0)-Dlow(0)+0.00001)

It’s a solution, I prefer something which can’t yield division by zero. I don’t think it’s likely to create any weird results either.

I bet your solution works too with the underlying primitives, altough without the exact knowledge of how they work it still looks like possibly erroneous code.

10/11/2017 at 7:47 PM #49061 quote

Francesco78
Participant

Master

stockdemon wrote:

Francesco78 wrote:

As I posted many times, this is the solution fullness = abs((Dclose(0)-Dopen(0))/(Dhigh(0)-Dlow(0)+0.00001)

It’s a solution, I prefer something which can’t yield division by zero. I don’t think it’s likely to create any weird results either. I bet your solution works too with the underlying primitives, altough without the exact knowledge of how they work it still looks like possibly erroneous code.

abs((Dclose(0)-Dopen(0))/(Dhigh(0)-Dlow(0)+0.00001) cant yield to a division by zero as High -Low is bounded to zero, so high +low + 0,0000001 is strictly positive

Nicolas and Yngve thanked this post
Author

Posts