Machine Learning in ProOrder ProRealTime

juanj · Sun, 06 Aug 2017 18:29:04 +0000

I would like to start a discussion around the possibilities of making use of machine learning in ProRealTime. I have personally experimented with some heuristics algorithms in ProOrder but find it very limiting when having to optimize more than one variable. Here is an snippet of the code I use to optimize an variable called periods (i.e. Average[periods]): // Heuristics Algorithm once Periods = 42 //Variable to be optimized once Optimize = 0 once Mode = 1 //switches between + and - increments once Increment = 1 //initial increment value once StratAvgB = 0 If Optimize = 10 Then //value is incremented by 1 at position entry or exit WinCount = 0 //Value used to determine minimum number of winning trades StratAvgA = StratAvgB StratAvgB = 0 For i = 1 to 10 Do //number of repittions to run past performance analysis If positionperf(i) > 0 Then WinCount = WinCount + 1 EndIF StratAvgB = StratAvgB + StrategyProfit[i] Next StratAvgB = StratAvgB/10 If (WinCount < 5 or StratAvgA > StratAvgB) and Mode = 1 Then Periods = Periods + (Increment+2) ElsIf (Winloss < 5 or StratAvgA > StratAvgB) and Mode = 2 Then Periods = Periods - (Increment*2) - 1 mode = 1 If Increment > 5 Then Periods = 42 //goes back to default value Increment = 1 //increment is reset Else Increment = Increment + 1 EndIF EndIF mode = 2 Optimize = 0 EndIf Please share your own experience. @Wing @Nicolas I know you said you have also experimented with this, would you mind to share?

This topic has 454 replies, 32 voices, and was last updated 2 years ago by Khaled.

Tagged: machine learning

Viewing 15 posts - 301 through 315 (of 455 total)

← 1 2 3 4 5 … 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 →

04/26/2020 at 2:03 PM #128210

Paul

still it seems it doesn’t respect the minimum value set.

Paul
Participant

Topics: 40
Replies: 812
Been thanked: 660 times

04/26/2020 at 2:31 PM #128214

Paul

@Bard you have ptrailing and I’am not sure if it’s written correct so it’s also a combined stoploss.

Looking at MAE however it seems correct and also when I add this for insurance;

if positionperf(0)<-2 then sell at market exitshort at market endif

1
2
3
4

if positionperf(0)<-2 then
sell at market
exitshort at market
endif

That’s a good ratio loss 50/ profit 500 🙂

Also because of the small stoploss the need for a reentry on new signal when in loss isn’t much needed.

So nice work! Now i’am gonna replace the ml engine to what I posted above to see if anything changes.

Paul
Participant

Topics: 40
Replies: 812
Been thanked: 660 times

04/26/2020 at 2:42 PM #128217

Paul

here’s a quick comparison. one big loss, result of buying near high at close and opening next day lower.

1 user thanked author for this post.

bertrandpinoy

Paul
Participant

Topics: 40
Replies: 812
Been thanked: 660 times

04/26/2020 at 2:57 PM #128222

Paul

Bard wrote:

What does adding once ValueX = StartingValue etc, do that it wasn’t doing before that code was added?

it’s setting the valuex/y when HeuristicsCycle >= HeuristicsCycleLimit is false, so the right value at the beginning for both. Well that’s my logic atleast 🙂

spotted now that you didn’t use

DEFPARAM CumulateOrders = False // Cumulating positions deactivated

1

DEFPARAM CumulateOrders = False // Cumulating positions deactivated

changed a few paramaters, results like this below; not bad!

1 user thanked author for this post.

Bard

Paul
Participant

Topics: 40
Replies: 812
Been thanked: 660 times

04/26/2020 at 2:58 PM #128224

Fran55

Aplícate to vectorial???

Fran55
Participant

Topics: 34
Replies: 210
Been thanked: 7 times

04/26/2020 at 3:02 PM #128225

Paul

vectorial is a great strategy. Yes would be very interesting.

Paul
Participant

Topics: 40
Replies: 812
Been thanked: 660 times

04/26/2020 at 3:05 PM #128227

Fran55

Please…

I no a programmer man.

I think aplícate this machine learning is very interesting.

Fran55
Participant

Topics: 34
Replies: 210
Been thanked: 7 times

04/26/2020 at 3:05 PM #128228

Fran55

In vectorial

Fran55
Participant

Topics: 34
Replies: 210
Been thanked: 7 times

04/26/2020 at 5:03 PM #128240

Bard

you have ptrailing and I’am not sure if it’s written correct so it’s also a combined stoploss.

It’s good: https://www.prorealcode.com/blog/learning/kinds-trailing-stop-proorder/

That’s a good ratio loss 50/ profit 500.

Something I read ages ago in an FXCM article about traders never allowing a good P:L ratio which made them losing traders. Minimum has to be 2:1 to allow for slippage, etc. In fact I worked out the the frequency of my day trading and then looked up one of those Risk of Ruin calculators and found in any one given year I would have 16 losing trades in any 12 month period just by the laws of probability!

Realised it can’t be called CycLimRes because the “Res” stood for Reset Period which I generally got better performance with when set to 0.5 or 1. Will have to experiment.

Bard
Participant

Topics: 131
Replies: 809
Been thanked: 99 times

04/26/2020 at 5:10 PM #128242

Bard

Can you do us a favour Paul and paste the code I posted here: https://www.prorealcode.com/topic/machine-learning-in-proorder/page/20/#post-128205

and set the same dates as those in the 4 screenshots? Only I’ve come back inside just now and when I tried cycling through different markets that I posted images of — and with what I am sure is untouched code — I get different results?
Odd… ?

Could you please post the code you have that produced the image in your post: https://www.prorealcode.com/topic/machine-learning-in-proorder/page/21/#post-128222

I want to try some different stops, like the Kase Dev Stop and Kaufman Volatility Stop to see if it can beat the ptrailing stop. Cheers!

Bard
Participant

Topics: 131
Replies: 809
Been thanked: 99 times

04/26/2020 at 5:52 PM #128251

Paul

Yeah I will check & post latest code, a bit later, eating & still working on the code 🙂

Just to mention, I removed the reset period in the code. It creates a variable and it’s maybe better to have it work continues.

Other thing, I wonder if you ptrailing stop works live, as accepted stoploss. Thanks for the link, it was the info I searched for! Changed it now I use

Set stop ploss xx ptrailing yy

1

Set stop ploss xx ptrailing yy

Yesterday I wanted to give up on this, you took it up and got it to work, and now together it’s more complete and working!

1 user thanked author for this post.

Bard

Paul
Participant

Topics: 40
Replies: 812
Been thanked: 660 times

04/26/2020 at 6:16 PM #128253

Bard

Thanks for the heuristic cycle additions, did it resolve “flat GraphY’s” over a range of settings? I am not sure I see the difference with or without but mainly because I don’t have any flay Ys’ in any of my tests with the current settings.

Re: Settings and comment: https://www.prorealcode.com/topic/machine-learning-in-proorder/page/20/#post-128201

If Boxsize Starting Value =40 but MaxValue also = 40, how is the ML able to step through different parameters to find the most profitable values? Just wondered also isn’t MinValue meant to = Increment?

Bard
Participant

Topics: 131
Replies: 809
Been thanked: 99 times

04/26/2020 at 7:11 PM #128258

Paul

Bard wrote:

If Boxsize Starting Value =40 but MaxValue also = 40, how is the ML able to step through different parameters to find the most profitable values? Just wondered also isn’t MinValue meant to = Increment?

that’s a reason to make the startingvalue lower then the max value. I think minvalue is really the minimum value.

While all looks nice, there is one problem.

It wants to place a stop order when the high crosses over or is higher then renkomax+boxsize, however it will always be refused because you can’t place a stop order at the moment and at level of the crossing. Tried to build in a solution.

Version; daily

1 user thanked author for this post.

Bard

Paul
Participant

Topics: 40
Replies: 812
Been thanked: 660 times

04/26/2020 at 7:35 PM #128264

Paul

Bard wrote:

Can you do us a favour Paul and paste the code I posted here: https://www.prorealcode.com/topic/machine-learning-in-proorder/page/20/#post-128205

1 user thanked author for this post.

Bard

Paul
Participant

Topics: 40
Replies: 812
Been thanked: 660 times

04/26/2020 at 9:15 PM #128276

Bard

Cheers for posting the screenshots of profits attained with the version 3 code I posted earlier (with the 4 screenshots). So it seems okay for you and that you can take my v3 code and paste it in your system and it gives 47.9k. I can’t get those results anymore like the Dow Daily 2 yr shown in one of my 4 screenshots, using the exact same code that produced them and which is now on this thread? (I’ve rebooted PRT too).

Did it give you a tbt warning during that 47k Dow backtest? I ask because I can’t run v3 now on 2 yr Dow without a tbt warning – using the same code, spread, dates and market?

It gave no tbt warning when I posted the code and those 4 screenshots earlier today. Your v4 also gives a tbt warning. Does it do that for you? I took your advice and reduced the date range.
From 2017 to present it gives a tbt warning but not for 2018 to present.

No worries, I’m glad I spurred you on Paul! Although I think I could do with a nudge to get me back on track. Today has been frustrating and one step forward with three behind… Time taken up sorting out issues like this tbt warning problem which is ruining the results… and yet IG say there is enough data back to 2010.

Bard
Participant

Topics: 131
Replies: 809
Been thanked: 99 times
Author

Posts