Machine Learning in ProOrder ProRealTime

• 05/01/2020 at 9:28 AM #129167
  JC_Bywan

  Bard wrote:

  Strange that I get so many email notifications though, but I am now wondering how many @ bard’s I might have missed here and which in fact went to a guy who hasn’t been posting here for 3 yrs and 8 months?

  Bard ( @Brad ) there are several other reasons than tagging for receiving email notifications, all of these can be customised with on/off in the “settings” section, go to your avatar in the top right corner, menu appears, choose “settings” just above “log out”, then go to the bottom to see the email notifications section, and if you modify anything don’t forget to click on “save settings” before leaving the page.

  You may also receive emails if you have subscribed to some topics. Top right corner of each topic, where you clicked “subscribe” it’s been replaced by a clickable”unsubscribe” in case you don’t want to receive emails anymore about a given topic.

  2 users thanked author for this post.

  Bard, GraHal

  

  JC_Bywan

  Moderator

  Topics: 20

  Replies: 2657

  Been thanked: 925 times
  05/01/2020 at 2:06 PM #129217
  Bard

  GraHal wrote:

  “Brad”

  Surely that’s a mistake and stop calling me Shirley! Lol.

  I think that admin alert distracted us from my last post, did you see the ML3, now all fully working? Also how are you’re ML\s coming along and did you ever suss out putting back in the //BestA = 300 code?

  Cheers, Axa Framlington Global Technology Z Incorporated
  Best, bARd!

  

  Bard

  Participant

  Topics: 138

  Replies: 814

  Been thanked: 99 times
  05/02/2020 at 4:30 PM #129447
  Fran55

  Go machine learning to vectorial????

  

  Fran55

  Participant

  Topics: 34

  Replies: 210

  Been thanked: 7 times
  05/02/2020 at 4:33 PM #129450
  zilliq

  The code in Vectorial is very heavy, so if we add this code it will take centuries to work…

  Actually all my backtest take 1-2 hours for 10-15 000 units 🙁

  

  zilliq

  Participant

  Topics: 58

  Replies: 683

  Been thanked: 152 times
  05/02/2020 at 4:42 PM #129452
  fifi743

  zilliq wrote:

  The code in Vectorial is very heavy, so if we add this code it will take centuries to work…

   

  

  fifi743

  Participant

  Topics: 10

  Replies: 913

  Been thanked: 426 times
  05/02/2020 at 4:53 PM #129455
  nonetheless

  I used to have a bike exactly like that, Peugeot 10 speed. Stolen in london about 30 years ago. I bet that’s it.

  Can’t believe you stole my bike. 😣

  2 users thanked author for this post.

  swedshare, fifi743

  

  nonetheless

  Participant

  Topics: 83

  Replies: 1237

  Been thanked: 978 times
  05/02/2020 at 5:05 PM #129460
  fifi743

  yes, it’s a very good bike

  

  fifi743

  Participant

  Topics: 10

  Replies: 913

  Been thanked: 426 times
  05/03/2020 at 2:50 PM #129588
  Bard

  Hey there, can someone please do me a favour and use this Mean Reversion ML2 code itf, (attached below) into their platform? It’s a simple and interesting system first developed by Francesco and based on Bollinger entries and Moving Average exits and was working great on the USD/CHF and LCO (Brent Crude).

  It used to plot X and Y perfectly, — after a long time getting the ML code to work and playing around with optimisations and the settings like the starting value and max value, — meaning that you could clearly see steps in values on Friday night but when I opened it up Saturday morning it flatlines with the same starting etc values?

  Some on this thread know that I have had erratic issues with the “save” option on my platform (since Oct 2019), like I can add indicators use them in systems and then find they’re not in my platform the next time I launch. I would like to know if it’s my platform not working or the code? And if it’s not the platform if anyone could figure out how I got the x and y to plot in steps Friday but can’t now with what I believe to be the exact same code (because I copied and pasted the settings into an Apple Text Edit app)?
  Here’s the working screenshot of the x and y plots moving in steps on Friday night: pls see screen. I just can’t work out why the x and y graphs won’t plot anymore and keep “deadlining” flat?

  Much appreciated, thank you in advance.

  

  Bard

  Participant

  Topics: 138

  Replies: 814

  Been thanked: 99 times
  05/03/2020 at 4:16 PM #129602
  bertrandpinoy

  I can install it in demo mode tomorrow morning if you want. USD / CHF daily is it good?

  

  bertrandpinoy

  Participant

  Topics: 24

  Replies: 121

  Been thanked: 8 times
  05/03/2020 at 4:21 PM #129607
  Bard

  Renko ML3.

  Two ML algos for Long/Short Boxsize created by Paul and the third for pLoss:

  Nice wide range of values for starting and max should provide robustness going forward. Would anyone like to apply a VRT to it and can ML pLoss be trusted? Still got to sort my VRT/Excel out.

  1 user thanked author for this post.

  mr blue

  

  Bard

  Participant

  Topics: 138

  Replies: 814

  Been thanked: 99 times
  05/03/2020 at 4:31 PM #129611
  Bard

   

  bertrandpinoy wrote:

  I can install it in demo mode tomorrow morning

  Cheers, that’s kind of you but I think I have sussed it. I had used the below:

  If (onmarket[1] = 1 and onmarket = 0) or (longonmarket[1] = 1 and longonmarket and countoflongshares < countoflongshares[1]) or (longonmarket[1] = 1 and longonmarket and countoflongshares > countoflongshares[1]) or (shortonmarket[1] = 1 and shortonmarket and countofshortshares < countofshortshares[1]) or (shortonmarket[1] = 1 and shortonmarket and countofshortshares > countofshortshares[1]) or (longonmarket[1] and shortonmarket) or (shortonmarket[1] and longonmarket) Then

  1 2 3

  If (onmarket[1] = 1 and onmarket = 0) or (longonmarket[1] = 1 and longonmarket and countoflongshares < countoflongshares[1]) or (longonmarket[1] = 1 and longonmarket and countoflongshares > countoflongshares[1]) or (shortonmarket[1] = 1 and shortonmarket and countofshortshares < countofshortshares[1]) or (shortonmarket[1] = 1 and shortonmarket and countofshortshares > countofshortshares[1]) or (longonmarket[1] and shortonmarket) or (shortonmarket[1] and longonmarket) Then

  to get the x and y values stepping. That’s how I saved it. But with the save issues I have it loaded an earlier version today which had the following incorrect code (for this type of system) instead:

  If onmarket[1] = 1 and onmarket = 0 Then optimise2 = optimise2 + 1 Endif

  1 2 3

  If onmarket[1] = 1 and onmarket = 0 Then optimise2 = optimise2 + 1 Endif

  Question is how can you tell in advance which code to use and whether the system will change direction with no bars on market (use longer version) or use the original shorter version?

  

  Bard

  Participant

  Topics: 138

  Replies: 814

  Been thanked: 99 times
  05/03/2020 at 4:35 PM #129613
  Bard

  Anyone else feel like they’re a kid in a candy shop at the moment with this amazing ML code?

  I had another idea can you set the ML onto the timeframe? Should I use a 2 x 1 hour green candles to determine if it’s a long breakout and long entry for the day or a 45 minute, or a 22 minute or 10 minute candle?

  I wonder what would happen with 4 x ML’s….

  

  Bard

  Participant

  Topics: 138

  Replies: 814

  Been thanked: 99 times
  05/03/2020 at 4:44 PM #129615
  Bard

  Bard wrote:

  I had used the below:

  Remember to change the optimise1 = optimise1 + 1 for “2’s” on the second ML algo.

   

  

  Bard

  Participant

  Topics: 138

  Replies: 814

  Been thanked: 99 times
  05/07/2020 at 2:45 PM #130455
  Bard

  Vonasi wrote:

  With VRT off all the results were identical and with it on they were all different.

  Hi Vonasi, I was just looking up my failed VRT test experiments. Failed because with VRT on I get twice or even 5x’s the profits compared to a non VRT system. That is a failure right because I was re-reading all your posts about this and one of your posts said the results will be different. By “different” do you mean double the profits compared to a non VRT system, because I can get results that are 5x bigger?

  Not being a coder I’m not sure what qty and random really mean? 10 run throughs with “5 random: whats?

  Or how the two flag codes even make a system jumble up it’s trade order?

  Or what the “mod” means or does?

  if barindex mod qty = 0 or barindex mod qty = j then tradeon = 1

  1 2	if barindex mod qty = 0 or barindex mod qty = j then tradeon = 1

  or why a test with settings:

  “I tested with settings of 10/5 and 20/5 for QTY and RANDOM. So that is 45 random samples and 95 random samples.”

  produces 45 or 95 random tests? Okay: 10 tests and 5 “randoms” (whatever that means), so 50 tests less “5 randoms”? (whatever that means) = 45? I actually don’t get what is happening here?

  I also am not able to turn VRT off without having to // out all the VRT code out whether I set your orig code

  tradeon = 0 if opendate >= startdate then...

  1 2	tradeon = 0 if opendate >= startdate then...

  to:

  tradeon = 1 if opendate >= startdate then...

  1	tradeon = 1 if opendate >= startdate then...

  or add a Robustness switch at the very top of the VRT code above StartDate

  RobustnessTest = 0

  1	RobustnessTest = 0

  and

  if not RobustnessTest then tradeon = 1 endif

  1 2 3

  if not RobustnessTest then tradeon = 1 endif

  as the very last line of instruction in the VRT code.

  Sorry for all the questions, I just don’t get this one and I really want to see the VRT results of my systems! (Subject to then wrestling with Excel where VRT results I copy and pasted over sometimes sync nicely with the template and other times don’t, (I’ve used Excel extensively for copy and pasting PRT optimised results and charting the results without issue). Cheers, Bard.

  

  Bard

  Participant

  Topics: 138

  Replies: 814

  Been thanked: 99 times
  05/07/2020 at 3:29 PM #130464
  Franro

  Anyone still using the 1m version?

  

  Franro

  Participant

  Topics: 25

  Replies: 219

  Been thanked: 23 times
• Author
  Posts