Why Does Removing ONE Line of Unused Space Cause Profits to Reappear?

Forums ProRealTime English forum ProOrder support Why Does Removing ONE Line of Unused Space Cause Profits to Reappear?

  • This topic has 30 replies, 4 voices, and was last updated 4 years ago by avatarBard.
Viewing 15 posts - 1 through 15 (of 31 total)
  • #130880

    So, I was running a system “MBFX Belkhayate” and then decided to apply machine learning (ML) to it last night, I called that one “MBFX Belk ML2.” I got around £1,700 profits in the ML version and £1,800 in the non ML system. (The ML had better stats but slightly lower profit).

    Long story short, I spent about an hour altering some MBFX Belk ML2 system starting values and max value settings, adding pProfit targets etc. BUT, when I went back to the original MBFX Belk ML2 settings I couldn’t get the £1,700 profit anymore? It was losing money.

    I tried for another hour to replicate the results, stripping out ML code and repasting it back in just as I had when I first created the  MBFX Belk ML2 system. Nope, still losses….

    So…. I started taking out chunks of ML code from the ML system this morning having given up  trying to replicate the original ML backtest results last night.

    Eventually I got to what I thought was a near identical version of the non ML, “MBFX Belkhayate” system.
    But still the MBFX Belk ML2 made losses. The code looked the same because everything superfluous was “//” out (pls see first image).

    I went further (2nd image), nope. still losses!?

    So then I removed one line of unused space between the “DEFPARAM…” (line 2) and the “//Conditions to go long” (line 5) and NOW the profits are the same. Huh?

    Why does one line of unused space have such a massive impact on profits? This would also explain why I’ve had this issue many times before when I couldn’t match profits with previously screen-shotted versions of systems.

    Thanks for any insights, because this kind of issues has cost me days of wasted time over the years!

    #130886

    Here’s the unprofitable code with an extra line of unused space:
    “No Profit because of one line of unused space after DEFPARAM”

    and now code with the space removed:
    “Now get profits because one line of unused space was removed:”

     

    #130909

    How can you trust this platform?

    Please see images and blue profitable equity curves from yesterday)and brown curve and loss for today.

    EDIT:  My first two images are small so I’ve posted two slightly larger ones showing just the System code and Detailed Reports from yesterdays profitable test (A) and todays losses with the same code / settings / dates / spread, (B).

    BLUE EQUITY CURVE: Screenshot 1 shows profits from a non machine learning simple “MBFX Belkhayate” system when it was profitable last night…. and this morning when it decided not to be –>>

    BROWN EQUITY CURVE: Screenshot 2 shows the same system today with no profits. It’s the same code — not edited/removed and repasted, — just completely the same unadulterated code that I used last night.

    Notice the settings on the Belkhayate indicator have not changed overnight either (0,7).

    So has anyone please got a logical explanation what is going on with PRT and it’s “sensitive” code? Thanks!

     

     

    #130923

    I copied y0ur code  and downloaded the Indicator, but now it’s asking for more variables.

    If you want to add the .itf so then I’ll have the variables in the Indicator (?) I’ll see if I can come up with anything?

    I have in the past had similar scenarios … made profits on backtest and then  a small change and it all changes beyond what would be expected.  You may have unearthed the reason!

    So glad you didn’t leave , how would we have found all this weirdness!? 🙂

     

    #130962

    but now it’s asking for more variables.

    Cheers, and thanks for taking a look, what variables is it asking for because I don’t have the ML version anymore as I have tried to rebuild it for the 5th time?..
    Such a waste of time. From now on, the minute I have something that works I will copy the code immediately instead of doing that at the end of my work session.
    4 hours gone for nothing and both system were profitable right at the very beginning last night… talk about chasing your tail…

    I swear it was literally changing that one line of unused space (by deleting it) in the ML version (on the LHS of image) that made a difference between it making  money or losing. Smh. And now none of the profitable non ML systems work today either? Double smh…

    Here’s the itf for the non ML version:

    First image was yesterdays test on the Swissy with profits. Second image, same system same settings, and losses?

    I think it’s the pLoss. I think it’s giving fantasy results like the trailing stop did in the Renko ML systems, I say that because the number of trades is typically 3x’s as high as it was yesterday. Its stopping trades out with losses and more new trades are being taken. It’s a hunch.

    #130967

    Not related to your question but why do you call the indicator twice? Just call it once and get all the values in one go.

     

    #130970

    Had an idea about this and then realised it wasn’t that…

    #130971

     

    why do you call the indicator twice?

    It’s just the way the “simplified creation” tool writes it when I create the strategy. But you’re right, it is cumbersome. Any clues/ideas on the unused line of spacing causing issues with backtests?

    #130980

    I just tested it. I’m guessing you have changed the indicator slightly because the library version has a slightly different name to the one used in your strategy and also returns only four values where your CALL is looking for six. Anyway once I got it working it tested just fine and I get the exact same result with any amount of spacing.

    #130988

    Thanks for taking a look. Yes I modified it but only started working with it after the mods had been done, that’s how I got the system to buy and sell on the lowest and highest lines. I started backtesting the MBFX Belkhayate systems last night and the results are different today, really different! Pls see pics.

    Re: Indicator might of been changed slightly, surely having the code state the Len and filter settings :” CALL “PRC_MBFX Timing v2”[5, 0]” means it will take precedent over the indicators settings even if I had changed them?

    Deleting that unused line made the diff difference between profits and losses. The screenshots above were taking before and after that change to the line.

    My platform is unstable. Period. There is now an “investigation.” For example I went to open one of 6 indicators an hour ago — ones that I created in the last few days (I posted some here: https://www.prorealcode.com/topic/8x-multicolour-market-states/#post-130681) and they were not in my indicator list. Only on quitting and logging back in could I find any of them when searching on the same search term: “x” Pls see image of the indicators reappearing after logging out and back in.

    Tbh, I am quite literally spending 75% of my time sorting/dealing with bugs and 25% with development of ideas. Not sure how much longer I can work like this.

     

    #131000

    Next bug you get, post the exact code on which you are experiencing the bug.

    If a modified Indicator is used then  post the modified Indicator.

    I would run the code on my Platform and report back … just as Vonasi  has done above (no bug found).

    Ideally we need a Mac user to provide a 2nd parallel trial?

     

    1 user thanked author for this post.
    #131009

    I just realized that I’m testing pretty much the same strategy that you are. https://www.prorealcode.com/topic/cog-mbfx-timing-v2/

     

    #131011

    Thank you. That was the exact code but short of perhaps potential issues with pLoss and the vast increase in trade number today, I believe the problem is inherent within this account and that even if I’d sent the itf’s it would have been fine your side. I don’t know if it’s possible to restore my faith in this platform after 7 months of what I can only describe as slow torture. I have lost all faith in PRT to ever resolve it. I’ve already started looking at other vendors. And you know what my gut instinct after using PRT and coming across loads of issues 10 years ago was? Avoid. It was only because of IG’s strong reputation that I became associated with the platform again. Mistakes like the scrolling issue just should not be happening in any software organisation. I am still incredulous that they allowed that gaff to slip through the net.

    #131021

    I’m beginning to think that you might be on one of those ultra reliable and stable Apple Mac computers that never have bugs or viruses?

    All seems to work fine on Windows10.

    Perhaps we need some more Mac users to test it for you – but sometimes when you go off peak and then you get stuck in a bit of poo poo rescuers can be a little harder to find! Any Mac users able to test for Bard?

    #131022

    I believe the problem is inherent within this account and that even if I’d sent the itf’s it would have been fine your side.

    Within this account … what does that mean?

    It sounds like the bugs are only experienced by you?

Viewing 15 posts - 1 through 15 (of 31 total)

Create your free account now and post your request to benefit from the help of the community
Register or Login