RUN BOTS LAST NIGHT, DIFFERENT RESULTS TODAY

Forums › ProRealTime English forum › ProRealTime platform support › RUN BOTS LAST NIGHT, DIFFERENT RESULTS TODAY

Viewing 13 posts - 1 through 13 (of 13 total)
  • 06/30/2016 at 2:04 PM #10029
    GraHal

    Hi

    Has anybody ever had this scenario or similar …

    I ran 4 bots last night at 10.00 ish over 100,000 1 hour bars … good results, all in profit!

    I ran the same 4 bots today at 14.00 ish over 100,000 bars … they all bomb-out / lose all my capital after about 12 monts or so.

    All on DEMO fortunately, but makes me feel like I was dreaming my good results from last night.

    I will not change any code and I will run them tonight at 10.00 and see what results I get and let you know.

    I have suspected this scenario previously so if anybody has had similar experience say so on here please.

    Cheers

    GraHal

    06/30/2016 at 6:24 PM #10047
    GraHal

    re above … that should read 12 months (not 12 monts).

    I am with IG in case anybody has similar experience and you are with IG also?

    06/30/2016 at 7:12 PM #10052
    Nicolas

    Are you talking about backtests difference from time to time? or running automated trading strategies on a demo account?

    06/30/2016 at 9:52 PM #10066
    GraHal

    Sorry yes Nicolas … I am talking Backtests.

    Don’t know why I said … all on DEMO as that’s not really relevant.

    07/01/2016 at 9:09 AM #10067
    Nicolas

    What about spread? Maybe you forgot to add it in one of your backtest?

    07/01/2016 at 4:19 PM #10092
    GraHal

    The spread on all 4 bots was set outside the code by ‘ticking the box’ and I never changed the spread from when I got good profit to the next day when they bombed.

    I tried again last night at 10ish and they all bombed again.

    Maybe I was dreanng after all!? ha

    You can delete this whole thread if you wish Nicolas?

    Thank you for your useful suggestions … as always.

    Cheers

    GraHal

    07/01/2016 at 4:31 PM #10094
    Nicolas

    Maybe you have lost the Holy Grail, shame on you! 🙂

     

    07/15/2016 at 9:01 PM #10496
    GraHal

    Yes I was thinking the same Nicolas 🙂 but no … the same thing happened today and I think I have sussed the reason.

    I backtested a bot earlier today and got big profits over 100,000 5M bars and just run the same bot and got loads less profit.

    I do  recall that earlier today the 100,000 5M bars started in 2012 but on my test 1 hour ago the 100.000 5M bars start on 13 Mar 2015 (about right?).

    I think it’s to do with selecting 100,000 bars @ a higher timeframe (e.g. 1 hour) and then dropping down to a lower timeframe (5M).  However, the start of the time period for 5M must stay as that for the (previously used) longer timeframe / 1H (or whatever)?  I nearly always leave the time period as ‘Earliest date displayed’.

    It must be a bug in the Platform (I’m with IG) because 100,000 5M bars should always be the same no matter what previous timeframe I have backtested on? I do have to select ‘x units’ and then 100,000 in order to get 100,000 bars but I guess that is the same for everybody?

    Am I missing a trick here?  Anybody else noticed this issue in anyway at all please?

    Thanks

    GraHal

     

    07/16/2016 at 2:55 PM #10499
    Derek

    I think it’s to do with selecting 100,000 bars @ a higher timeframe (e.g. 1 hour) and then dropping down to a lower timeframe (5M).  However, the start of the time period for 5M must stay as that for the (previously used) longer timeframe / 1H (or whatever)?  I nearly always leave the time period as ‘Earliest date displayed’.

     

    I think that’s it. You can load 10.000 bars and the reduce the number to 1.000 but the remaining 9.000 bars will still be “on the chart” if you zoom out.

    The complete timeframe also remains in backtests.

     

    How do you load 100.000 bars with IG? I am maxing out at 10.000.

    07/19/2016 at 5:28 PM #10627
    GraHal

    Hi @Derek you have to select ‘X Units’ and then change the number of bars to 100,000.

    You can do a similar thing re X minutes etc to get odd timeframes, but ProOrder wont always accept the odd timeframes.

    Cheers

    GraHal

    07/19/2016 at 6:35 PM #10633
    Elsborgtrading

    Hi GrayHal. I experienced similar situation, also with IG. Specially: “I do  recall that earlier today the 100,000 5M bars started in 2012 but on my test 1 hour ago the 100.000 5M bars start on 13 Mar 2015 (about right?)”

    you test and test and test and sometimes you don’t produce the same result. I think with all that testing and sometimes multiple time frames in the screen it easily gets mixed up, especially with that behaviour you stated before. What I do is, keep a Log and screenshoot of all sucessfull test and codes+remarks, and I put it in MS onenote- very easy. Before running the test again, I close the “result window” the Equity” and the “position” windows in the chart you are testing. That for me gives the same result every time.

    Cheers

    Elsborgtrading

    07/20/2016 at 9:22 AM #10670
    GraHal

    Hi Elsborgtrading, thanks … yes, I need to adopt your good disciplined practices also.

    Cheers

    GraHal

    07/21/2016 at 12:20 PM #10759
    Derek

    @GraHal: Thanks for the reply. I must have misread your comment because 100.000 bars is just right.

  • Author
    Posts
Viewing 13 posts - 1 through 13 (of 13 total)

Find exclusive trading pro-tools on ProRealCode Marketplace

Similar topics:
different backtest results using different pc’s
“proscreener” and “probuilder” deliver different results with identical code
different backtest results depending of atr code placement
optimization results different from backtest
same script but different results in demo and live ig cfd account
so no thrown out systems last night on fed’s decision – but no trades either
two seemingly identical screener – but different results
adaption/addition to last man standing indicator
ai bots – anybody tried them?
back testing results

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