Strategy MAEX – experiment with MFE & MAE

• 01/18/2021 at 6:34 PM #158345
  nonetheless

  Anyone else get the dreaded div by zero for this? Apart from the indicators, the only lines I can see with division by anything are

  averecentmael=close-(maerec/nn) averecentmfel=close+(mferec/nn) averecentmaes=close+(maerec/n) averecentmfes=close-(mferec/n)

  1 2 3 4 5

  averecentmael=close-(maerec/nn) averecentmfel=close+(mferec/nn)   averecentmaes=close+(maerec/n) averecentmfes=close-(mferec/n)

  but n and nn have permanent values.

  🤔🤔🤔???

   

  

  nonetheless

  Participant

  Topics: 83

  Replies: 1237

  Been thanked: 978 times
  01/18/2021 at 7:28 PM #158368
  Paul

  always that error!

  It’s either in that section or in the vectorial part.

  How about adding before

  if range[0]>0 and range[1]>0 and range[2]>0 and range[3]>0 and range[4]>0 and range[5]>0 then

  1	if range[0]>0 and range[1]>0 and range[2]>0 and range[3]>0 and range[4]>0 and range[5]>0 then

  maybe up to 8/9/10, to include vectorial. Not very nice but it makes sure when the calculation starts every bar has a range.

  Otherwise use min(x,x) for certain parts?

   

   

  

  Paul

  Participant

  Topics: 40

  Replies: 812

  Been thanked: 660 times
  01/18/2021 at 7:39 PM #158370
  nonetheless

  So, where would i add that line exactly? at the top of the //mfe & mae section?

  

  nonetheless

  Participant

  Topics: 83

  Replies: 1237

  Been thanked: 978 times
  01/18/2021 at 10:48 PM #158408
  Paul

  Something like this,

  I thought I tried it in the past and in that situation it didn’t work, but I’ve got no other solution.

  vectorial seems ok and also Stochastic RSI

  ps you get check the time of the error and see if in the chart test=1

  TIMEFRAME(default) // mfe & mae if range[0]>0 and range[1]>0 and range[2]>0 and range[3]>0 and range[4]>0 then n = 4 nn= 1 if method=1 then maerec=0 mferec=0 for i=1 to n maerec=maerec+(high[i]-open[i]) mferec=mferec+(open[i]-low[i]) next for j=1 to nn maerec=maerec+(open[j]-low[j]) mferec=mferec+(high[j]-open[j]) next averecentmael=close-(maerec/nn) averecentmfel=close+(mferec/nn) averecentmaes=close+(maerec/n) averecentmfes=close-(mferec/n) avgl=(averecentmael+averecentmfel)/2 avgs=(averecentmaes+averecentmfes)/2 elsif method=2 then maerec=0 mferec=0 maerec2=0 mferec2=0 for i=1 to n maerec=maerec+(high[i]-open[i]) mferec=mferec+(open[i]-low[i]) next for j=1 to nn maerec2=maerec2+(open[j]-low[j]) mferec2=mferec2+(high[j]-open[j]) next averecentmael=close-(maerec2/nn) averecentmfel=close+(mferec2/nn) averecentmaes=close+(maerec/n) averecentmfes=close-(mferec/n) avgl=(averecentmael+averecentmfel) avgs=(averecentmaes+averecentmfes) endif endif if range[0]=0 or range[1]=0 or range[2]=0 or range[3]=0 or range[4]=0 then test=1 else test=0 endif graph test

  1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66

  TIMEFRAME(default) // mfe & mae   if range[0]>0 and range[1]>0 and range[2]>0 and range[3]>0 and range[4]>0 then   n = 4 nn= 1   if method=1 then   maerec=0 mferec=0   for i=1 to n maerec=maerec+(high[i]-open[i]) mferec=mferec+(open[i]-low[i]) next   for j=1 to nn maerec=maerec+(open[j]-low[j]) mferec=mferec+(high[j]-open[j]) next   averecentmael=close-(maerec/nn) averecentmfel=close+(mferec/nn)   averecentmaes=close+(maerec/n) averecentmfes=close-(mferec/n)   avgl=(averecentmael+averecentmfel)/2 avgs=(averecentmaes+averecentmfes)/2   elsif method=2 then maerec=0 mferec=0 maerec2=0 mferec2=0   for i=1 to n maerec=maerec+(high[i]-open[i]) mferec=mferec+(open[i]-low[i]) next   for j=1 to nn maerec2=maerec2+(open[j]-low[j]) mferec2=mferec2+(high[j]-open[j]) next   averecentmael=close-(maerec2/nn) averecentmfel=close+(mferec2/nn)   averecentmaes=close+(maerec/n) averecentmfes=close-(mferec/n)   avgl=(averecentmael+averecentmfel) avgs=(averecentmaes+averecentmfes) endif endif   if range[0]=0 or range[1]=0 or range[2]=0 or range[3]=0 or range[4]=0 then test=1 else test=0 endif   graph test

   

  

  Paul

  Participant

  Topics: 40

  Replies: 812

  Been thanked: 660 times
  01/19/2021 at 10:55 AM #158448
  Tanou

  Hi guys!

   

  What is this error? I dn’t see it when I backtest! How does it affects the code?

  

  Tanou

  Participant

  Topics: 18

  Replies: 153

  Been thanked: 29 times
  01/19/2021 at 11:34 AM #158458
  nonetheless

  it doesn’t show up in backtest, that is what’s so infuriating. Just when you think you’ve done the hard work and everything is sweet, you launch it but it wont run live because ‘division by zero’. They don’t tell you where the problem is and it’s almost never apparent in the code because the values are changing all the time. Try doing a search, there’s dozens of threads on the subject but I don’t think anyone’s found a cure.

  For me it is definitely worse in v11. Algos that ran fine in 10.3 are now getting stopped for this regularly.

  

  nonetheless

  Participant

  Topics: 83

  Replies: 1237

  Been thanked: 978 times
  01/19/2021 at 12:33 PM #158473
  Monobrow

  Yeah, ive had a bunch of div 0 and similar errors recently. Very frustrating but a bit of hard work and help from the forum has solved them all.

  Fact is there will be something attempting to div/0 and a well placed MAX(<something above zero that doesn’t break your system>, <offending code>) will fix most of them.

  graph all your variables/returning values,  find the zero and it should be possible to fix.

  You can also try using If variable <> 0 then <do dodgy code>  Endifs also.

  its a pain but it is what it is.

  

  Monobrow

  Participant

  Topics: 15

  Replies: 157

  Been thanked: 23 times
  01/19/2021 at 2:56 PM #158513
  Paul

  I don’t think it’s the solution above/below works though in my post, because it only  solves to have a value before dividing it. And the error appears to be the other way around.

  Problem is a zero is never used as far I can tell.

  Anyway another way to have it programmed, instead testing and coding each bar individually, like this below. Should only take trades if the last xx bars have range and not have open=close=high=low.

   

  // mfe & mae count = 0 For zz = 1 to xx do IF range[zz]=0 THEN count= 1 ENDIF next if count=0 then n = 4 nn= 1

  1 2 3 4 5 6 7 8 9 10 11 12 13

  // mfe & mae   count = 0 For zz = 1 to xx do IF range[zz]=0 THEN count= 1 ENDIF next   if count=0 then   n = 4 nn= 1

  wild guess, but what if with live data feed i.e. there’s no tick/bar data is not the same as the backtest data?

  Something like backtestdata has open=low=high=close and running live the bar is simply skipped or the other way around.

  Maybe such bar is processed another way but not visible and maybe the error is broader then the error message itself gives indication.

   

   

  

  Paul

  Participant

  Topics: 40

  Replies: 812

  Been thanked: 660 times
  01/19/2021 at 4:00 PM #158522
  nonetheless

  Thanks Paul, I’ll try that and see if it runs 🤞

  

  nonetheless

  Participant

  Topics: 83

  Replies: 1237

  Been thanked: 978 times
  01/19/2021 at 5:48 PM #158536
  Tanou

  @nonetheless, can you please upload a .itf file? As I don’t see the error I’d like to see what’s the difference… thanks 😉

  

  Tanou

  Participant

  Topics: 18

  Replies: 153

  Been thanked: 29 times
  01/19/2021 at 5:53 PM #158539
  nonetheless

  DJ-1m-Maex-v4.5-atr.itf

  

  nonetheless

  Participant

  Topics: 83

  Replies: 1237

  Been thanked: 978 times
  01/19/2021 at 6:31 PM #158551
  Tanou

  Have you change something regarding the previous version? I don’t see any changes!

  

  Tanou

  Participant

  Topics: 18

  Replies: 153

  Been thanked: 29 times
  01/19/2021 at 6:46 PM #158553
  nonetheless

  at the moment I have it in demo with Paul’s suggestion above but hasn’t taken a trade yet. When you say, ‘I don’t see the error’, do you mean you dont get division by zero rejection?

  

  nonetheless

  Participant

  Topics: 83

  Replies: 1237

  Been thanked: 978 times
  01/19/2021 at 7:29 PM #158558
  Tanou

  Yes exactly! Can you upload the version with the implementation of Paul’s suggestion?

  

  Tanou

  Participant

  Topics: 18

  Replies: 153

  Been thanked: 29 times
  01/19/2021 at 7:41 PM #158559
  nonetheless

  it’s in my demo account that’s not open just now. but it’s exactly the same as v4.5 atr but add

  count = 0 For zz = 1 to 5 do IF range[zz]=0 THEN count= 1 ENDIF next if count=0 then

  1 2 3 4 5 6 7 8

  count = 0 For zz = 1 to 5 do IF range[zz]=0 THEN count= 1 ENDIF next if count=0 then

  under the // mfe & mae section, with another endif at the bottom. There’s no other changes.

  

  nonetheless

  Participant

  Topics: 83

  Replies: 1237

  Been thanked: 978 times
• Author
  Posts