Problem with condition

Despair · 2017-07-11T11:00:32+00:00

I have a strange problem. I want my condition to be true if a moving average goes flat (sideways). So i have something like: once i=0 if ema=ema[1] then i=i+1 endif graph ema as "ema" graph ema[1] as "ema[1]" graph i as "i" When i run the code the condition obviously never comes true since i remains zero all the time. Anyway when i look at the graphed variables the backtest shows many periods when ema=ema[1]. How can this be? Is that maybe that PRT intern calculates the emas with more than 5 digits and then cuts off the last digits when graphing them? This way there is a tiny difference between ema and ema[1] though graph shows the same value for them? This takes me to the question, with how many digits behind the comma is PRT calculating?

Viewing 5 posts - 1 through 5 (of 5 total)

Author

Posts
07/11/2017 at 12:00 PM #40345 quote
Despair
Blocked

Master
I have a strange problem. I want my condition to be true if a moving average goes flat (sideways). So i have something like:
```
once i=0
if ema=ema[1] then
  i=i+1
endif
graph ema as "ema"
graph ema[1] as "ema[1]"
graph i as "i"
```
When i run the code the condition obviously never comes true since i remains zero all the time. Anyway when i look at the graphed variables the backtest shows many periods when ema=ema[1]. How can this be? Is that maybe that PRT intern calculates the emas with more than 5 digits and then cuts off the last digits when graphing them? This way there is a tiny difference between ema and ema[1] though graph shows the same value for them?

This takes me to the question, with how many digits behind the comma is PRT calculating?
07/11/2017 at 1:51 PM #40355 quote

Despair
Blocked

Master

Obviously there is a problem with hidden digits. PRT shows 5 digits but internally it must calculate with more digits. Does anybody know with how many?

07/11/2017 at 3:06 PM #40374 quote

robertogozzi
Moderator

Legend

Which TF did you use?

I your assumption might be possible for very low TF’s, say 1 minute or lower, I think it would be extremely odd to have EMA’s matching the first 5 digits in a, say, 4-hour TF!

07/11/2017 at 6:54 PM #40381 quote

Despair
Blocked

Master

I was using 5 minute TF. And it is definitely th case with some hidden digits. When you graph the variables it comes back sometimes as “0” and sometimes as “-0.00” so obviously there is something going on one doesnt see (again!).

07/11/2017 at 10:27 PM #40392 quote
robertogozzi
Moderator

Legend
I had to modify your code as follows to be executed:
```
once i=0
ema = average[20]
if ema=ema[1] then
   i=i+1
endif
if i < 0 then
   buy 1 contract at market  //useless condition just to add a BUY order
endif
graph ema    as "ema"
graph ema[1] as "ema[1]"
graph i      as "i"
```
I run it on DAX 5-min TF and it was executed regularly, I found a match on Friday July 7th at 09:35 (gmt +2), as from the picture.

Variable i was set at 11, so there must have been 10 matches before.

Prices seem regular with the full 5 digits + 5 decimal digits.

Ema-test.jpg
Author

Posts