Let’s say

buyCondition = average[10] crosses over average[30]
if A and B and C and buyCondition then
    ...
endif

the result will be totally different from:

if buyCondition and A and B and C then
    ...
endif

I found this abnormal behavior when I do my experiment on Nicolas’s nice post (This abnormal behavior is nothing to do with the code of Nicolas’s post. I show the post here is just to give a code base for a quick experiment):

https://www.prorealcode.com/blog/learning/how-to-improve-a-strategy-with-simulated-trades-1

First, you can change the code block (at # line: 22):

elsif  not realtrading and ordercount>equityCurvePeriod then //fake trading
if not longontrading and buysignal then

into:

elsif  buysignal and not realtrading and ordercount>equityCurvePeriod then //fake trading
    if not longontrading  then

or

elsif  buysignal and not realtrading  then //fake trading
    if ordercount>equityCurvePeriod and not longontrading  then

and run both. You can easily see the difference.

note that logic of code blocks is equivalent to:

elsif buysignal and not realtrading and ordercount>equityCurvePeriod and not longontrading then

You may switch each position of those conditions around. (and see the abnormal when buysignal is in 1st or 2nd position)

I don’t know whether such abnormal is due to PRT’s intrinsic bug or maybe there is a basic rule for if… condition that I don’t know yet.

Does someone know the cause? And how to avoid such unpredictable behavior?

ps. also there is a bug in backtesting when running for i=0 to aNumberLargerThan500, the “i” will never exceed 500. Everyone can reproduce it by putting this code into backtesting :

for i=0 to 999 do
    cantBiggerThan500 = i
next
graph cantBiggerThan500

Is this bug will be fixed on next PRT version?

Thanks!