Brian, thank you for the very thorough response and examples.
I have included below a section of code which demonstrates the very different
results computed using two different techniques.
One uses TimeFrameSet() and TimeFrameRestore().
The other uses TimeFrameCompress() and TimeFrameExpand().
Now matter how I setup the database market hours the time frame set/restore
method gives inaccurate results, as compared to values calculated manually in
an excel spread sheet.
The time frame compress/expand is the only method I've found which produces
accurate results. However, when you have to call these functions dozens of
times in a AFL code it drastically slows the processing time of backtests and
optimizations.
It is a multi-stage process. First determine values during market hours, then
compress the result into daily time frame. Once in the daily time frame it is
now possible to perform operations involving multiple days, like calculating
moving averages. Once the values are calculated you must then expand the
variable before it can be used in the intra-day time frame.
This is the reason I am searching for a better solution. If I could nest all of
my calculations within a single time frame set/restore statement I could
drastically reduce processing time.
If someone has a solution to this I would be very grateful to know it.
Thanks.
Pete :-)
_Section_Begin("Time Frame Test");
MrktOpn = TimeNum() == 093000;
MrktCls = TimeNum() == 160000;
TimeBar = TimeNum() >= 160000;
TimeBar2 = TimeNum() < 160000;
EOSTimeBar = TimeBar == 1 AND Ref(Timebar2,-1) == 1;
//the only way I've found to calculate the true difference
//between daily H and L which are only taken during market hours
DlyHighest = HighestSince(MrktOpn , H);
HighAtClose = ValueWhen(EOSTimeBar , Ref(DlyHighest,-1));
DlyLowest = LowestSince(MrktOpn , L);
LowAtClose = ValueWhen(EOSTimeBar , Ref(DlyLowest,-1));
TRUE_DlyDiff = HighAtClose - LowAtClose ;
//then it needs to be compressed before using it to calculate
//a daily moving average of the difference between H and L
tfc_DlyDiff = TimeFrameCompress(TRUE_DlyDiff, inDaily, compressLast);
TRUE_DlyMA = MA(tfc_DlyDiff,10);
//then it gets expanded for use in Ploting to other method of
//display
tfeDlyMA = TimeFrameExpand(TRUE_DlyMA,inDaily);
//What I would prefer to use is the timeframeset, use it once
//and perform various calcluations like the one above and have
//it accurately produce results. If you check the values calculated
//using the timeframeset section below against values manually
//calculated in a spread sheet you will see they are including
//data from outside of market trading hours.
TimeFrameSet(inDaily);
FALSE_DlyDiff = H-L;
FALSE_DlyMA = MA(HighAtClose - LowAtClose, 10);
/*
it does not work no matter what you try here:
FALSE_DlyMA = MA(H - L, 10);
FALSE_DlyMA = MA(FALSE_DlyDiff, 10);
*/
TimeFrameRestore(inDaily);
//plot to compare false results to true.
Plot(TimeFrameExpand(FALSE_DlyDiff,inDaily), "FALSE Daily Diff", colorBlue,
styleDashed);
Plot(TRUE_DlyDiff, "TRUE Daily Diff", colorBlack, styleDots);
Plot(TimeFrameExpand(FALSE_DlyMA,inDaily), "FALSE Dly MA", colorRed,
styleDashed);
Plot(tfeDlyMA , "TRUE Dly MA", colorRed, styleLine);
_Section_End();
