I am interested in a pcf for testing moving averages by computing

the net gain loss (over 500 periods) of a system that buys when the heiken ashi open is above the moving average

and selling when heiken ashi open is below the moving average.

TC2000 does not have any built in backtesting features and I do not know of any way to write such a formula.

Thank you.

Oh. I am wondering if it may be possible to get it by computing the average

buying prices versus the  average selling prices.

What is the period and average type of the moving average? Is that moving average applied to the HA Open, the HA Close or something else?

For example, it would be avgc10, where c is normal (not HA) close.

The moving average crossing up through the HA open.

SUM(IIF(XUP(AVGC10, (XAVGO3.1 + XAVGH3.1 + XAVGL3.1 + XAVGC3.1) / 4), C, 0), 489) / CountTrue(XUP(AVGC10, (XAVGO3.1 + XAVGH3.1 + XAVGL3.1 + XAVGC3.1) / 4), 489)

The moving average crossing down through the HA open.

SUM(IIF(XDOWN(AVGC10, (XAVGO3.1 + XAVGH3.1 + XAVGL3.1 + XAVGC3.1) / 4), C, 0), 489) / CountTrue(XDOWN(AVGC10, (XAVGO3.1 + XAVGH3.1 + XAVGL3.1 + XAVGC3.1) / 4), 489)

Thank you.  I am only interested in comparing to the value of

the moving average at the beginning of the period (i.e C.0=O.0)  - it seems that the above

may occur intraperiod(?).

No, it is only testing at the crossups and crossdowns. Why do you think otherwise?

If you change all instances of 489 to something lower (that matches) you can see the points at which the averages change.

The formula looks correct. But it is updating more often than I expected (count increasing).

How can I send you an example? 

Keep in mind if you change the period to something lower, the value will change both when a new price gets added and when an older price that is no longer in the period gets dropped.

Please send it to my attention at:

support@tc2000.com

Thank you. Actually it looks good now...  There were simply more crossovers than

I had expected to see. But a minor problem occurs when there is a tie between the ha open

and moving average.  I would only want  to count  such ties as having the previous sign.

(otherwise a bounce produces a buy and sell - but one certainly can live with that).

Well the numbers probably aren't actually tied in the literal sense if you take it out to enough digits.

Not sure if we would want to try to figure out a way to count stuff that is just close as a "tie". We would need to add a lookback of some sort to see which side the tie should be on and make sure it actually does count the cross it eventually "unties". That would reduce the number of bars we could check.