posted 11-04-2005 02:15 AM         

I've been experimenting with Heiken-Ashi averaging. Specifically, I've tried averaging H-A OHLC to get a second level H-A, an "average of an average", if you like. Then I repeat the process, to get third, fourth, fifth, … level averages. I find that the pattern stabilizes completely when something like the 12th level is reached, i.e. after that the averaging process has no effect.

Levels 2,3,4 can further eliminate false signals (candles with short bodies whose body color opposes the trend can undergo a color change), but my conclusion is that the "lag" introduced (as candles at the start and end of trends can change color also) more than offsets any benefit in the elimination of false signals.

So it would appear that level 1 H-A offers the most effective compromise between lag and smoothness. As the chart in your link illustrates.

This is all fairly abstract – hope it makes sense!

posted 11-04-2005 09:06 AM         

In order to do that, I have written a simple program to place the color reversals as buy/sell signals on the graph of an equity in order to do back tests to determine the profits based on H/A reversals. There are few parameters to 'optimize,' but I had the same results-that the short candles could be ignored, but with more delays in the buy/sell signals-reducing profitability.

Perhaps an eyeball determination from the real time H/A chart that a short candle is forming, and delaying or ignoring a buy/signal when a color reversal occurs, might be helpful? I'm still experimenting on this in my trial & error non-scientific manner.

On a H/A graph, it is easy to see the progression of the candle for a buy/sell signal by observing whether the last H/A price is moving above or below the H/A open-looking for a reversal. Might be a help in making the more profitable "buy at typical price" rather than the "buy on the next open"-as well as making ignoring these short candles easier?

David, I had to send my computer back to HP for in warranty service & they erased everything on it(despite my request not to) and I have lost your email address. I would appreciate your again sending me an email at fep@usa.com, if you are so inclined.

Thanks,

Frank