posted 09-21-2005 06:20 PM         

Like you, I have found H-A candles to be a useful summary technique.

However, you might be interested in the following article: http://www.ensignsoftware.com/tips/tradingtips57.htm

I think Mr Arrington is being a little harsh with his criticism. However, I believe that his point that you should not interpret H-A candles (especially the wicks) in the way that you would interpret standard candles, is nonetheless valid. It again illustrates the importance of understanding the underlying method behind the tools that you're using.

In researching H-A, I also found this, although I note that refers mainly to day-trading: http://home.cinci.rr.com/upandup/

If you're interested, the calculation behind the T3 can be found here: http://paritech.com/paritech-site/education/technical/custom/indicators/98jan2.asp

And the StochRSI here: http://www.stockcharts.com/education/IndicatorAnalysis/indic_stochRSI.html

posted 09-22-2005 08:52 AM         

Nice to hear from you again, as always. I completely agree with your comments on the H/A candles. They obviously average out the actual data in order to provide a more consistent trend summary. I use them as a stand alone indication along with other indicators. For all other analysis, I do not use them-in deference to a bar or regular candle chart that indicates actual data. I do find the daily & weekly H/A candles as a very useful tool for my short/medium term work with QQQQ & IWM. I do appreciate all of the useful info that you gave, and am going to carefully read it.

The broad market in the USA is in a real mess right now. Head & shoulder & trendline breakdowns, etc. Our use of limited petroleum resources needs to be reigned in, but we are really hooked & it will be painful, I'm afraid, and will adversely impact the market. Puts & shorts are doing well, tho. We are waiting to see how bad the effects of "Rita" will be.

Just about everything I look at concerning the USA market is down right now.

My best to you & yours, and thanks again for your outstanding B/C efforts and help....

posted 09-27-2005 11:44 PM         

The best information I have found on Heiken-Ashi candlesticks is here: http://www.ensignsoftware.com/tips/tradingtips57.htm

Their formula is as follows:

Heikin Ashi close: haClose = ( O + H + L + C ) / 4
Heikin Ashi open: haOpen = ( yesterday's haOpen + yesterday's haClose ) / 2
Heikin Ashi high: haHigh = the higher of H, and today's haOpen
Heikin Ashi low: haLow = the lower of L, and today's haOpen

where O, H, L, C are today's opening, highest, lowest and closing prices.

Another option you might like to consider, although it perhaps needs some preliminary research, is to replace your standard (linear) moving averages with non-linear versions. This is ESPECIALLY RELEVANT to BC, as most of the buy/sell signals are generated by the indicator (RSI, MFI, OBV, etc) crossing through a moving average of itself.

Standard MAs have the following problem: the shorter the period of the MA (e.g. 5 days), the more receptive it is to undesirable "noise", i.e. minor fluctuations that cause false signals. But if one chooses a MA with a longer period (e.g. 25 days), the result is a smoother curve with less noise, but the problem is that a "lag" factor is introduced, i.e. the MA trails the price to such an extent that, by the time the price crosses it, much of the trend has already been consumed, i.e. the buy/sell signal is too late to catch the trend. In other words, choosing the optimum value for the MA is partly concerned with finding a balance that minimizes "lag" and maximizes "smoothness". This applies to MAs whose underlying math is LINEAR.

One aspect that I noticed with BC is that some of the indicators zigzag across their MAs several times during back-testing, with the end result that a high number of trades is generated. Many of these trades are unprofitable.

Latest research shows that a NON-LINEAR MA can reduce lag without compromising smoothness. I am guessing here, but I think that using non-linear MAs might turn BC into a more profitable system.

**** NOTE: I can not promise this; what you would need to do would be to test it. ****

One of the best non-proprietary non-linear MA systems is the T3 (devised by Tim Tillson). The formula is as follows:

Inputs: (1) period length; (2) phase (default value = 0.618, although you may like to experiment)

e1 = EMA of indicator value over the period length
e2 = EMA of e1 over the period length
e3 = EMA of e2 over the period length
e4 = EMA of e3 over the period length
e5 = EMA of e4 over the period length
e6 = EMA of e5 over the period length
c1 = –phase*phase*phase;
c2 = 3*phase*phase + 3*phase*phase*phase;
c3 = –6*phase*phase – 3*phase – 3*phase*phase*phase;
c4 = 1 + 3*phase + phase*phase*phase + 3*phase*phase;
Result = c1*e6 + c2*e5 + c3*e4 + c4*e3;

e.g. if we want to do a 5-day T3 of RSI using a phase of 0.618:

e1 = 5 day EMA of the RSI value
e2 = 5 day EMA of the e1 value
e3 = 5 day EMA of the e2 value
e4 = 5 day EMA of the e3 value
e5 = 5 day EMA of the e4 value
e6 = 5 day EMA of the e5 value
c1 = – 0.618 * 0.618 * 0.618 = –0.23603
c2 = 3 * 0.618 * 0.618 + 3 * 0.618 * 0.618 * 0.618 = 1.853859
c3 = –6 * 0.618 * 0.618 – 3 * 0.618 – 3 * 0.618 * 0.618 * 0.618 = –4.853631
c4 = 1 + 3 * 0.618 + 0.618 * 0.618 * 0.618 + 3 * 0.618 * 0.618 = 4.235801
Result = –0.23603 * e6 + 1.853859 * e5 – 4.853631 * e4 + 4.235801 * e3

From what I have read about non-linear MAs, I suspect that they might enhance BC's back-testing profitability considerably, especially given that BC is heavily dependent on indicators crossing their MAs to generate signals. But you would need to try it, and I can't promise that the end result will justify your research time and effort.

posted 10-09-2005 08:53 PM         

Following my post above, here are some more ideas for your consideration:

USES FOR TILLSON T3

1. As explained in the post above, where a BC buy/sell signal is generated by an indicator crossing over its own MA, use the T3 instead of the MA. Set the phase parameter at 0.618 (or experiment with other variations), and vary the period length in the same way that BC tests for optimum profitability by varying the period length of the MA.

2. Instead of passing the closing price into the indicator formula, consider using a smoothed value by passing the price through the T3, and then the resulting value into the indicator formula. For example, instead of using RSI(Close,14), use RSI(T3(Close,3,0.618),14).

3. Where there is averaging taking place inside an indicator formula, use T3 instead of the averaging process.

You could try any combination, or indeed all three, of the above. All of these should smooth out "noise" from BC's calculations, making for fewer and "cleaner" crossovers (signals). This could also help alleviate the "problem" (described by "Johnny" in http://www.stock-anal.com/ubb/Forum1/HTML/000901.html) where the signals change every day, due to the rolling data window on which the back-testing/AIO is performed. Cleaner, noise-reduced curves could filter out many "false" signals caused by ripples in the data, and improve overall consistency. As I said in my previous post, you will probably need to experiment to find settings that deliver the optimum result.

HULL MOVING AVERAGE

As an alternative to the T3 formula, you might also like to experiment with the Alan Hull's moving average (HMA). Here is the formula:

Input value is the period length ("period"); WMA is a weighted moving average; "price" is the value being passed to the formula:

WMA(2* (WMA(price,period/2) ) – WMA(price,period),int(sqrt(period)));

Example: assume a period of 10 days, then:

a) take a 5-day WMA of the price [note: 5 days is half the period length], and multiply the result by 2
b) take a 10-day WMA of the price, and subtract this from the result of (a)
c) take a 3-day WMA of the result of (b) [note: 3 days is the square root of the period length, to the nearest integer]

The HMA can be used instead of the T3 (see uses section above).

OTHER NON-LINEAR APPROACHES

Using indicators based on non-linear math is becoming state-of-the-art Technical Analysis. Have a look at each of the following sites. The indicator formulae are (as far as I can tell) proprietary, but you may be able to get some ideas for your own development:

a) http://www.wave59.com/technical.asp
(click on the "Free Downloads" and "Weekly Showcase" buttons for details)

b) http://www.jurikres.com/catalog/catalog.htm#top

c) http://www.fractalsedge.com/howitworks.html
(download their free trading course to see how it works)

Note: I have not used any of these products. I do not endorse them. I include this only so that you might be able to use some of their concepts and methodology in BC's future development.

MARKETING IDEA

In the same way that you changed the name from Stock Analyzer to Best Charts, you may want to use the non-linear approach to develop yet another "new" product (while still leaving the other features of BC, e.g. portfolios etc) intact.

Cheers
David