Experiments in Technical Analysis

CreditViolet · #**941** 3rd May 2007, 11:14 PM

Quote:

Originally Posted by oxusmorouz

Hi CV,
If you don't mind, can you please elaborate on step II (Determine type of targeted behavior, data processing and mining, 1st level testing and end result).

Obliged,
Oxymoron

Sure Oxy, will do laters probably on the weekend, it would all be quite lengthy to explain. I was trying to do the same on the blog, bcos of the time issue never managed to post it.

Rgds

CreditViolet · #**942** 3rd May 2007, 11:17 PM

Quote:

Originally Posted by uasish

CV,

Cant read properly,plz give another link (with magnified version).

Asish

The attachment limit of 100kb forced me to reduce image quality. Try this, a bit more magnified - http://img369.imageshack.us/img369/2995/strategywp5.png

uasish · #**943** 4th May 2007, 12:36 AM

Thks.Now my week end is jeopardized.Have to try to grasp this concept of data more fluidly .

kkseal · #**944** 4th May 2007, 01:16 AM

Quote:

Originally Posted by CreditViolet

The problem is indeed the concept. A MACD tells nothing but the difference b/w two MAs.It has nothing to do with trend at all, sure indicators can be useful but then its more about 'reading' it than anything else. A good example is Chick Goslin who uses the 3-10-16 oscillator beautifully.

Rgds

In fact going by its construction the MACD should actually be categorized as a Momentum (rather than trend) indicator; 'coz it measures the difference between two time denominated entities (EMAs) - which is a momentum characterestic. Now if momentum is what you want to gauge (specially on your trading timeframe) there are other better (less lagging) options.

I find the MAs themselves to be better indicators of the trend (also more visually depictive) as kind of proxy trendlines.

Regards,
Kalyan.

kkseal · #**945** 4th May 2007, 02:12 AM

Quote:

Originally Posted by CreditViolet

Ok, so a stationary series is one whose statistical properties are constant along its length. Its quite well documented that financial time-series are heavy-tailed and not stationary.Volatility clustering is another important topic, i.e tendency of autocorrelation in volatility.Given all these different conditions, I think its absolutely treacherous to try and detect signals in all this noise.Theres just too much randomness to deal with here than most people think and most are 'getting' fooled by it.

Very true. I have a method that i'd say is doing ok, but still i'm not sure if the premises are right. (And what is the most basic premise? The price-time data series on which everything else is based) I can sense this with my hesitation with stop-losses. It's not just indiscipline, actually i'm afraid. And what do i fear? The randomness. Even if i have identified the dominant cycle correctly, the randomness can still get me, stop me out prematurely That is what i'm afraid of. So i just let it play around a bit as long as the dominant trend is intact. However not using a SL also leaves me insecure and vulnerable. Who knows that buy signal itself might have been one the games that this randomness plays! So i try to do the best i can by taking a mean of the trend and subtracting (or adding, for shorts) a certain multiple of the average (short-term) volatility. (And don't hesitate to veer from this scheme if a more logical stop-level is nearby. Even for entry-stops i think one should go by the logical levels and risk-reward estimates rather than any fixed %. Randomness doesn't count percentages)

Regards,
Kalyan.

kkseal · #**946** 4th May 2007, 02:30 AM

Today (inspired by earlier posts made by CV, Karthik & Anant) i was reading about the ZEMA - the zero-lag EMA. Ehler's method is based on analogy of a shooter shooting a non-stationary target and a method that uses not just the price but also the ROC of price. Even with all that what do we get? A MA crossover system perhaps that gives earlier signals, but does that solve everything? Nope. 'Coz those MAs are still based on fixed time cycles (which i don't think is constant at all) In order to get over that problem one needs to make those MAs adaptive. How? There can be two ways (or more, 2 is what i can envisage)

Either base them on a more 'transformed' noise-filtered view of the data (Ehler's MAMA & FRAMA) OR
Make them adaptive to the predominant component of the randomness itself - voatility (Chande's VIDYA) (so that they kind of 'factor in' the randomness)

And MA is just one kind of indicator, there are many more to deal with.

Alternatively, there can be another approach and that is to avoid the noise ridden price data completely for basing ones entry/exit decisions. To use something that has a high correlation with but not directly based on the price data series. Like CVs sentiment indicator.
Now this may be ok if one is trading the index (i.e. the mkts directly), but what when one is trading stocks? There i think the only way is one of cross-verification. Not relying on any one signal but using a convergence of multiple - what may be called the 'cluster' approach

Any better ideas folks for poor stock traders like myself?

Regards,
Kalyan.

kkseal · #**947** 4th May 2007, 02:41 AM

Science is the refusal to believe on the basis of hope

- C.P.Snow (don't know who he was but liked the quote)

karthikmarar · #**948** 4th May 2007, 03:05 AM

Quote:

Originally Posted by CreditViolet

The attachment limit of 100kb forced me to reduce image quality. Try this, a bit more magnified - http://img369.imageshack.us/img369/2995/strategywp5.png

Jesse, my friend...Too many + signs in the chart...Hope you will post the chart with all - signs..

Oilman too on this thread... Looks like the thread is becoming elite..

Of course the discussion is getting more fascinating...

regards

Karthik

sujitbasu · #**949** 4th May 2007, 06:18 AM

HI GUYS GOT A GOOD LINK WHERE YOU CAN LEARN TECHNICAL ANALYSIS FOR FREE

"http://www.incademy.com/pages/home.htm?ginPtrCode=10002"

sujitbasu · #**950** 4th May 2007, 06:26 AM

Hi,

CreditViolet I am a gr8 fan of yours and wouldlove to learn trading from you. Our forum is lucky to have ppl like you who is always willing to help

bravo

Similar Threads for: Experiments in Technical Analysis
Thread	Thread Starter	Forum	Replies	Last Post
Technical Analysis Course	snowy	Technical Analysis	25	29th December 2007 10:19 PM
Technical analysis course	sh50	Seminars & Tutors	18	10th June 2007 03:46 PM
Hello everybody.Technical Analysis	vijay chauhan	Introductions	2	30th May 2007 03:09 PM
Technical Analysis	crosscurrency	Forex	0	30th October 2005 11:00 PM