How much data should one analyze? 2

In a comment to this week’s Outlook post a Sentient Trader user asked two questions, the answers to which deserve their own post, hence this quick (and hopefully brief) post.

Here is the comment from Jeffrey:


I analyze the markets daily and this week I came upon something interesting. The cluster of circle and whiskers for the 40 day trough seemed to move from Tuesday when it was indicating an approximate October 4th low and then it seemed to have shifted the next day (Wednesday) into the September 24th-28th time frame. Then on Thursday shifted back into the October 4th area. Now after Friday’s close it shifted back again into late September. Is this usual? I have been using the software for several weeks and that was the first occurance I saw that happening.

I have another question as well. What is the best amount of data to use to form ones analysis. I am using Metastock and asked for data back to 1997. Is there a best point to use because I notice things shift around a bit if I pick different troughs to pin the analysis off of. Should I do a certain amount of data based on the cycle length? The book I am reading by Chris Grafton seems to suggest using 200-250 points of data but I was not comfortable with that when I saw the results. You will be getting quite a bit of participation from me here because I was about to write asking where I could post questions about the weekly analysis. Just want to say that this software is awesome and I don’t even know 1/100th of what it can do. Thank you!

Shifting Troughs

The first question asks whether it is usual for the future position of a trough to shift back and forth between two dates. I wouldn’t describe it as usual, but it does sometimes happen, and I think that it is very important to understand why it happens so that despite the shifting forecast you can still make sensible trading decisions.

First of all: why does it happen? The forecast position for a trough is the direct result of the phasing analysis that has been performed, and is calculated by simply projecting forward from the most recent trough of a cycle the recent average wavelength of that cycle. When a phasing analysis changes two things will happen: the position of a trough will change, and the recent average wavelength will probably change as well. And so the reason why the projected future position of a trough (the circle-and-whiskers) changes is because the phasing analysis has changed, in particular because the position of the most recent trough of a particular cycle has been moved.

Most usually the positioning a trough remains constant once it has been placed by Sentient Trader, but sometimes a trough will shift as Sentient Trader changes its mind about where that trough should be (I know computers don’t have minds, but I cannot help wondering …). When a trough shifts it usually stays there, but there are occasions when a trough will shift back and forth as you describe. In fact while this may be disconcerting at first, once you understand why it is happening it is an opportunity to have greater confidence in the trading that you are doing on the basis of the analysis, which I will explain in a moment.

One of the things that Sentient Trader uses to perform the phasing analysis is the developing cycle shape, and if that cycle shape changes then it will cause a phasing analysis shift. Sometimes the market can be deceptive, and the cycle shape will switch back and forth, causing the phasing analysis to do exactly the same. This is what happened this week: on Thursday the leap upwards looked to be creating an initial bullish shape from the trough that I went on about in my last post. But then on Friday the bullish move fizzled out, and the cycle shape does not look bullish at all.

That is why the phasing analysis changed back and forth this week, and it is why the future position of the 40-day cycle trough shifted back and forth. But I mentioned using this to your advantage. How can you do that?

The reason the trough position is shifting is uncertainty in the analysis. The moment I read your comment I knew why it was: On Friday 24 August 2012 a trough was formed at 1395.25 in the S&P 500 futures, then on Tuesday 4 September 2012  a trough formed at 1394.50. Those two troughs are only 11 days apart, and one of them (or a trough in between them) is probably the trough of an 80-day cycle (according to the analysis we are considering here). Which of those troughs is the correct one to phase as the 80-day cycle trough?

The answer is that we cannot give an answer at the moment! Purists might say that the Tuesday 4 September 2012 trough is the one because it is lower, but both troughs occurred on a weekend (Friday obviously the day before a weekend, Tuesday was the first day of trading of the week following a holiday on Monday 3 September), and so the actual low prices could have occurred over the weekend, at a level that we cannot see on our charts (if this surprises you then remember that one of Hurst’s basic principles is that cycles affect prices continuously, not only when we are trading. Watch the fourth video here). One might also argue that the trough should immediately precede the strong move up, but this is not always the case.

So if we cannot give the answer how can we possibly use this to our benefit: it is actually very simple. The moment you see this shifting trough phenomenon, create an extra chart in your Sentient Trader workspace, and pin the questionable trough onto the trough that Sentient Trader was using previously. Here is the result:

Trough on 24 August 2012

And here is the other option, which has the default analysis which now has the trough in September:

Trough on 4 September 2012

That takes a few minutes to do, and now you have two charts which present viable analyses, instead of one chart that switches back and forth. As time passes either both charts will be showing the same analysis as that turns out to the true analysis, or the chart with the pinned trough stops making sense and Sentient Trader will refuse to position the trough where you pinned it. At that point you remove the chart from your workspace.

How does this help you make more certain trading decisions? Because you only make trading decisions that are supported by both charts. If a trading decision isn’t supported by both charts, then don’t make the trade! A missed opportunity is not a losing trade after all, and there will be some times when you cannot make decisions because of the uncertainty in the analysis.

How much data should one analyze?

This is a question that we often get and the answer is actually fairly straightforward. There is a direct correlation between the amount of data that you provide to Sentient Trader, and the amount of information that Sentient Trader provides you with in return. Therefore the best way to answer this question is to ask how much information you want.

When trading a particular cycle you must have information for at least two cycles longer than your trading cycle. Personally I like to work with five cycles longer than my trading cycle, and so when trading the 40-day cycle I want to know when the last 54-month cycle occurred. When trading the 20-day cycle I need to have information about the 18-month cycle.

Now you know how much information you need, how much data does that equate to? Sentient Trader will position a trough for a cycle if there are at least two full cycle wavelengths worth of data, and so:

  • If trading the 40-day cycle we want information about the 54-month cycle. That means that we need 9 years worth of data (with 9 years of data Sentient Trader will include the 54-month cycle in the analysis)
  • If trading the 20-day cycle we want information about the 18-month cycle, therefore we need at least 36 months of data.

There is one other consideration: bear in mind the principle of commonality which tells us that all markets move with a great deal in common and make sure that you are able to position the analysis start date before an important trough. All world stock markets had major troughs in 2002/2003 and again in 2009. Therefore make sure you have enough data to include one of those major troughs. Or if you are looking at very long cycles then include prominent troughs at those times.

That is why I usually load daily data going back to early 2002. That includes the 2002/2003 trough, and gives me the required 9 years to find the 54-month cycle trough.

I would agree with you that 250 bars is not enough. That is about a year of trading, and so the longest cycle that Sentient Trader will position would be the 20-week cycle.

I have used Sentient Trader to analyze the Dow going back to 1898, and I found that in fact it made little difference having the extra 34 years of data prior to the 1932 low. Bear in mind that Sentient Trader places emphasis on the most recent data, because that is more important for making trading decisions now. The length of the 40-day cycle in 1898 might be a fascinating thing to know, but it isn’t going to help us trade that cycle more profitably today. Sometimes one can have too much data!


About David Hickson

I have been trading for over 20 years, but only had any success after discovering Hurst's cyclic principles. Unable to find any software to speed up the analysis process I created Sentient Trader software, which now pretty much does all the analysis for me. I am a film maker and a TV director, but nowadays I mostly provide consultation services to professional traders and fund managers, helping them to integrate Hurst analysis into their trading. I'm South African and live with my family in Italy.

Leave a Comment

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>

2 thoughts on “How much data should one analyze?