Ask Colin

Is it a clear divergence?

The detailed question was:

I'm looking at the weekly chart of the ASX All Ordinaries index on 4 August 2005. I use three momentum indicators on this chart with the standard settings - MACD, RSI, and Stochastic.

One of the things I try to look for is divergence, but I'm still learning the signals at this stage. What is concerning me is that the new high in the price chart is not being matched on the MACD or RSI. Both have a downward slope from the highs of March to the new high of last week.

The trend is still intact with my trend line from March 2003 touching in May 2004, August 2004 and twice in May 2005.

My concern is this momentum divergence seems to confirm your BRW article of this week regrading the Arms index and the 20-day moving average of Advances. I also noted that Dr Alexander Elder's last newsletter was mentioning the MACD divergence on the SP500.

How do you feel about the MACD and RSI? Is this a clear divergence? Or just me gazing at the charts for too long?

I am writing this answer in the last week of October. I am sorry it took so long to respond to your questions. There are two reasons for this. Firstly, I have been very busy. Secondly, as you will be aware, I must be careful not to comment on current markets without a licence to advise people. The comments I am making below are therefore not to be construed as advice, but seen only as a discussion of the methods referred to for educational purposes. Although I know in part the outcome of the situation you have asked about. I will do my best to respond in such a way that I am not using hindsight to drive the analysis.

First, I have two rather general points I would like to make:

Divergences
I find there is a problem with bearish divergences in an uptrending market. It is true that there tends to be a divergence before the trend ends. However, prior to that, there is often one or more divergence that did not presage the end of the trend, but only a correction in the trend. That is not to say they are not valuable and that we should not take notice of them. However, it can result in very premature exits if we act on the first divergence in an uptrend. Ideally, we should be using a trend following indicator to trade an uptrend. So, if we see a divergence, I would not act on it as a trend ending unless I also had a trend following signal or a clear price signal like a trend ending or completion of a reversal pattern.

Trend lines
A trend line is very unfortunately named (as is Relative Strength Index) because it suggests that the indicator is doing something it is not. A trend line is not a trend following indicator. It is a momentum indicator. A trend is defined by peaks and troughs. A trend line, measures the speed or momentum with which the trend is unfolding. So, if price hugs the trend line, the trend is very consistently moving at the slope of the line. If the price is moving up away from an upward sloping trend line, then the speed of the trend is increasing. Note this assumes you have the trend line drawn on a semi-log chart, otherwise it may be just a scale effect. If the price is falling towards or through, such a trend line, then it tells us the trend is slowing down or losing momentum. There will often be a divergence when this occurs – for the same reason. This can be forewarning of the end of a trend, because trends often slow down before they turn, but it is not the definitive end of the trend, by definition, when the price falls through the trend line. As with the other momentum indicators, I would take notice of a break in the trend line, but I would not act on it as a trend ending unless I also had a trend following signal or a clear price signal like a trend ending or completion of a reversal pattern.

Having said these things, it must be stressed that there are many ways to trade markets. If you are a swing trader, your trading plan may be to sell on divergences and trend line breaks and this could be a rewarding way to capture most of any one swing an a trend. However, if you are trying to stay in the whole trend, these rules would tend to throw you out prematurely. As in all these things, a question or answer is only relevant in terms of what you are trying to do. You have not told me, but it sounds as though you are implying you are looking to hold in corrections and through to the end of the bull market.

Looking at your trend line first, you are clearly drawing it on a linear chart. As I said before, I always draw trend lines on semi-log charts. In that case, trend line analysis has been very difficult. Successive trend lines have been broken forming the fan pattern described in the textbooks. That can be an important signal too – like divergences, fan lines often occur before a trend ending, but each fan line, or any particular one in the sequence, is not invariably followed by a trend ending. The rule for me is that I would only use a trend line on a linear chart if it had a very small dynamic price range. The range in question is 2600 to 4800 on my software. This is not clearly a small range, but nor is it clearly a big range. So it is arguable which scaling is appropriate. If in doubt, I err towards semi-log.

Because of these problems, I prefer an alternative approach. As you will know from my weekly columns in BRW, I do not use conventional trend lines. Instead I prefer to use regression lines with channel lines through the maximum excursions from the regression line. I use a semi-log chart unless the price range is very small on a short term chart. This is just a personal preference, but I prefer this method. It is an expression of the well known tendency to revert to the mean, except when the trend changes. It seems to keep me in major trends, with shorter term regression lines used to track each upswing. At the time of your email the index was quite close to the regression line. This was not a place where we might normally expect a correction or trend ending. There were some signs of a possible correction which I noted in BRW. This might have meant a correction back to the lower channel line. In the event we got more uptrend through to October – the first expectation.

Now, to consider your divergence on the weekly MACD. I am sorry, but I cannot see this divergence on the MACD on either the MACD line or the signal line.  I have reconstructed the chart as at 4 August and the MACD line is still rising above the rising signal line. There is a potential divergence (and see the next item in this newsletter), but no turn down to form a second peak after March, as you describe. However, close observation using the MACD Histogram shows that the weekly bar for the week to 5 August had ticked down. The reason this was not immediately apparent on the MACD chart was that it was caused by a rise in the Signal line faster than the MACD line rather than the more common situation where the MACD line turns down first.

2004 saw several premature consecutive divergences on the MACD. Likewise the MACD Histogram. This rather backs up my observation of lots of charts, in that all it meant was that the trend had slowed down for a while. What happened was that instead of leading to trend failure, it led to a period of much faster growth in the trend.

I do see the potential divergence forming on the RSI August versus March (you said standard setting, but I am not sure what that is. I assume you mean Welles Wilder’s original suggestion of 14 weeks?). However looking at the period since May, there is only a low grade divergence (equal peaks on RSI) June versus August.

Going back to the big picture on the RSI, the long term divergence in April 2004 versus September 2003 did not lead to a trend ending. Why would we worry about this one except that the bull market is older?

It is my experience that momentum oscillators are best used to enter uptrends on corrections, rather than to exit them on divergence – always assuming that we are trying to stay in the long trend. If we are only trading each swing, weekly divergences are probably much more important to tell you which divergence on the daily oscillator to act on.

Keyword: