S&P/ASX 200 (ASX:XJO) - XJO Weekend Charting and Chat - 16th Feb 2018, 31179400, page-27 - HotCopper | ASX Share Prices, Stock Market & Share Trading Forum

Originally posted by redbacka ↑

"Adam Grimes' article was good, and if the 200 MA has become a poor predictor of anything over the last 30 years, then we need to be also skeptical of many other TA concepts, such as Standard Error Channels."

48 - that's a huge leap in logic and I don't follow how you can make a claim leaping from a critic of the 200-Day MA to suggesting that we should be skeptical of SECs.

The use of those two concepts are totally different.

The 200-Day MA has no basis as far as I know in statistical analysis, except ex-post facto. It's become a meme in Technical Analysis which, understandably, should come under scrutiny.

The Standard Error Channel is a completely different kettle of fish.

It seems to me that you have singled out SECs as a target for a spray gun at TA concepts - only because I use it.

How many other analysts do you know that regularly quote the SEC? Frankly, other than myself, I don't know of any.

The SEC is a quantitative concept used as a statistical device because it utilises standard statistical knowledge. Its basis has nothing to do with the stock market per se.

It relies on the notion that a distribution will "normally" expand and contract about two standard deviations from a regression line. That's a basic statical concept.

Once a population extends above 2STDs above a regression line - we tend to get a reversion to the mean - and often an overshoot.

That's a completely different sort of concept from the idea that the 200-Day MA is a market between bull and bear markets.

Yet you want to claim, based on your post, that the SEC is also to be suspect because the 200-Day MA is also suspect.

Do you have any evidence for that?

If you do you are going up against the weight of evidence from the extraordinarily deep knowledge of the statistics community.

If you want to make a claim that reliance on the SEC is flawed because of fat-tail events in the stock market - then you have some sort of claim to a criticism.

But - fat-tail events, more or less, confirm the notion of normal distribution and regression-to-the-mean that we get from normative statistics. But the fact that fat-tail events are rare, more-or-less confirms our faith in normal distributions in our markets - except in highly unusual events.

Perhaps - you would like to comment on why you criticise SECs because 200-Day MAs are criticised in the Grimes article. They are completely different concepts. And I fail to see the logical connection.

RB

So it is hardly a concept that

Expand

Hi RB,
Firstly I would like to say that my posting to which you have responded with some degree of upset was never meant as a spray into your very informative charting and TA.

Now my point is that traders will use a myriad of tools and techniques to try to interpret patterns of price movements to attempt to gain an insight into the next probable movements in price or an index. It does not have to be a method with a grounding in classical statistics, only that other technical analysts have proposed a method or even just a specific line of data points that can provide a good probability of predicting some particular outcome.

My next point is that there is considerable danger in assuming that a "normal" distribution which might be a best fit in describing the variance of a characteristic in a human population can be adapted to variance in a stock market index with a similar degree of confidence.
The Standard Error Channel I believe has an upper and lower line approx 2 standard deviations away from the mid line, the linear regression line. Please correct me if that is not very close to the fact. So 95% of price movements should fit inside the channel according to definition. I assume classical statistics apply here.

A trade initiated when the price touches an upper or lower line that assumes reversion to the mean should be successful the large majority of times, so going against the trend in the immediately preceding time period.
In fact I doubt that this is a reliable trading plan, and an equal weight of opinion would say to bet with the trend continuing, which sees the price move outside the upper or lower band, a lot more frequently than 5% of occasions.

So I again reiterate that I am skeptical that the "normal" distribution is a valid model for a stock market index, and if I am incorrect, it would be a relatively simple matter to place trades that have a high success rate and amass large profits in a fairly short time.
Real world experience indicates that large profits are much more difficult to attain.