Moving averages

I do hope you're joking?!?

Why?

Just because an indicator is 'lagging' in the sense that it is (in part) a function of past data, doesn't mean that it should be represented graphically in the same position as that past data. Today's 50 day MA belongs exactly where it tends to appear on the date axis - next to today's date! Obviously, it is a function of data from the past (1 to 50 days ago), but that doesn't mean that it's not 'current'. Obviously, any indicator needs to be understood by the user. As long as he understands that it is an average of the previous 50 prices, then it's not in the least misleading to place at by today's date.

Any why would you put it at today's date minus 25 days?!. For a start, that data point would mysteriously be 'predicting' in part, the 25 data points ahead of it.

why would you put it at today's date minus 25 days?!

That's the best place for it. Graphs and charts were invented to better illustrate the truth or give a feel for data than just a tabulation. I think Dr C has a point.

It reminds me of the old joke about the punter who out of the blue buys his wife that nice leather bag she's coveted on Carnaby St. Why, she asks. He says, I'm feeling generous as I'm in the middle of a winning streak. She asks, how do you know it's the middle?

Moving averages when properly used are simply a smoothing indicator.
They smooth out the data so that the direction is clear but only to the point charted half a span of the average back.

A 100 day moving average needs 100 days of data before the first point can be calculated.
That first point being the average of the 100 points. This needs to be placed exactly in the centre of that data.
We lose 50days at the beginning and 50 at the end give or take a point.
And we proceed from there.

As I said before, the dependency of this statistic on past data doesn't mean that it belongs on a graph other than at the current date. Everyone understands that TODAY's n-period moving average is an average of the previous n prices. There are several commonly-used price statistics (RSI for one) that are a partial function of past prices. Do you place these at their 'average'? That would be hugely confusing. What about exponential moving average? Do you place that at the expontentially weighted position in the past??

And imagine how you would use a long MA in your system. Suppose I'm using a month-long graph of prices, and I want to look at the 200 day MA. According to you, this can't even be represented on my graph (because it should be placed at 100 days ago!).

What about exponential moving average?

An exponential moving average tries to confirm a change in direction earlier by putting more emphasis on the nearer at hand data than it does on the earlier data in the calculation. Like a shorter moving average would.
It is still a lagging indicator.
It merely reduces the lag before the out of date trend is confirmed.   
Where it should be placed on a chart I do not know.

It's still pretty pathetic way of trying to predict where the price might be going next.

Suppose I'm using a month-long graph of prices, and I want to look at the 200 day MA. According to you, this can't even be represented on my graph (because it should be placed at 100 days ago!).

Absolutely,  but why would anyone want to know what the market was doing 100 days ago?  And possibly be misled into believing that it is still heading in the same direction when it might not be.

A glance at a bare chart would tell you more about the nearer to day action anyway.

The last point on a 100 day moving average is using data from up to 100 days ago.
It should no more be placed at the end of a chart than it should be placed at the begging.   
It is an average and the plotted line should be centred on the chart and follow the data.
By doing this it tells a tale of the data's true course, but only up to half a span of the average back in time, that is all it tells us.

Absolutely,  but why would anyone want to know what the market was doing 100 days ago?

But it doesn't just represent what the market was doing 100 days ago. It represents what it was doing from 1 to 200 days ago. Now you might not be interested in this, but a lot of people are, and they obviously want that piece of data to appear at the current date! Similarly, imagine you have a binary indicator that is TRUE if today's market is in the top 1% of the year's range, and FALSE otherwise. This will be partially determined by past prices, but the indicator belongs beside today's date.

If you just have a problem with lagging indicators, then that's fair enough. But if they are allowed, then they're current conventional placement is obviously correct. As I said earlier, if you place the n-day MA n/2 days earlier in the chart, then it is appearing n/2 days in advance of some of prices of which it is a function! If you put it there, it might as well not be on the chart (as my 200 day example illustrates).

The last point on a 100 day moving average is using data from up to 100 days ago.
It should no more be placed at the end of a chart than it should be placed at the begging.

And it's also using data from 1 day ago. This, fundamentally, is why it belongs by today's date: it can't be determined until today/yesterday.

To prove everything I say, and how generally misleading these moving averages can be when drawn incorrectly, I'll use a 50 day moving average and the FTSE100 chart to date to illustrate my point.

On the 16 October this year when the FTSE 100 changed direction, the 50 day moving average was heading south.
Yet it didn't change direction until 33 data points later on the  2 December and a rise of 547 points had been recorded by then.
That's nearly an 8.8% rise before the 50 day moving average responded with a change in direction.

When a moving average is drawn correctly it follows the  data, and visually indicates perfectly what I am saying which is.
A moving average is only relevant up to half a span of the average back in time and that is where the last data point of the moving average should be drawn.

So what is the point of drawing them incorrectly?

I give up.

My last post should have driven home the point that any moving average is only valid up to a point half of its span back along a chart.
So I can only think that people who routinely consult these moving averages when they are drawn incorrectly do not understand that point.

A moving average can never be more than a means of smoothing past data and must never be used as a guide to where the data was going beyond half its span back in time.

So why draw them incorrectly and visually lose that point?

If say there was a monthly average, ( whether temperature, or rainfall or economic data ), there would be little question. You would stick your data point in the middle of the month. It's a graph - it's meant to show meaning to something in a graphical way. Having to play a mental trick with yourself of shifting what you're looking at by half a month ( if the end point of the month is used ) detracts from that.

Fight!
Therefore, let the protagonists consult their charts: make a prediction therefrom and we'll see who wins this argument in 25 days time.
Chickens? Not these guys, Dr Crippen and Contrarian.
Over to you..

let the protagonists consult their charts: make a prediction therefrom

That would be difficult considering that I haven't offered any alternative to the moving average here.
Even contrarian who defended the way they are usually drawn hasn't come out and stated that he uses moving averages as a means of predicting the future direction of markets either.