That's a big question. Have a look here. I got this off a Google search. There's other info there as well. http://www.investopedia.com/university/movingaverage/
If I'm using them I like Fibonacci numbers. I use 21 and 55 crossover. Also like to see where the 200 is.
I should think it would depend on what you propose to do with the information?
The smaller the sample size (lower the number of periods included) the less reliable the result in terms of trend but the more reflective in terms of recent movements.
I'm not sure that an MA is the most appropriate tool for a one day perspective and might suggest one MA of 5 periods would reflect the short term view, one of 10-20 the mid term view and one of 90 the longer.
The problem with specifying fixed periods is that they are not always appropriate for the share in question. If you are using the MA's as buy/sell signals, I suggest you need to do some back testing on the share of choice and find out how many false or marginal trade signals they generate.
In my limited experience, the one size fits all approach to MA's isn't as helpful as something you tweak to fit your style and the share in question.
Hope this helps
Not sure what you wish to do with emas,but here is something you might already know.
One of the biggest problems traders have is choosing indicators/system that works both in trending markets and sideways markets. Figuring which is which is not always blindingly obvious to begin with so it's useful to have some warning signs as to what sort of market might be ahead of you.
So put both a 100 and 50 day ema on your chart. For the moment don't fuss too much about the timeframe.
Nice daylight between them (be more specific and measure that with ATR if you wish) with the 100 under the 50 ema the market is trending up whatever 'wriggling' about you see going on with the price.
Reverse the order with the ema on top and the markets going down.
All pretty obvious.
However, when you see ever shrinking daylight between those indicators you have an advance warning to stand aside, or switch systems,because for at least awhile the trading range is going to get tighter than a ducks backside. That might only be for a day or two, or it might just go sideways for two months.
Some people would use a cross of these to maybe enter or exit the market. I prefer to use them to say I have not got a clue what will happen next other than there should be some narrowing of trading activity for an unknown period of time,hence stay out, or switch systems for that period. You could use an ATR filter which says when the value between 100 and 50 ema is below X * ATR I will not take new position and obviously the converse holds true.Alternatively, when ATR value hits a certain point you will only use system B which is for trading ranges.
If in doubt stick with a trending system and find a chart showing plenty of daylight between the 100/50 ema. Use other emas if you feel they show you a better picture.
Nothing blindingly great about this suggestion,but it's just another one for the toolbox to try to keep the probabilities with you.
Use it ,or chuck it.
Maybe I've missed something that's obvious to you but I'm still not entirely sure what it is you want the Moving averages to tell you in each situation.
Looks like some of the others are puzzled too.
Maybe if you were to give a little more background on how you trade and what sort of signal you would like the MA's to give you?
If you want to use crossover for entries/exits then as someone already said, it's how many signals you want in a day. The shorter your 1st MA, the nearer the price it is and the more signals you'd get. Try some numbers out and see what different results you get.
How many times per session/day do you want to trade. How long a time frame do you intend to run a position. I don't rely on MAs but still leave a 21/55 on a 1min running, on one set up, for interest. For me it helps to reinforce the picture. Holds my hand when I'm in a position. A chart is a picture of the market. What it's done before, is it likely to happen again? Indicators lag. If I'm idecisive or changing my mind on a trade, then the MA might help the picture.
If I understand you right, I think you need to look at questions of sensitivity.
Find a chart package that allows you to set your own MA periods.
Select the currency or whatever that you want to trade.
Select the trigger signal; eg price crosses MA or MA crosses price or MA long crosses MA short.
Run it over say three months and see how many buy and sell signals it generates.
Subtract the costs of your trades etc from the margin you've made and ask yourself if you traded proftably over the period.
Change the MA's to reduce or increase the number of signals and apply the same test.
When you get to an optimum between number of trades and profit generated, that's the optimised MA.
Back test it again over longer timeframes and see if it still works.
Obviously, this is very simplified but I think its the direction you want to take.
Hope this helps. :?: