-1

I was wondering if when calculating technical analysis indicators having more data points makes any difference. Let's take as an example a 9 periods exponential moving average (EMA). To calculate it we could use a price series with the latest 9 prices (or does EMA-9 require 10 prices?) but we have access to more data we could use a much longer price series with those same 9 prices as the last. Would the two series produce the same EMA curve? My previous example is somewhat of an edge or limit case but I suppose you understand where I want to go, my question is if having longer or shorter price series makes any difference.

2 Answers 2

3

When you calculate an (n) period EMA, you start off by calculating a simple moving average for the first (n) days and you begin exponential calculations on day (n) + 1.

In your example, it would be a 9 day SMA and exponential beginning on day 10. Therefore, a 10 day EMA with 10 days of data would NOT have the same value as a 10 day EMA on a much longer period series of data because the EMA on only 10 days of data would be a 9 period SMA and one day of EMA. And the more volatile the data, the greater the disparity.

Going forward, the differences would diminish as the effect of "exponentiating" enabling this difference to equalize. As an example, if you calculated a 10 day EMA on the last 6 months of stock XYZ and compared it with a 10 day EMA of the last 9 months of that same stock, the values would be identical.

All of this would be quite easy to observe to see if you set up a spreadsheet.

2
  • My question comes from the fact that I've never seen any requirement on length of the price series as input to indicators apart for parameters such as the period and also no discussion on how reliability of indicators is impacted by series with too few data points
    – noplace
    Commented Apr 30 at 16:09
  • Most indicators will vary in value if the number of data points varies. These will be those that reference previous data. Anything that is calculated solely on today's value will not have such variability. Commented Apr 30 at 19:22
0

Think of the data plotted on a graph. Adding more data points will make the graph "smoother". If you have only a few data points, the graph will be very jagged, jumping from point to point. As you add more data points, it's going to smooth out. If you had thousands of data points, it would be a continuous curve.

Does it make a difference to any analysis? Up to a point. With too few data points, it's difficult to know whether the indicator is jumping around wildly, or if you're just seeing the product of too few data points. Most especially if you're looking at the most recent data. If some indicator is up sharply, does that mean the company is having some "flash in the pan", some short-term factor affecting this month's earnings (or whatever metric) that will likely go away next month? Or is this a long uphill climb, that the company's fortunes have been steadily improving as they solved some problems or introduced a new product or whatever? With more data you could see the difference.

At some point more data just becomes superfluous. If you're trying to evaluate the long-term prospects of a company over a period of years, having one data point for each month for the past five years is probably plenty. Having a data point for each week is surely plenty. Having a data point for every hour of every day is just unnecessary and pointless. You're just bogging down your spreadsheet (or whatever) with redundant data.

5
  • Very few indicators are functional with monthly data. Offhand, the only one that I can think of that's not would be long term support and resistance. For short and intermediate support and resistance, you'd need data with much shorter time periods. Commented May 9 at 17:35
  • @BobBaerker I'm not sure what you mean. Most indicators are true at a point in time or a specified time range. Like, if you tell me the total market cap of a company, that's going to change in a few minutes the next time the price per share changes. So the question is how often we collect or calculate the metric. My point was, if I'm trying to draw a graph of a company's total market cap (say) over the past 5 years. having one data point for each year would make way too fuzzy a graph. Having a data point for every hour would be unnecessary overkill. ...
    – Jay
    Commented May 10 at 16:37
  • ... One data point per week would probably be about as much as would be useful. If you're trying to make a decision about which stock to buy today, having just one data point for each stock, the market cap (or whatever) last month, yeah, that would be of limited value.
    – Jay
    Commented May 10 at 16:38
  • This question is about moving averages. Any indicator involving moving averages will be more responsive to price change if the moving average(s) is shorter (with more whipsaws) and less responsive if the average(s) are longer. Furthermore, the nature of one's trading will dictate what is or is not superfluous or overkill. Day traders utilize very far more data points, sometimes as low as seconds whereas swing traders might utilize hourly or daily data. Commented May 10 at 19:45
  • @BobBaerker Yes, as I said, add more data points and the curve will be smoother. "nature of one's trading will dictate what is ... overkill" Certainly. Perhaps I didn't make this clear in my answer. As I said in my answer, if you're looking at 5 years of data, having a data point for every hour is overkill. I suppose I didn't clarify that if you're a day trader looking at three days of data, one data point per day is surely not enough, as that would be a total of 3 data points. The issue is basically total number of data points that are relevant to whatever it is you are trying to do.
    – Jay
    Commented May 11 at 4:29

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .