# How can I determine best performing groups in a series

Not sure if this is right place for this question, but taking an example of the following trend, how can I determine that groups A and B are the areas where the result has been best?

I have already tried by doing an average and finding all those above the average. But finding out a group and driving a conclusion is tricky part. How can I do this?

Not sure to understand what you are asking. Your purpose is to find the only one best group ? From the picture you provided, two things are unclear to me:

• A and B areas don't have the same width, how are you defining your width ?

• A and B are covering areas where the signal looks exactly the same. Is it that for the sake of the example you copy-pasted some fake data or does it have to do with what you are finally trying to do ?

If your width is fixed then you should be able to compute a series of local averages and pick out the best one.

• (1) The widths can be different (2) Yes signals can be different. It's just for the sake of example as I just copy-pasted same graph twice. Just to show there can be several groups that look better than the rest of the series. – arxoft Nov 21 '19 at 7:01
• Then I'm still not getting what you are trying to do. If the size can be different then there can be an infinite number of frames overlapping one another and for which the average is above a certain number. What prevent you from selecting a section of width almost 0? – Jeanba Nov 21 '19 at 7:09
• That's the challenge actually. How can I determine the window size? Size can be variable. – arxoft Nov 25 '19 at 13:45

From a particular set get local maxima & local minima, then compare it to the average & global maxima & global minima you have. If those values are better than your average then that can be a right section for you,