APBRmetrics

I developed a small trends-based search engine to find performance similarities between NBA basketball teams.

The user specifies the name of an NBA basketball team, and receives a list of similarly behaving other NBA teams. The similarity analysis is based on trends of performance from 1946 to 2009, and the performance measure is the wins/losses rate per year.

http://stockatowebsite.cloudapp.net/nbateams.html

Your opinions and suggestions are welcomed!

Uri K.

It doesn't open for me.
Can you explain what you mean by "trends-based search" ?

As far as I can tell, it's using time series correlation (similar to Google Trends). I can't get it to work either though.

Take a look at the image:



"trends-based search" or ("pattens-based search") is a search based on finding similar patterns, for eaxample, Trends2 is more similar to Trend1 than Trend3:

Trend1: 1,5,6,3,2
Trend2: 2,4,7,3,1
Trend3: 8,3,6,1,8

-U.

Is it anything more complicated than correlation between two time series? That's what it looks like to me.

High correlation is the result, not the search procedure itself.

It's like saying on Google - "is it anything more complicated than finding expressions that are similar to one another?" - basically, this is what it is, but it's more than that.

For a small amount of data simple, correlations might be a solution, but for many signals, which are also long (say thousands of observations per signal) you need more than using correlation coefficients.

-U.