I have an idea for a little project

[TheKeyboardDemon]

Hi guys,

I'm kind of new to data mining so really just getting used to the idea and wanted to take on my first little project. I was hoping for some advice on whether I am approaching things correctly or if my idea will uncover the results I have in mind.

The problem I'm trying to solve is looking at the performance results for computer graphics cards. I build performance computers as a hobby and want to see that I have selected the best performing components by using benchmark software. However most benchmark software only shows maximum and average frame rates, but some cards have issues which means that the have very bad minimum frame rates at the points where the issues manifest themselves.

I thought it might be interesting to create a log of a series of benchmark tests where I record the number of frames per second every 100ms and compare this to the results achieved using different graphics cards. The problem with that is that a more expensive graphics card would produce much higher figure than a cheaper card so I thought maybe the numbers should be normalised by using a log function to bring them in to an easier or more manageable range of values. Then once I have enough sets of data perform an analysis of variance test on the data sets to see if any cards are performing outside of the what is assumed to be normal performance or behaviour.

The thing that I am unsure of, is am I approaching this the right way? What are the likely pitfalls in such an approach? Is there a better way to do this?