I suspect to get accurate date you are going to need not so much a larger sample size but a much larger gap in spell speed between the two tests. I would test this on an astro where you can arrow yourself for 60 seconds (though if you didn't detect any change you might reason that arrow is bugged). Likewise Astro has a built in 5% spell speed bonus you can tinker with.

25 is just too small a number from what I have seen on spell speed increments to detect a change.