it's also misleading in the sense that there is no "one number". You can't say something increased in performance after just one run. You need a few to know t

it's also misleading in the sense that there is no "one number". You can't say something increased in performance after just one run. You need a few to know that it's consistent, but at least 200 runs to be statistically sound, and you're not gonna get exactly 40x on each of those runs. Sometimes it will be 40x, sometimes 35x, sometimes 1.5x. What the article doesn't mention is, was that the highest measured performance increment, or was it the mean or median or most common? Also, what was the margin of error in that reading?