Regression To The Mean

This is a phenomena in statistics. If one sample collected is at an extreme, the next sample will generally be moderate.

Example: Let's say you are counting how many words you are writing a day, and the average is 500 words per day. One day you wrote 800 words. Next day its likely to be closer to 500. This works in the other direction too - if you wrote just 200 one day, next day its again likely to be near 500.

This is a critical thing to know if you are trying to test different methods to improve performance. Do not judge a tested method with just few sample data points(eg. I wrote 900 words the day I used a different editor to write) - this is the Insensitivity to sample size. A bigger problem is that once we notice such a pattern, the Narrative Fallacy comes into play - and we become extremely convinced about our theory(eg. The new editor has X and Y feature - so it makes writing faster).

Meta