Large Averages Question

[riffz]

So in an earlier thread I said I'm making a console cube timer program in C++. I'm wondering whether it would be more convienent to provide an average of all solves during that session like CCT does, or whether I should cut out the slowest and fastest times when averaging the entire session.

I guess I'm asking whether it's more helpful to see a mean of every solve or an average of every solve.

 

[rjohnson_8ball]

Personally, I like taking the latest 12 (or 5 or 22 or 102) solves, dropping the fastest and slowest in that set, then finding the average of the remaining 10 (or 3 or 20 or 100) solves. Using 102 might give the most accurate indication of someone's speed.

 

[qqwref]

I'm pretty sure the current timers use a mean of all non-DNF solves when calculating the session mean, so you should probably do that for consistency. But of course drop the best and worst for averages of 5/12/100.

 

[MichaelErskine]

I like to do sessions of 5 (weekly competition or WCA competition) or 12 (JNetCube default) but I save all times to make graphs with 3 plots: best times, worst times, and average of the remainder.