- Joined
- Jul 23, 2008

- Messages
- 496

- Likes
- 1

- Location
- California, USA

- WCA
- 2008YOUN02

- YouTube
- notkevin888

Not sure where best to put this, but I've been idly thinking about the "most balanced" cuber for a while, based on some comment that someone made about their KinchRanks standard deviation (alas, I can't seem to find it now T__T). But now that I wanted to calculate this, I'm having trouble justifying exactly what it should look like.

So, the naive approach is to simply take the standard deviation of the per-event values - but then cubers who only do a couple events and are somewhat slow are greatly advantaged here. So I thought maybe using a metric like "standard deviation of all non-0 values", but it still greatly rewards slower cubers over faster ones (which isn't bad in and of itself, I guess, but I'd prefer rewarding doing more events more). I thought about standard deviation / overall KinchRank, which would help that, but there's the issue that competitors who do few events are rewarded. I also thought of things like total range of event KinchRanks over overall Kinchrank, but it still runs into similar problems, and also isolates out the best and worst (eg, a competitor who has event KRs of 40 for all but two events, which are at 0 and 100, will have a score of 100; meanwhile, a competitor who evenly has KRs ranging from 0 to 80 would be "more well rounded" by this metric).

Anyone have any thoughts for a good statistic for measuring both well-roundedness and speed?

So, the naive approach is to simply take the standard deviation of the per-event values - but then cubers who only do a couple events and are somewhat slow are greatly advantaged here. So I thought maybe using a metric like "standard deviation of all non-0 values", but it still greatly rewards slower cubers over faster ones (which isn't bad in and of itself, I guess, but I'd prefer rewarding doing more events more). I thought about standard deviation / overall KinchRank, which would help that, but there's the issue that competitors who do few events are rewarded. I also thought of things like total range of event KinchRanks over overall Kinchrank, but it still runs into similar problems, and also isolates out the best and worst (eg, a competitor who has event KRs of 40 for all but two events, which are at 0 and 100, will have a score of 100; meanwhile, a competitor who evenly has KRs ranging from 0 to 80 would be "more well rounded" by this metric).

Anyone have any thoughts for a good statistic for measuring both well-roundedness and speed?