All-Round Rankings (KinchRanks)

Kit Clement said:
Consider this simple example -- to get a 50% score in 2x2, you need to have a 3.20 average. That takes quite a bit of skill, and likely takes knowing some EG, at the very least CLL. To get the same score in 5x5, you need to have a 1:48.40 average. That's a considerably easier level to attain -- and it's just because the numerator for 2x2 is so small. It's even sillier for FMC -- getting that score takes a solve of 40, or a mean of 50, which someone can do having never tried an FMC attempt, assuming basic 3x3 knowledge.

I do agree though -- It is rather desirable that the score function for an event is convex, but I'm just arguing that it may be a bit too convex. I do see the appeal for ease of interpretability though, and I'm starting to see how applying logs is just getting arbitrary.
Click to expand...
Ah thanks, I understand now I guess. I actually wouldn't agree that 3.20 even requires CLL. Good ortega scrambles could achieve that. I think the effect is not so much that 2x2 inherently has a small numerator, but rather that the best averages are likely to be lucky and better than e.g. the general home average of theirs.
Fewest moves is a different kettle of fish, because it is not a speedsolving event and I think we're quite lucky that the spread of results is anywhere near the spread of speedsolving events.
I don't think there's any obvious or neat way to "fix" these cases, and I really think that the bias in them is tiny compared to the bias removed compared to e.g. sum of ranks by the whole system generally. So I'm not really bothered by it, although it's nice to have it at the back of the mind.

not_kevin said:
Thanks for the update on the main post - cool to see how the countries compare to each other. All the more confirmation that Poland is a huge cubing powerhouse - even the huge populations and competitive depth of the United States and China can't overcome Polish talent :P

Mostly as a personal request (getting something like this set up would take quite a few more resources, I'd imagine), but I'd love to see how people compare within their regions (like how sum of ranks is set up on the WCA site). We now know how strong the regions are compared to each other - but how strong am I within my region? I can calculate my own regional KinchRanks without too much difficulty (not including the 3bld and 3fm mean change, my score is 45.38 in NA), but how does that rank against the rest of NA?
Click to expand...

Yes, indeed the Polish bld skills really come to light once you treat them as equal events to all others!

Yep, regional KinchRanks is relatively easy to calculate - I already did UK although I can't remember whether I ever posted it. Hopefully I'll get round to generating a few countries and continents soon. I'd love to get something set up like the WCA Sum of Ranks page, although I don't have the knowledge to do that.

Keroma12 said:
I like this. Is it possible for this to be on the WCA page along with sum of ranks? If not, it would be really nice if it were hosted on the web somewhere. I would also like to be able to see ranks within each country (with respect to NRs instead of WRs).

Edit: http://hem.bredband.net/_zlv_/rubiks/stats/kinch_ranks.html (thanks Kit) (still waiting on by country though!)
Click to expand...
WCA page is an eventual consideration. I didn't suggest it immediately as I wanted to continue discussing it for some time first to make sure it's right. Indeed this thread has helped a lot, so thanks everyone for the feedback.

As mentioned above, I can calculate ranks within countries one at a time.

On the topic of regional KinchRanks: I actually don't like them too much, because you end up comparing against NRs, which are much more volatile and also prone to biases within countries. It's a nice rank for sure, but it's not quite as objective as comparing to WRs. Ideally I wouldn't use WRs at all in the calculation, but it is the concession I have to make to provide a benchmark that is as close to equal from event to event as possible. So using CRs and NRs is a further step away from the theoretical perfect benchmark, but still provides nice rankings.
 
kinch2002 said:
Yes, indeed the Polish bld skills really come to light once you treat them as equal events to all others!
Click to expand...

Related to this, I'm somewhat curious what events are most strongly correlated to each other - as in, if I'm good at [some event], what events am I most likely to also be good in, or what events am I most likely to do poorly in? For example, it seems that a good 333mbf score translates almost always to great 333bf/444bf/555bf scores, and that 666 and 777 are very similar, but what would feet and Square-1 correlate with?

kinch2002 said:
On the topic of regional KinchRanks: I actually don't like them too much, because you end up comparing against NRs, which are much more volatile and also prone to biases within countries. It's a nice rank for sure, but it's not quite as objective as comparing to WRs. Ideally I wouldn't use WRs at all in the calculation, but it is the concession I have to make to provide a benchmark that is as close to equal from event to event as possible. So using CRs and NRs is a further step away from the theoretical perfect benchmark, but still provides nice rankings.
Click to expand...

I do agree with the negatives with the volatility, and the lack of objective comparison to the "ideal cuber", but the main reason that I'm interested in regional KinchRanks is precisely because of the biases within countries - for example, because Southern California (at least in the past ~7 years) has been rather weak at BLD, there's not much reason for competitions to hold BLD - which, of course, creates a negative feedback loop. Meanwhile, OH is pretty heavily emphasized here, which means that even for someone who doesn't consider themselves very good at OH, the nature of the scene means that one tends to actually be fairly good globally (an easy example here is myself - even when I got sub-20, I didn't consider myself very good because 19 averages rarely podium, but sub-20 is actually pretty good - a KinchRank of >54, and a rank of <577 out of ~9500). Although my example is of a region we probably wouldn't calculate for, it hopefully gives some idea as to why I'm interested in the statistics, even though they deviate somewhat from that comparison with the "ideal" in the normal KinchRanks.
 
I've massively overhauled the OP, and made a website to host the info and latest rankings on.

not_kevin said:
Related to this, I'm somewhat curious what events are most strongly correlated to each other - as in, if I'm good at [some event], what events am I most likely to also be good in, or what events am I most likely to do poorly in? For example, it seems that a good 333mbf score translates almost always to great 333bf/444bf/555bf scores, and that 666 and 777 are very similar, but what would feet and Square-1 correlate with?
Click to expand...
I may have a look into that sometime - sounds interesting. Or else you could do some correlation analysis yourself :)

not_kevin said:
I do agree with the negatives with the volatility, and the lack of objective comparison to the "ideal cuber", but the main reason that I'm interested in regional KinchRanks is precisely because of the biases within countries - for example, because Southern California (at least in the past ~7 years) has been rather weak at BLD, there's not much reason for competitions to hold BLD - which, of course, creates a negative feedback loop. Meanwhile, OH is pretty heavily emphasized here, which means that even for someone who doesn't consider themselves very good at OH, the nature of the scene means that one tends to actually be fairly good globally (an easy example here is myself - even when I got sub-20, I didn't consider myself very good because 19 averages rarely podium, but sub-20 is actually pretty good - a KinchRank of >54, and a rank of <577 out of ~9500). Although my example is of a region we probably wouldn't calculate for, it hopefully gives some idea as to why I'm interested in the statistics, even though they deviate somewhat from that comparison with the "ideal" in the normal KinchRanks.
Click to expand...
That does make it sound interesting. I've done UK and US ranks on my website, and I'll add some more soon.
 
Can we have an update with this? I noticed my 2x2 rankings aren't updated.

EDIT: Yeah, as a note: For sum of ranks, my 2x2 barely changes (2 --> 1 does basically nothing), but for this it makes me got from 94.7 --> 100, which is pretty big. I like this; it motivates you to still keep getting faster at events you're already good at as much as get faster at events you're bad at as well, unlike sum of ranks, which heavily relies on getting fast at events you're bad at.
 
Last edited:
I wrote a short python script last night to calculate continent KinchRanks. The script is here. I tested it by generating the KinchRanks for Africa. The results are here.

I've also added all 6 continents now.

Edit1: Except MBLD results seem to be problematic... :( Getting scores over 100 for some reason...
Edit2: The error with MBLD scores is fixed now.
 
Last edited:
Having just gotten a single in feet, but failing to make the cutoff (and therefore no mean), I'd like to see some kind of single weight involved. That being said, I understand that since the purpose of this is to compare against the "ideal" cuber (and of course the ideal cuber would always make cutoffs and never deal with this issue), the clear solution is just to git gud, but it's still a bit of a feelbad to not have any improvement in an event, despite making "improvements" in sum of ranks.
 
so how long until the wca realizes this is better than sum of ranks and implements into their website?
 
You must log in or register to reply here.

Similar threads

Carrot
  • Poll Poll
World Ranking vs Round Ranking
Replies
15
Views
3,385
Carrot
Carrot
Z
Most 3x3 sub5 avgs in one round?
Replies
3
Views
404
zhanglei1943
Z
Cubing_Grandson
Cubing Grandson Progression
Replies
14
Views
873
Cubing_Grandson
Cubing_Grandson
speedcubingworld.com
Could someone explain what dual rounds are?
Replies
4
Views
431
speedcubingworld.com
speedcubingworld.com
ILCubing
Could a competition with 5 rounds of an event happen?
Replies
5
Views
426
Ordway Persyn
Ordway Persyn
Back
Top