EPTSC2 Ranks (site) - Page 8
| Forum Index > SC2 General |
|
Shadowed
United States679 Posts
| ||
Empyrean
17070 Posts
On August 10 2010 14:46 Shadowed wrote: Median seems better since average could get propped up (slightly) by being in the first diamond division and such. Yeah, but the problem with the median is that you'll get a lot of divisions that are the same (for example, 366 or something). The older diamond divisions will also have higher medians as well, simply because ALL scores are higher (so the extra effect the really high scores have on the mean won't really matter). You could try taking the average of all players in the middle 50%, though. That kind of combines both median and mean, but there's no way you're going to account for older divisions vs. younger divisions. | ||
|
Shadowed
United States679 Posts
 | ||
|
StaR_Robo
Australia229 Posts
| ||
|
Shadowed
United States679 Posts
On August 10 2010 14:59 Empyrean wrote: Yeah, but the problem with the median is that you'll get a lot of divisions that are the same (for example, 366 or something). The older diamond divisions will also have higher medians as well, simply because ALL scores are higher (so the extra effect the really high scores have on the mean won't really matter). You could try taking the average of all players in the middle 50%, though. That kind of combines both median and mean, but there's no way you're going to account for older divisions vs. younger divisions. Since everything works out roughly the same no matter what you apply, and I've been recording point averages for a while it's easier to just use averages, as such: http://sc2ranks.com/div/all/diamond/1/points/0 The interface is slightly inconsistent with the rest of the rankings/divisions, but that's because it'll be the new base look that I'm moving the other listings to. | ||
|
Shadowed
United States679 Posts
Everyone else will still have snapshots taken every 24 hours, but updating the top 100 hourly will let people see progression in more finer detail. [edit] And for good measure, the top 100 will also have their world rank saved so it can be shown. | ||
|
Silu
Finland165 Posts
| ||
|
Shadowed
United States679 Posts
| ||
|
Silu
Finland165 Posts
On August 11 2010 17:02 Shadowed wrote: I'm not actually sure if that would be considered inaccurate. If you have someone tied for a rank then it's the same and not higher/lower. If normally that's untrue then I can change it though. Sure they should have the same rank if they're tied. But if my rank is 10, I expect to be the TENTH BEST. So if there are two guys at rank 8, the NEXT guy after them should have rank 10 instead of rank 9. | ||
|
MaxField
United States2386 Posts
| ||
|
Shadowed
United States679 Posts
On August 11 2010 17:04 Silu wrote: Sure they should have the same rank if they're tied. But if my rank is 10, I expect to be the TENTH BEST. So if there are two guys at rank 8, the NEXT guy after them should have rank 10 instead of rank 9. Fair enough, I changed it and it'll go live in about 20 minutes. | ||
|
k!llua
Australia895 Posts
Total number of players Number of players in each division Breakdown of race in each division Number of games played/lost | ||
|
Shadowed
United States679 Posts
On August 14 2010 09:58 k!llua wrote: is there any way we can get a simple stats for each server, so like, you click on SEA or Asia server and you can get: Total number of players Number of players in each division Breakdown of race in each division Number of games played/lost Sure, added to the list of things to do. Also, I'm in the process of adding graphs. Because you really can't go wrong with graphs, unless you were to make a graph of how awesome graphs are. All teams will get graphs of how their world rank and points changed, for example: http://sc2ranks.com/char/us/388538/HuK clicking 'Reset Zoom' you can see that Huk dropped from #2 world wide to #10 for about an hour. You can also see the top 10 1v1 players and how all of their points have changed http://sc2ranks.com/ranks/graph I'm still tweaking the charting library to get cleaner looking graphs right now as well. | ||
|
Lucid90
Canada340 Posts
I'm currently a 600 diamond zerg and my winrate vs T is fucking HORRENDOUS. By far my best matchup is ZvZ, and I'm okay ZvP, but my ZvT is by far my worst matchup. I really want to see what's my winrate against T's and I've looked but I couldn't find any feature like that. | ||
|
Shadowed
United States679 Posts
Few other changes: Redid all of the column listing layouts to use tables for my sanity and so they look better, let me know if you have complaints or you notice some weird glitches in how they look. Exposed the last updated at/last game played/joined league on data for teams http://sc2ranks.com/team/111866 You can also see percentile and how many players the percent/percentile are out of. Added the league icons next to ranks so it's hopefully clearer those are league ranks not across every league. Times for age are now 'accurate' and will update with Javascript even after the page is cached, granted this just means they can be about 10 minutes more accurate, but still counts! Updating tweaked again as well, every division will be updated every 12 hours and a snapshot taken (top 100 1v1 diamond are still hourly), down from every 24 hours. I might reduce this to every 6 hours, but not sure yet. On August 14 2010 15:51 Lucid90 wrote: is there any way to check your win rate against other races? I'm currently a 600 diamond zerg and my winrate vs T is fucking HORRENDOUS. By far my best matchup is ZvZ, and I'm okay ZvP, but my ZvT is by far my worst matchup. I really want to see what's my winrate against T's and I've looked but I couldn't find any feature like that. Unfortunately no, Blizzard does not expose this information so there's no way of doing it short of parsing every single replay. | ||
|
7mk
Germany10157 Posts
didnt expect him to be able to keep up after all this time | ||
|
lowlypawn
United States241 Posts
| ||
|
ThunderGod
New Zealand897 Posts
In the end I came up with a rather simple method which would be cool if you implemented on sc2ranks.com. If you could rank divisions in descending order by the average rating points of all players in the division muiltiplied by the average winning percentage of all players in that division. This ranking isn't perfect but I think it gives a fair representation. The advantage of this is its better than the current ranking of just average points and easy to calculate. However its quite simplistic so could probably be improved later on. (New divisions have fewer people and lower rating, however they generally have a higher win percentage as they haven't been playing top diamond players yet. Of the divisions which are full rating points are clearly not equatable, but good players tend to have EITHER high rating points OR a high win ratio). Edit: I guess if this feature were incorporated it could also be used for overall player ranking in a server and player ranking within a division. | ||
|
Shadowed
United States679 Posts
I'm open to discussion if other people want to weight in on this, that's just what I'm thinking. | ||
|
Escapist
Portugal548 Posts
Good effort and excellent work, i added a link to your site on my youtube shoutcasting channel. | ||
| ||
