EPTI hope this isn't a stupid thread but I guess some people might want to know this:
People sometimes ask whats the median iccup rank so I went ahead and tried to answer that:
If you get the number of players with at least 1 game played 63710, and half it to find the middle player you get 31855, which is a D player with 1261 points.
This seems stupid because there are so many accounts with like 1 game played and stuff.. so I guess the best that can be done is finding the median rank, starting from the lowest D+ 2000 points. This filters out 44083 D and D- accounts, many of which we would like to not filter out, however finding the median from D+ seems much more accurate due to the fact that it filters out players that are better than D+/C- and have only played 9 or fewer games...
#19627 is the first account with 2000, so /2 = 9813.5.... ok the proper way to do this is (n+1)/2... so thats 9814 which is 3025 C-. The accounts next to this account are also 3025 so it doesn't matter really... So there you go.
median from lowest D- not including accounts with 0-0 is D 1261
median from lowest D+ is C- 3025
To get any better results than this would require something like... filtering out accounts with less than 15 games played... and that would be really hard to do manually so the best we can do is make an educated guess:
if the median from D+ filters out too many legitemate D and D- players, it means that the low C- we got is inflated, or in other words is higher than what the real median would be if we had a million hours to work it out.. and the average from all accounts with games played is pulled down by the many accounts with very few games played, so the low D that we got is deflated/lower than the actual median (you may say that there are D- level players with D accounts, but I'm sure there are many many more higher level players with D accounts that outweigh this)...
so the median iccup rank is probably somewhere between a mid-ish D+ like 2500ish , and...somewhere in mid D, lets say 1500.... :/ :\ :/ :\. Anyway whatever it is, it is clearly below C- and above 1261 D
Yeah this was quite a waste of time... HOWEVER, it will stop (some) people from asking this question I hope.
--------------------------
EDIT: Kwark requested rank percentiles from D+:
This is slightly time consuming because sometimes pages don't want to load instantly x_x
Well since I first posted this, the first D+ moved from #19627 to #19639.. and theres no olympics.. and while I'm doing this the numbers keep moving, making me think I should be doing this at the end of season or something.. but whatever.. I'm doing the counting from lowest rank to highest rank, even though it seems more logical to go from high rank to low rank because lower ranks fluctuate more... but uh the numbers are big enough that it doesnt really matter.. and the higher ranks don't fluctuate as much so who cares...
There are 19693 players from the lowest D+ (that is, 19693 players, not including D and D- accounts).
19693 to 10213 is D+, which is 9481 D+ accounts
10212 to 5277 is C-, which is 4936 C- accounts
5276 to 3179 is C, which is 2098 C accounts
3178 to 1975 is C+, which is 1204 C+ accounts
1974 to 1101 is B-, which is 874 B- accounts
1100 to 667 is B, which is 434 B accounts
666 to 359 is B+, which is 308 B+ accounts
358 to 127 is A-, which is 232 A- accounts
126 to 33 is A, which is 94 A accounts
32 to 1 is 32 A+ accounts
note that it's a mistake to go 10212-5277=4935, because this would mean that if there were say 3 olympic ranks, then 3 to 1 would be 3-1 = 2 olympics durr durr.. so that is why I add 1 to each result... I didn't notice this till I got to A+ though lol.. and uh.. shit... I don't know if I should do this for D+ too.. lol I suck.. ok... 19693 accounts... 10212 are NOT D+... so that means I do in fact have to go 19693-10213+1 to get the same result as 19693-10212.. ok it's fine...
ANYWAY: 3 decimal places rounded is more than enough, considering that a difference of 3 players in D+ is a difference of about .07, so there's no point having more numbers... and although I could give a million decimal points for the sake of it, it does give the numbers a misleading sort of 'accuracy', which is bad.. so it's better not to.. OK...
9481 / 19693 * 100 = 48.144% of iccup is D+ (excluding D and D-)
4936 / 19693 * 100 = 25.065% of iccup is C- (excluding D and D-)
2098 / 19693 * 100 = 10.654% of iccup is C (excluding D and D-)
1204 / 19693 * 100 = 6.114% of iccup is C+ (excluding D and D-)
874 / 19693 * 100 = 4.438% of iccup is B- (excluding D and D-)
434 / 19693 * 100 = 2.204% of iccup is B (excluding D and D-)
308 / 19693 * 100 = 1.564% of iccup is B+ (excluding D and D-)
232 / 19693 * 100 = 1.178% of iccup is A- (excluding D and D-)
94 / 19693 * 100 = 0.477% of iccup is A (excluding D and D-)
32 / 19693 * 100 = 0.162% of iccup is A+ (excluding D and D-)
This does not tell you your percentile, AKA "I'm in the top blah percent." This tells you what percentage of the total a rank is. It is different.
---------------
EDIT: ok I just thought of how to figure out the percentiles.. Now you can go 'hay guyz I'm in the top blah blha blah percent of players coughoniccupcough'...doing that now...
ok, for working these out I'm going to use the 3 decimal point figures except for the A+ figure which is easy to get again because I cbf doing all the numbers again.. so the accuracy will go down slightly.. but uh it was never that accurate to begin with so it doesn't matter.
+ Show Spoiler +
32 / 19693 * 100 = 0.1624942873102117503681511196872
0.1624....... + 0.477 = 0.6394942873102117503681511196872
0.6394...... + 1.178 = 1.817494....
+ 1.564% = 3.381494287310211750368151119687
+ 2.204 = 5.585494287310211750368151119687
+ 4.438 = 10.023494287310211750368151119687
+ 6.114 = 16.167494287310211750368151119687
+ 10.654 = 26.821494287310211750368151119687
+ 25.065 = 51.886494287310211750368151119687
0.1624....... + 0.477 = 0.6394942873102117503681511196872
0.6394...... + 1.178 = 1.817494....
+ 1.564% = 3.381494287310211750368151119687
+ 2.204 = 5.585494287310211750368151119687
+ 4.438 = 10.023494287310211750368151119687
+ 6.114 = 16.167494287310211750368151119687
+ 10.654 = 26.821494287310211750368151119687
+ 25.065 = 51.886494287310211750368151119687
======= "I'm in the top blahblah percentage of sc players that are at least D+" =========
================================================================
If you are A+ you are in the top 0.162% of players who are at least D+
If you are A you are in the top 0.639% of players who are at least D+
If you are A- you are in the top 1.817% of players who are at least D+
If you are B+ you are in the top 3.38% of players who are at least D+
If you are B you are in the top 5.59% of players who are at least D+
If you are B- you are in the top 10.02% of players who are at least D+
If you are C+ you are in the top 16.17% of players who are at least D+
If you are C you are in the top 25.60% of players who are at least D+
If you are C- you are in the top 50.67% of players who are at least D+
================================================================
=================================================================
to give you an idea of how accurate/innacure this was after losing all those decimal places earlier: if you are D+ you are in the top 100.03049428731021175036815111969... and obviously this should be 100%... so these figures are off by about -0.0304942873102117503681511196 by the time it got to D+.. and this is compunded.. so.. the closer it is to A+, the more accurate the percentage is.. so it's not bad.
EDIT: FUUUUCK I just did all that and yeah my B+ number was wrong lolol... fixed everything up...
EDIT: Other peoples' contributions:
On August 03 2009 10:35 stroggos wrote:
A+ .05%
A .14%
A- .36%
B+ .48%
B .68%
B- 1.4%
C+ 1.9%
C 3.3%
C- 7.7%
D+ 14.8%
D 62%
D- 7.1%
average amount of games per iccup account this season: 29
there is a huge amount of people at D/D+ who havn't lost a game.
It looks like from D+ onwards the amount of people in each succeeding rank halves~.
If we double D+ that is about 30% and i think that's the amount of players that should be D. but there is also a large amount of clearstats accounts at D+, and even C-.
on a random page i selected of D between ratings 1200 and 1800 only 50% of the players had a winning rate of below 50%. (the winning rate should be about 33% for a D player that cannot get any higher and also plays motw)
on a random page i selected of D+ between ratings 2200 and 2800 only 30% of the players had a winning rate of below 50%. (the winning rate should be about 33% for a D+ player that cannot get any higher and also plays motw)
so using those examples, 50% of D players should definitely have a higher rank than they have do atm. It is unclear for D+ players because their winning % would've been boosted from laddering through D rank before getting to D+
A+ .05%
A .14%
A- .36%
B+ .48%
B .68%
B- 1.4%
C+ 1.9%
C 3.3%
C- 7.7%
D+ 14.8%
D 62%
D- 7.1%
average amount of games per iccup account this season: 29
there is a huge amount of people at D/D+ who havn't lost a game.
It looks like from D+ onwards the amount of people in each succeeding rank halves~.
If we double D+ that is about 30% and i think that's the amount of players that should be D. but there is also a large amount of clearstats accounts at D+, and even C-.
on a random page i selected of D between ratings 1200 and 1800 only 50% of the players had a winning rate of below 50%. (the winning rate should be about 33% for a D player that cannot get any higher and also plays motw)
on a random page i selected of D+ between ratings 2200 and 2800 only 30% of the players had a winning rate of below 50%. (the winning rate should be about 33% for a D+ player that cannot get any higher and also plays motw)
so using those examples, 50% of D players should definitely have a higher rank than they have do atm. It is unclear for D+ players because their winning % would've been boosted from laddering through D rank before getting to D+
On August 03 2009 11:16 Kwark wrote:
Unless I'm being incompetent I think it's pretty easy to estimate them (provided the stats are provided by a useful op
)
1974-1 are B- or better out of 19000ish players included in our stats (D+ or better players).
So B- is the top 10%.
1100-1 are B or better so about top 6%.
432-1 are B+ or better so about top 2%.
Unless I'm being incompetent I think it's pretty easy to estimate them (provided the stats are provided by a useful op
)1974-1 are B- or better out of 19000ish players included in our stats (D+ or better players).
So B- is the top 10%.
1100-1 are B or better so about top 6%.
432-1 are B+ or better so about top 2%.
![[image loading]](http://img.photobucket.com/albums/v462/3xcommodore/iccuprankspie.gif)
On August 03 2009 11:54 skronch wrote:
XY chart a better representation of the data in my opnion.
Across all ranks (using Stroggos's data set)
![[image loading]](http://i26.tinypic.com/1sy0lk.png)
D+ and higher (using OP's data set)
![[image loading]](http://i29.tinypic.com/300cwtv.png)
basically it's exactly what we expected?
XY chart a better representation of the data in my opnion.
Across all ranks (using Stroggos's data set)
D+ and higher (using OP's data set)
basically it's exactly what we expected?
On August 03 2009 11:57 stroggos wrote:
assuming that every single D ranked player should be D+ or higher then yes...B- and higher is about 10%. However, using my example posted earlier 50% of D ranked people should be D+ or higher assuming they play a reasonable amount of games.
3.35% are B- and higher if we include D but if only 50% of D rank players are meant to be D+ and higher then B- and higher should be 6.675% roughly
assuming that every single D ranked player should be D+ or higher then yes...B- and higher is about 10%. However, using my example posted earlier 50% of D ranked people should be D+ or higher assuming they play a reasonable amount of games.
3.35% are B- and higher if we include D but if only 50% of D rank players are meant to be D+ and higher then B- and higher should be 6.675% roughly
hmm in response to all this, I think the median iccup rank leans more to D+ than D.
below is some more stuff I did that resulted in massive fail.. but I've spoilered it for those who want to laugh at my big waste of time:
+ Show Spoiler +
EDIT: ok I'm going to do percentiles for these estimates that include D and D-:
fuuck ok here is how I will do it: I will go with stroggos and say that half of D is placed wrongly. I'm going to get that half, and distribute it amongst the other ranks according to each ranks percentage:
59240 to 19665 = 39576 D's.
half of that is 19788
so there are
19788 accounts that are actually anywhere from D+ to olympic
24134 D-
19788 D accounts
9481 D+ accounts
4936 C- accounts
2098 C accounts
1204 C+ accounts
874 B- accounts
434 B accounts
308 B+ accounts
232 A- accounts
94 A accounts
32 A+ accounts
Now we need to distribute the 19788 D accounts:
I'm going to do something that's probably retarded and get the distribution of people above D+, and apply it to the 19788 fake D's, rounding up to whole numbers:
These are the new 'actual' numbers of people at each rank (although still not accurate, as there are of course D+ players who could be higher, C- players who could be higher, etc. but D is where there are the most of these):
If everyone at D mass gamed, then it would be something like this:
24134 D-
19788 D
.48144 * 19788 + 9481 = 19008 D+
.25065 * 19788 + 4936 = 9896 C-
.10654 * 19788 + 2098 = 4206 C
.06114 * 19788 + 1204 = 2414 C+
.04438 * 19788 + 874 = 1752 B-
.02204 * 19788 + 434 = 870 B
.01564 * 19788 + 308 = 617 B+
.01178 * 19788 + 232 = 465 A-
.00477 * 19788 + 94 = 188 A
.00162% * 19877 + 32 = 64 A+
and if you like, .000015696123057604771621409511850573 * 19788 + 0 = 0.31 olympics lol
ok so ummm... the percenTILES are FUUCK.. this takes ages... ok first I need to do percenTAGES:
ok 63710 total players... FUUUCK ok I have again rounded numbers before I've finished doing stuff.. sigh whatever.. it's still pretty good:
If D people mass gamed and reached their highest rank, the distribution of players would be like this:
+ Show Spoiler +
24134 / 63710 * 100 = 37.881023387223355831109715900173% D-
19788 / 63710 * 100 = 31.059488306388322084445142049914% D
19008 / 63710 * 100 = 29.835190707895149897975200125569% D+
9896 / 63710 * 100 = 15.532883377805681996546852927327% C-
4206 / 63710 * 100 = 6.601789358028566943964840684351% C
2414 / 63710 * 100 = 3.7890441061057918694082561607283% C+
1752 / 63710 * 100 = 2.7499607596923559880709464762204% B-
870 / 63710 * 100 = 1.3655627060116151310626275309998% B
617 / 63710 * 100 = 0.96845079265421440904096688118035% B+
465 / 63710 * 100 = 0.72986972217862188039554230105164% A-
188 / 63710 * 100 = 0.29508711348296970648249882279077% A
64 / 63710 * 100 = 0.10045518756867053837702087584367% A+
0.31 / 63710 * 100 = 0.00048657981478574792026369486736776% Olympic
37.88% D-
31.06% D
29.84% D+
15.53% C-
6.60% C
3.79% C+
2.75% B-
1.37% B
0.97% B+
0.73% A-
0.30% A
0.10% A+
and now we get to the "what percentile of players am I in?":
+ Show Spoiler +
0.10045518756867053837702087584367 A+
+ 0.29508711348296970648249882279077 = 0.395542301051640244859519698633 A
+ 0.72986972217862188039554230105164 = 1.125412023230262125255061999684 A-
+ 0.96845079265421440904096688118035 = 2.093862815884476534296028880864 B+
+ 1.3655627060116151310626275309998 = 3.459425521896091665358656411863 B
+ 2.7499607596923559880709464762204 = 6.209386281588447653429602888083 B-
+ 3.7890441061057918694082561607283 = 9.998430387694239522837859048811 C+
+ 6.601789358028566943964840684351 = 16.600219745722806466802699733162 C
+ 15.532883377805681996546852927327 = 32.133103123528488463349552660482 C-
+ 29.835190707895149897975200125569 = 61.968293831423638361324752786042 D+
+ 31.059488306388322084445142049914 = 93.027782137811960445769894835952 D
+ 37.881023387223355831109715900173 = 130.90880552503531627687961073612 D- WHAT THE FUUUUCK IM PISSED
FUUUUUUUUUUUUUCK
ok screw this. I'm not doing more detailed ones. I'm going to just do percentiles off of what stroggos estimated:
+ Show Spoiler +
FUUUUUUUUUUUUUUUUUU
yeah this really doesn't work well if the numbers aren't accurate.. it's no good to say yeah I'm in the top 0.55% of sc players give or take 15% lolol. So um yeah.. the only accurate percentile is the one I posted earlier that excludes D and D-
fuuck ok here is how I will do it: I will go with stroggos and say that half of D is placed wrongly. I'm going to get that half, and distribute it amongst the other ranks according to each ranks percentage:
59240 to 19665 = 39576 D's.
half of that is 19788
so there are
19788 accounts that are actually anywhere from D+ to olympic
24134 D-
19788 D accounts
9481 D+ accounts
4936 C- accounts
2098 C accounts
1204 C+ accounts
874 B- accounts
434 B accounts
308 B+ accounts
232 A- accounts
94 A accounts
32 A+ accounts
Now we need to distribute the 19788 D accounts:
I'm going to do something that's probably retarded and get the distribution of people above D+, and apply it to the 19788 fake D's, rounding up to whole numbers:
These are the new 'actual' numbers of people at each rank (although still not accurate, as there are of course D+ players who could be higher, C- players who could be higher, etc. but D is where there are the most of these):
If everyone at D mass gamed, then it would be something like this:
24134 D-
19788 D
.48144 * 19788 + 9481 = 19008 D+
.25065 * 19788 + 4936 = 9896 C-
.10654 * 19788 + 2098 = 4206 C
.06114 * 19788 + 1204 = 2414 C+
.04438 * 19788 + 874 = 1752 B-
.02204 * 19788 + 434 = 870 B
.01564 * 19788 + 308 = 617 B+
.01178 * 19788 + 232 = 465 A-
.00477 * 19788 + 94 = 188 A
.00162% * 19877 + 32 = 64 A+
and if you like, .000015696123057604771621409511850573 * 19788 + 0 = 0.31 olympics lol
ok so ummm... the percenTILES are FUUCK.. this takes ages... ok first I need to do percenTAGES:
ok 63710 total players... FUUUCK ok I have again rounded numbers before I've finished doing stuff.. sigh whatever.. it's still pretty good:
If D people mass gamed and reached their highest rank, the distribution of players would be like this:
+ Show Spoiler +
24134 / 63710 * 100 = 37.881023387223355831109715900173% D-
19788 / 63710 * 100 = 31.059488306388322084445142049914% D
19008 / 63710 * 100 = 29.835190707895149897975200125569% D+
9896 / 63710 * 100 = 15.532883377805681996546852927327% C-
4206 / 63710 * 100 = 6.601789358028566943964840684351% C
2414 / 63710 * 100 = 3.7890441061057918694082561607283% C+
1752 / 63710 * 100 = 2.7499607596923559880709464762204% B-
870 / 63710 * 100 = 1.3655627060116151310626275309998% B
617 / 63710 * 100 = 0.96845079265421440904096688118035% B+
465 / 63710 * 100 = 0.72986972217862188039554230105164% A-
188 / 63710 * 100 = 0.29508711348296970648249882279077% A
64 / 63710 * 100 = 0.10045518756867053837702087584367% A+
0.31 / 63710 * 100 = 0.00048657981478574792026369486736776% Olympic
37.88% D-
31.06% D
29.84% D+
15.53% C-
6.60% C
3.79% C+
2.75% B-
1.37% B
0.97% B+
0.73% A-
0.30% A
0.10% A+
and now we get to the "what percentile of players am I in?":
+ Show Spoiler +
0.10045518756867053837702087584367 A+
+ 0.29508711348296970648249882279077 = 0.395542301051640244859519698633 A
+ 0.72986972217862188039554230105164 = 1.125412023230262125255061999684 A-
+ 0.96845079265421440904096688118035 = 2.093862815884476534296028880864 B+
+ 1.3655627060116151310626275309998 = 3.459425521896091665358656411863 B
+ 2.7499607596923559880709464762204 = 6.209386281588447653429602888083 B-
+ 3.7890441061057918694082561607283 = 9.998430387694239522837859048811 C+
+ 6.601789358028566943964840684351 = 16.600219745722806466802699733162 C
+ 15.532883377805681996546852927327 = 32.133103123528488463349552660482 C-
+ 29.835190707895149897975200125569 = 61.968293831423638361324752786042 D+
+ 31.059488306388322084445142049914 = 93.027782137811960445769894835952 D
+ 37.881023387223355831109715900173 = 130.90880552503531627687961073612 D- WHAT THE FUUUUCK IM PISSED
FUUUUUUUUUUUUUCK
ok screw this. I'm not doing more detailed ones. I'm going to just do percentiles off of what stroggos estimated:
+ Show Spoiler +
A+ .05%
A .19%
A- .55%
B+ 1.03%
B 1.71%
B- 3.11%
C+ 5.01%
C 8.31%
C- 7.7%
D+ 16.01%
D 78.01%
D- 85.11%
A .19%
A- .55%
B+ 1.03%
B 1.71%
B- 3.11%
C+ 5.01%
C 8.31%
C- 7.7%
D+ 16.01%
D 78.01%
D- 85.11%
FUUUUUUUUUUUUUUUUUU
yeah this really doesn't work well if the numbers aren't accurate.. it's no good to say yeah I'm in the top 0.55% of sc players give or take 15% lolol. So um yeah.. the only accurate percentile is the one I posted earlier that excludes D and D-
EDIT: ok although I cbf doing a detailed rank percentile thing, I can't resist doing at least this. I'm going to do the ranks percentiles based on stroggos estimates that got put in that pie graph..
A+ .05%
A .14%
A- .36%
B+ .48%
B .68%
B- 1.4%
C+ 1.9%
C 3.3%
C- 7.7%
D+ 14.8%
D 62%
D- 7.1%
so that's...
If you're A+, you're in the top .05% of players
if you're A, you in the top .19% of players
if you're A-, you're in the top .55%
B+ is top 1.03%
B 1.71%
B- 3.11%
C+ 5.01%
C 8.31%
C- 16.01%
D+ 30.81%
D 92.81%
D- 100% (99.91. This just means that the number for D may be about .1 off, the number for D+ might be about .08 off, the number for C- might be about .06 off, etc.)
