<< Back to Clans Forum   Search

Posts 80 - 99 of 99   <<Prev   1  2  3  4  
CL Predictions SpreadSheet: 2/12/2017 23:21:25


Math Wolf 
Level 63
Report
Beware: wall of text and results incoming!

The detailed results per division can be found below and will also be discussed in each division topic.



#### General data-analytic remarks ####
(Skip if this is not your thing.)

There were 90 retained responses of which 38 anonymous. One troll response was deleted (so 91 surveys were collected).

Of these 90, 5 respondents showed or reported some uninformed predictions for one or more divisions, which were subsequently deleted. This resulted in 90 predictions for division A, 89 for B, 85 for Q1, 86 for Q2 and 86 for Q3.

Tests were performed to check if anonymous replies differed significantly from non-anonymous replies. No significant differences were found. Additionally, both exploratory analysis and ranking squared deviations from the mean predictions showed no different trends for anonymous responses.

Some respondents reported ties, i.e. 2 clans in the same division finishing at the same place. I see this as a feature, not a bug. Those results were retained. Consequently, the percentages per clan will sum to 100, but per rank they will not.

Some descriptives:
* Respondents were from 24 different clans, including 2 clans not participating in Clan League. 30 respondents, mostly anonymous, did not report a clan. Clans with at least 3 respondents include (in alphabetical order) 101st, Apprentice, Blitz, CORP, FCC, Hydra, Icelandic Turtles, Lynx, Masters, Outlaws, VS.
* Respondents were mostly male (64), with 2 reported females. The males includes one respondent who is assumed to be female by many. 22 repondents either left this blank or preferred not to say. The remaining 2 correspondents were of gender, other: "slug" and "tranny". Both gave their username, but I don't know either well enough to comfirm or deny this reported gender.
* Of the 54 respondents who shared their age, 51 were between 15 and 44 years old. Three outliers, all non-anonymous, were detected: one person being 88 years (likely not correct) and two others respectively 3 and 5 years old. The latter is likely the correct mental age for both.
And, of course, what all you horny pervs want to know: the non-anonymous woman chose to not share her age, the anonymous woman is 17 and therefore jailbait. So please behave!

Below, a table is reported for each division with all percentages per clan per rank (percentages calculated absolute number divided by non-missing responses as above). Additionally, a mean rank and standard deviation is given. Clans are sorted by mean rank.
A chi-squared analysis (not reported) obviously showed that the predicted ranks are dependent on the clan in each division, proving that in no divisions all clans have a completely interchangeable level (even not in Q3).
An ANOVA test on the means followed by a Tukey pairwise comparison (multiple testing correction) shows which clans are predicted to rank significantly higher than others. This test needs either a normality or large sample size assumption for correct p-values. Since this only approximately fullfilled, borderline significance cannot be trusted. Luckily, few tests showed borderline significant results.

The mean squared error (difference) between ranks given by each respondent and the average predicted ranks was calculated. No obvious outliers were detected. The person with responses closest the means was anonymous, with malakkan and Hunta respectively second and third. The person with responses farthest from the means was non-anonymous and known by me. His responses, while deviating from most, still made sense and were retained.


#### Division A ####
           1    2    3    4    5    6    7 mean   sd        
Masters 78.9 14.4  4.4  1.1  0.0  0.0  1.1 1.33 0.85
Lynx    13.3 56.7 15.6  7.8  3.3  3.3  0.0 2.41 1.13
Turtles  3.3 15.6 52.2 17.8 11.1  0.0  0.0 3.18 0.94
French   2.2  6.7 16.7 34.4 21.1 16.7  2.2 4.24 1.29
Outlaws  0.0  2.2  7.8 20.0 28.9 21.1 20.0 5.19 1.31
One      1.1  3.3  3.3 16.7 22.2 38.9 14.4 5.30 1.29
Blitz    2.2  1.1  1.1  3.3 13.3 17.8 61.1 6.22 1.30

The difference for 5th between Outlaws and One was non-significant, all other differences were highly significant. These results seem to be in line with the ones from KKND.
Full discussion at https://www.warlight.net/Forum/244331-clan-league-9-division-official-thread?Offset=9


#### Division B ####
            1    2    3    4    5    6    7 mean   sd        
GG       83.1 10.1  3.4  0.0  0.0  3.4  0.0 1.34 0.99
Mhunters  9.0 40.4 25.8 14.6  5.6  2.2  2.2 2.83 1.30
WG        3.4 25.8 21.3 41.6  3.4  1.1  3.4 3.33 1.22
VS        2.2 18.0 38.2 24.7  7.9  7.9  1.1 3.46 1.23
Sninja    1.1  2.2  3.4 13.5 52.8 15.7 11.2 5.07 1.13
VIW       2.2  1.1  7.9  2.2 15.7 36.0 34.8 5.75 1.42
BIA       0.0  2.2  2.2  2.2 13.5 34.8 44.9 6.11 1.11

The difference for 2nd between M'hunters and WG was not significant at the 5% level with a p-value of 9%. The difference for 3rd between WG and VS was not significant either. However, the difference between M'Hunters and VS is significant with a p-value of 0.9%. The difference for 6th between VIW and BIA is not significant.
Full discussion at https://www.warlight.net/Forum/244418-clan-league-9-division-b-official-thread?Offset=13


#### Division Q1 ####
              1    2    3    4    5    6 mean   sd        
Hydra      74.1 17.6  5.9  1.2  0.0  1.2 1.39 0.82
FCC        12.9 41.2 23.5 12.9  5.9  3.5 2.68 1.24
ACME        5.9 22.4 44.7 12.9 11.8  2.4 3.09 1.13
Poon Squad  4.7  8.2  7.1 32.9 20.0 27.1 4.36 1.41
Olympus     1.2  5.9 12.9 29.4 24.7 25.9 4.48 1.24
FBG         2.4  5.9  5.9 10.6 36.5 38.8 4.89 1.28

The difference for 2nd between FCC and ACME was not significant. The difference for 4th to 5th and 5th to 6th was not significant either with 4th to 6th (Poon Squad vs. FBG) borderline (p-value 4.8%).
Full discussion at: https://www.warlight.net/Forum/244379-clan-league-9-division-q1-official-thread?Offset=26


#### Division Q2 ####
                   1    2    3    4    5    6 mean   sd        
CORP            32.6 43.0  9.3  9.3  3.5  2.3 2.15 1.21
7th Heaven      52.3 20.9  3.5  8.1  7.0  8.1 2.21 1.67
LuFredd          2.3 11.6 27.9 23.3 20.9 14.0 3.91 1.32
German Warlords  5.8  8.1 25.6 19.8 22.1 18.6 4.00 1.45
Dutch [NL]       4.7 12.8 15.1 17.4 23.3 26.7 4.22 1.54
Preussen Garde   3.5  4.7 18.6 22.1 22.1 29.1 4.42 1.38

Two obvious blocks here. The difference for first between CORP and 7th Heaven is not significant (note the higher estimated variation of 7th Heaven, possibly due to some respondents not knowing who's in the clan). All differences from 3th to 6th are not significant either with LuFredd - Preussen Garde at a p-value of 18%, delivering no proof that the Warlight community sees any of these clans as having a significantly different skill level of any other.
Full discussion at: https://www.warlight.net/Forum/244354-clan-league-9-qualifier-q2-official-thread?Offset=17


#### Division Q3 ####
(the only division that matters)
                 1    2    3    4    5    6 mean   sd        
101st         25.6 29.1 23.3 15.1  4.7  2.3 2.51 1.28
Apprentice    32.6 23.3 22.1  8.1  2.3 11.6 2.59 1.62
Statisticians 24.4 22.1 23.3 20.9  7.0  2.3 2.71 1.35
TJC           15.1 14.0 12.8 20.9 30.2  7.0 3.58 1.58
DWF            3.5  9.3  8.1 27.9 37.2 14.0 4.28 1.28
Hawks          1.2  3.5  9.3  7.0 17.4 61.6 5.21 1.24

There is no significant difference between the top 3 clans here. Note that 4 clans got at least 10 first place votes here and each clan has at least 40% of the respondents placing them outside the top 2. Did we say brutal? Brutal.
Full discussion at: https://www.warlight.net/Forum/244404-clan-league-9-division-q3-official-thread?Offset=31


I also asked people to share their deepest thoughts (on Clan League) if they wanted. Some gave some interesting analyses. A summary:

*On the clans, we got the following input:
Poon Squad, 101st, ACME and Masters got an extra vote of confidence. Hydra is much stronger than in CL8, M'Hunters are cool, French is underrated, Weirdo's shouldn't be underestimated, Outlaws didn't field their best line-up, ACME will lose to Hydra, TJC to Apprentice. Poon Squad nearly beats FCC, but FCC will rock your socks. WG should dissolve and give back smileyleg to TJC. M'Hunters get mad about a rule and blame their 3rd place finish on that. They will also miss first in their division by at most the amount of point they'll lose on 3v3 EU and proceed to blame Smoove for leaving the clan.

*On specific players and groups, we got the following input:
- Q3 is the toughest: One of TJC's players is Nich (former WG player) making TJC even stronger
- Math Wolf will lose all his games.
- MoD somehow doesn't win Elitist Africa, finishing 4-2.
- MOTD SUX
- I miss Sultan ),:

* About Clan League in general, the following input was given:
- Two people hope all goes smoothly without complaining and drama.
- Meanwhile, two others already claim it's rigged.
- It's going to be a good one though and we all are wished good luck by one friendly respondent!

* Some random knowledge that was shared.
(1) What is dead needs to shut up.
(2) If tomatoes are fruit is ketchup a smoothie?
(3) 42 is obviously the answer.
(4) Falling down the stairs hurts a lot more when there's a wall halfway through your fall.
(5) Does a slug want to eat lettuce, or is it fate?
(6) Go Trump!
And finally, my favourite:
(7) Masters will win it. Hell I prefer Mistresses but what the heck!
I couldn't agree more.




EDIT: Some minor spelling and grammar fixes, descriptives for age added, links to other topics added.

Edited 2/13/2017 01:53:48
CL Predictions SpreadSheet: 2/13/2017 00:02:41


Glam 'd Slam Islam
Level 61
Report
wow! extraordinary and detailed galop from both -bravo!
CL Predictions SpreadSheet: 2/13/2017 00:39:06


Deadman 
Level 64
Report
That was a fantastic read. Keep up the good work!
CL Predictions SpreadSheet: 2/13/2017 01:21:43


Aura Guardian 
Level 62
Report
The remaining 2 correspondents were slug and tranny. Both gave their username, but I don't know either well enough to comfirm or deny this reported gender.


That's interesting...
CL Predictions SpreadSheet: 2/13/2017 04:22:43

mslasm 
Level 60
Report
Great job Math Wolf!
CL Predictions SpreadSheet: 2/13/2017 15:08:26


Howling Mad
Level 59
Report
tx math wolf
CL Predictions SpreadSheet: 2/13/2017 15:35:51


Kaerox
Level 59
Report
Really awesome job, it was a pleasure to read this all on each thread !

Also : "The latter is likely the correct mental age for both."

hi plat
CL Predictions SpreadSheet: 2/13/2017 18:16:56


Onoma94
Level 60
Report
22 repondents either left this blank or preferred not to say. The remaining 2 correspondents were of gender, other: "slug" and "tranny".


*On the clans, we got the following input:
Poon Squad, 101st, ACME and Masters got an extra vote of confidence. Hydra is much stronger than in CL8, M'Hunters are cool, French is underrated, Weirdo's shouldn't be underestimated, Outlaws didn't field their best line-up, ACME will lose to Hydra, TJC to Apprentice. Poon Squad nearly beats FCC, but FCC will rock your socks. WG should dissolve and give back smileyleg to TJC. M'Hunters get mad about a rule and blame their 3rd place finish on that. They will also miss first in their division by at most the amount of point they'll lose on 3v3 EU and proceed to blame Smoove for leaving the clan.
*On specific players and groups, we got the following input:
- Q3 is the toughest: One of TJC's players is Nich (former WG player) making TJC even stronger
- Math Wolf will lose all his games.
- MoD somehow doesn't win Elitist Africa, finishing 4-2.
- MOTD SUX
- I miss Sultan ),:
* About Clan League in general, the following input was given:
- Two people hope all goes smoothly without complaining and drama.
- Meanwhile, two others already claim it's rigged.
- It's going to be a good one though and we all are wished good luck by one friendly respondent!


Beautiful.
Thank you Math Wolf for that.
CL Predictions SpreadSheet: 2/13/2017 19:09:42


Min34 
Level 61
Report
two others respectively 3 and 5 years old. The latter is likely the correct mental age for both.

Math Wolf the savage
CL Predictions SpreadSheet: 2/13/2017 21:51:48


Jefferspin 
Level 61
Report
hahaha missed that part.
CL Predictions SpreadSheet: 2/13/2017 22:24:15

Memele 
Level 60
Report
It was a good laught reading that. Keep the good work Math wolf ;)
CL Predictions SpreadSheet: 2/14/2017 04:41:53

smileyleg 
Level 61
Report
WG should dissolve and give back smileyleg to TJC.


Thanks whoever, it's nice to be appreciated!

(Assuming this is from a TJC person that is...)
CL Predictions SpreadSheet: 2/14/2017 15:25:25

Mike
Level 59
Report
* Respondents were mostly male (64), with 2 reported females. The males includes one respondent who is assumed to be female by many.


And, of course, what all you horny pervs want to know: the non-anonymous woman chose to not share her age, the anonymous woman is 17 and therefore jailbait. So please behave!


(4) Falling down the stairs hurts a lot more when there's a wall halfway through your fall.


Quality ^_^

Why not disclose names in your report ?

Can you explain why top seeded sometimes falls 2nd in your rankings ? Doesn't make sense to me.
CL Predictions SpreadSheet: 2/14/2017 16:13:56


Onoma94
Level 60
Report
Can you explain why top seeded sometimes falls 2nd in your rankings ?


What do you mean?

Edited 2/14/2017 16:14:16
CL Predictions SpreadSheet: 2/14/2017 16:50:34


psykkoman
Level 59
Report
Jailbait at 17?
You live in weird country.
CL Predictions SpreadSheet: 2/14/2017 17:03:15

Mike
Level 59
Report
What do you mean?


7th Heaven for example. Unless they are not displayed in descending order ? Why ? WHY ?!? lol

Edited 2/14/2017 17:05:14
CL Predictions SpreadSheet: 2/14/2017 17:05:18


Min34 
Level 61
Report
                   1    2    3    4    5    6 mean   sd
CORP            32.6 43.0  9.3  9.3  3.5  2.3 2.15 1.21
7th Heaven      52.3 20.9  3.5  8.1  7.0  8.1 2.21 1.67


Because clans are based on this (the bold) number and not on who was voted what place most.

Edited 2/14/2017 17:06:32
CL Predictions SpreadSheet: 2/14/2017 17:07:30

Mike
Level 59
Report
Yeah
Clans are sorted by mean rank.


Still doesn't make sense to me though or there should be different weights on each position (top positions heavier than lower positions), which would result in just sort by % for each position.
CL Predictions SpreadSheet: 2/15/2017 01:34:42


linberson 
Level 61
Report
#### Division A ####

           1    2    3    4    5    6    7 		mean   sd        
Masters 78.9 14.4  4.4  1.1  0.0  0.0  1.1 = 100%	1.33 0.85 addition of values 1-7 shold give a total of 100%
Lynx    13.3 56.7 15.6  7.8  3.3  3.3  0.0 = 100%	2.41 1.13
Turtles  3.3 15.6 52.2 17.8 11.1  0.0  0.0 = 100%	3.18 0.94
French   2.2  6.7 16.7 34.4 21.1 16.7  2.2 = 100%	4.24 1.29
Outlaws  0.0  2.2  7.8 20.0 28.9 21.1 20.0 = 100%	5.19 1.31
One      1.1  3.3  3.3 16.7 22.2 38.9 14.4 = 100%	5.30 1.29
Blitz    2.2  1.1  1.1  3.3 13.3 17.8 61.1 = 100%	6.22 1.30


Ranked after mean.

Mean = you add up all the numbers and then divide by the number of numbers.

The percentage below each rank then shows you how many players voted them into that place.
CL Predictions SpreadSheet: 2/22/2017 12:38:38


Math Wolf 
Level 63
Report
Thanks for the nice words everyone! It was my pleasure. And obviously, I have no idea why you (mostly) guys find a report with some boring statistics funny.

To answer some of the comments:
(Assuming this is from a TJC person that is...)

This respondent is a member of TJC.

Why not disclose names in your report ?

In the survey itself, it was stated that results would be anonymized. I alone keep the non-anonymized data for quality control. I'm willing to share the fully anonymized data to anyone who can give me a good reason why they want them (i.e. an idea for a data-analysis) through PM.

Can you explain why top seeded sometimes falls 2nd in your rankings ? Doesn't make sense to me.

As linberson explained, they are ranked after mean. I went back and forth a little on this because the mean is not a great choice for this, but it is the best one there is, I believe. Sorting by number of first place votes doesn't make sense, it would put Blitz before Outlaws for example in A. Sorting by mode or median doesn't make sense either, see Q3 for example (and an additional tiebreaker would be needed).

Another system I considered was sort of a surplus votes ranking. The first ranked team is the one with the most first place votes. The second ranked team is the ones with the most first and second place votes, where first place votes count double, etc. The problem with this system is that going from first to last can give a different ranking than going from last to first, which is undesirable for a ranking system.

I considered putting weights on positions as well, as you mentioned. My initial idea was something along the lines 10-7-5-3-2-1-0. I decided against this as any choice of weights would have been arbitrary. Weights would assumingly work better if there is indeed uninformed last-place voting, as is probably the case for 7th Heaven and Apprentice.

Jailbait at 17?
You live in weird country.

In many jurisdictions, this is the case. Note that for all practical purposes the jurisdiction of the act where it happened, applies. I'm not sure under which jurisdiction Warlight.net falls in case of sexual harassment, so I erred towards caution.
Posts 80 - 99 of 99   <<Prev   1  2  3  4