Play
Multi-Player
Coins
Community
Settings
Help
Community   Maps   Forum   Mail   Ladders   Clans   Recent Games
Sign In | Sign Up
<< Back to Clans Forum   

Posts 1 - 30 of 35   1  2  Next >>   
A Method of Ranking Clan Players: 12/21/2015 00:44:20


Benjamin628 
Level 59
Report
So let's say you have a clan and you want to find out who is the best. Just win rates, or ladder performances are unhelpful, especially depending on your definition of skill. The AWF has 3 "ladders" which are just random teamed/match-up 1v1, 2v2, and 3v3, and are ranked by win-rate. Just straight win-rate isn't good enough for me, so I came up with multiple parameters to rank the clan members:

AWF Score = (((4 * '1v1 Ladder Wins') + (5 * '2v2 Ladder Wins') + (6 * '3v3 Ladder Wins')* 10 * 'Ladder Win Rate') + ('1v1 Win Rate(General)' + '2v2 Win Rate(General)' + '3v3 Win Rate(General)'* 10) + ('1v1 Ladder Peak'/10))/('Boot Rate' + 10)


This created the following leaderboard:

Appalachian Score:

tschombawomba		1702
master of desaster	1524
jz              	1500
Stealth90		1206
Blortis	        	1059
kachan	                957
Benjamin628	        904
Sugi			868
Doge			857
Jayvan			848
ppface			819
TheThedde		655
HankyPinky		403
Legend of the Phoenix	360
Andrew			341
Lord of Shadows		326
JT			322
BilboSwaggins		294
Pajama Dance		205
Yellowswag		187
L'Esophagus		165
shyb			155
Legacy			108
Character		71
Jim			60
marsad PL		49
Jammawind		48
Jplatt821		43
Oscar			38


The reason that tschombawomba is above the two Masters that play on the AWF Ladder is because he is 3-0 on the ladder, and AWF Ladder Games matter the most in the AWF Score (Jz is 3-2). Also, MoD has a higher Boot Rate than tschombawomba, which is a pretty big difference (Even if its 11.9/14.4 vs 1.9/4v4)

The Entire AWF Spreadsheet can be seen here:

https://docs.google.com/spreadsheets/d/1tz-I-btd5cASH5UhN4RLtOQo-3yyt5DH3_uxGfHABW0/edit#gid=166654506

The Entirety of players ranked on the ladder can be seen here:

https://www.warlight.net/Clans/?ID=76 AWF

https://www.warlight.net/Clans/?ID=11 Masters

https://www.warlight.net/Clans/?ID=206 Acme. Co

https://www.warlight.net/Clans/?ID=84 Stahlgewrittern

Edited 12/21/2015 00:45:01
A Method of Ranking Clan Players: 12/21/2015 00:54:39

Luna {TJC}
Level 56
Report
Intresting stuff
A Method of Ranking Clan Players: 12/21/2015 01:00:23


kynte
Level 51
Report
Feels like the formula is a bit contrived. Also, going to add clan team ID scraping to the PlayerScraper so I can automate these for players.
A Method of Ranking Clan Players: 12/21/2015 05:38:22


knyte 
Level 58
Report
As, promised: http://pastebin.com/6EaDzjY5

Based on the formula:

score = ( ( ( (# of 1v1 ladder wins * 4)
+ (# of 2v2 ladder wins * 5)
+ (# of 3v3 ladder wins * 6) )
* 10 * overall ladder win rate) +

((general 1v1 win rate +
general 2v2 win rate +
general 3v3 win rate) * 10) +

(peak 1v1 ladder rating / 10) ) /

(bootRate + 10) )
A Method of Ranking Clan Players: 12/21/2015 05:51:20


Norman 
Level 57
Report
From knyte's link:
1. sarink: 18639 points
2. Leo41: 16223 points
3. Chromeo: 15995 points
4. Javor: 14389 points
5. AyanM: 11953 points
6. [WL贴吧]jhk: 7984 points
7. ziskonian republic: 7545 points
8. Rogue Nikolai Krogius: 7416 points
9. fifi: 5963 points


So sarink is the M'Hunters clan champion and I'm..... number 20. For the Master Hunters the list seems to make no sense at all, it just rewards mediocre success on all kinds of ladders.

Edit: The guys ranked first for M'Hunters have just the most games on the 1v1 ladder played. This gives sarink a better rating than [WL贴吧]jhk who took first place.

Edit 2:
--> 1. {101st}Blowfly: 57508 points

Edited 12/21/2015 05:58:01
A Method of Ranking Clan Players: 12/21/2015 06:05:01


knyte 
Level 58
Report
Yeah the rating system doesn't make any sense to begin with. I just did this to debug my code. Found a few interesting edge cases.
A Method of Ranking Clan Players: 12/21/2015 06:31:29


Diplomatic Immunity 
Level 54
Report
1. It_Is_I: 4163 points
2. xsandman: 2660 points
3. Fen: 1334 points
4. VaporWolf [FROG]: 1289 points
5. Lolicon love &lt;3: 1116 points
6. Diplomatic Immunity: 842 points
7. Tiwato: 757 points
8. BielekPL: 395 points
9. Alvin Nobody: 195 points
10. MoneyMakingMitch: 61 points
11. IV: 44 points
12. Epicular: 34 points
13. Highlander: 19 points

Sounds right, seeing as those at the top (maybe with the exception of loli) are actually the best in the clan.
A Method of Ranking Clan Players: 12/21/2015 06:46:16

Ryzys
Level 57
Report
From knyte's Link (and some editing from myself):
Ranked 29 members in ILLUMINATI ▲ (edited now to include me):

1. Sebus Maximus: 90795 points
2. Iron Mike: 35409 points
3. mech: 25198 points
4. Nyetworker: 22385 points
5. Derfellios: 17323 points
6. Great Expanse: 11954 points
7. Gimli&acute;s Koala: 8104 points
8. MightySpeck (a Koala): 7374 points
9. The Chosen Koala: 6683 points
10. Zappyelf: 6468 points
11. Ryzys: 5356 points
12. Tiny Koala: 5180 points
13. Stormers: 5027 points
14. Miles Edgeworth: 4941 points
15. Doc: 4714 points
16. theRanDMC: 3559 points
17. SamDatt: 3235 points
18. PancakeLover: 3122 points
19. The Forbidden Koala: 2751 points
20. cookie: 2468 points
21. Megatron: 2446 points
22. Angry Koala: 2440 points
23. WhoHitMyCar: 2401 points
24. eneskristo: 2068 points
25. Sitzkrieg: 1427 points
26. Grandpa Koala: 1139 points
27. The Hysterical Koala: 1015 points
28. Green Turtle: 178 points
29. Diablo Rojo: 77 points
============

(Disregard this, it was a question)

Edited 12/21/2015 08:36:32
A Method of Ranking Clan Players: 12/21/2015 06:54:02


kynte
Level 51
Report
Because you don't have any Ranked Games data displayed on your profile. Any player that didn't have both Ranked and Ladder data on their profile was automatically excluded because their score would be too skewed by the exclusion of one of the two major categories.
A Method of Ranking Clan Players: 12/21/2015 07:02:50


Fan the Apostle
Level 56
Report
I challenge you to do HAWKS
A Method of Ranking Clan Players: 12/21/2015 07:17:57


knyte 
Level 58
Report
It's already in there:

Ranked 14 members in HAWKS
   
1. AlexB: 16708 points
2. Mek Blaze: 4808 points
3. Scotty: 3264 points
4. Gazpacho: 257 points
5. Cicero XXII: 157 points
6. Crown King: 83 points
7. buddyboyb1: 79 points
8. Mitko: 66 points
9. Awsome911: 49 points
10. AlexRS: 45 points
11. Rear Admiral John: 44 points
12. Tristan: 31 points
13. kolosser: 17 points
14. A Talking Fan: 16 points
============
A Method of Ranking Clan Players: 12/21/2015 07:25:30


Fan the Apostle
Level 56
Report
OK I may be 14thbut I am not 84th!


Why only 14?
A Method of Ranking Clan Players: 12/21/2015 07:26:07

Ryzys
Level 57
Report
That makes sense now, thanks knyte/vitriol. Just another Q for you: When people's win rates are displayed after these stats are on there, does the player in question (in this case, myself) get their score added to there or not?

Edited 12/21/2015 07:26:22
A Method of Ranking Clan Players: 12/21/2015 07:29:32


kynte
Level 51
Report
@Fan: because only 14 people had enough data on their profile for this to work

@Ryzys: I see you've made your Ranked Games public again. :P I can't really retroactively edit the paste, but your score is 5356, which would get you ranked 11th.
A Method of Ranking Clan Players: 12/21/2015 08:35:07

Ryzys
Level 57
Report
Cool, thanks vitriol!
A Method of Ranking Clan Players: 12/21/2015 10:36:30


Math Wolf 
Level 62
Report
Nice attempt, but agree with Norman and knyte that it doesn't make much sense:

* number of games skews towards older and more active players (overall, me or Heyheuhei would get the top position)

* Win % doesn't make much sense either as everyone except the top 25 and bottom 25 should have a win rate close to 50% anyway.

What would be a better (but more extensive) attempt:

* Use TrueSkill since the start of the ladders to get everyone's rating. BayesElo resets to 0 after you left the ladder for 5 months, but TrueSkill does not and still gives a decent accurate ranking.

* Use both the best rating reached and the current rating.

* Use the seasonal ladders (for seasonal only end rating makes sense, peak rating does not) and RT ladder as well.

* A formula could be
(2*best rating 1v1 + current rating 1v1 + best rating 2v2 + best current rating 2v2 + 2nd best rating 2v2 (other team) + best rating 3v3 + best current rating 3v3 + 2nd best rating 3v3 (other team) + 2*best rating RT + current rating RT + best seasonal rating + 2nd best seasonal rating + 3rd best seasonal rating)/15

Someone without a rating in something should have a reasonable baseline value imputed (the 25% quantile for example) and someone who scores lower than this baseline, can have the baseline imputed as well. This would reduce the influence of one bad or inactive ladder.

Additionally, it may be worthwhile not to input the absolute numbers, but rather standardized scores. After all, some ratings are higher than others. Seasonals for example, are a lot higher and the absolute numbers cannot be compared over seasons. Another, simpler solution, would be to recompute Seasonal ratings with TrueSkill as well so everything is on the same scale. This would be the best solution as far as I know.

Doing this wouldn't be all that difficult, after all, there is a TS program for WL available on the blog. The difficult part is exporting and combining the results in an automatic way. Without that step, I would have done this a long time ago already.

Edited 12/21/2015 10:59:02
A Method of Ranking Clan Players: 12/21/2015 10:52:21


Master Ryiro 
Level 62
Report
also,apparently hiding my ladder stats gave me a rank very low in the list
A Method of Ranking Clan Players: 12/21/2015 10:56:37

Master of the Dead 
Level 62
Report

A formula could be
(2*best rating 1v1 + current rating 1v1 + best rating 2v2 + best current rating 2v2 + 2nd best rating 2v2 (other team) + best rating 3v3 + best current rating 3v3 + 2nd best rating 2v2 (other team) + 2*best rating RT + current rating RT + best seasonal rating + 2nd best seasonal rating + 3rd best seasonal rating)/15


I think you have a typo Math Wolf. The bold text should be : 2nd best rating 3v3 (other team)?


I don't think many people would have a 2nd 3v3 team yet though, due to it being so new. A statistical method is probably the only feasible way to rank a huge group of players. However, I think it does not provide a complete picture. I know of some great players who've never paid much attention to the ladder.

Edited 12/21/2015 11:00:34
A Method of Ranking Clan Players: 12/21/2015 11:04:19


Math Wolf 
Level 62
Report
I think you have a typo Math Wolf. The bold text should be : 2nd best rating 3v3 (other team)

Changed, thanks.

I don't think many people would have a 2nd 3v3 team yet though, due to it being so new. A statistical method is probably the only feasible way to rank a huge group of players. However, I think it does not provide a complete picture. I know of some great players who've never paid much attention to the ladder.

Agree on the 2nd 3v3 team. But the idea is to have a basic formula that is valid as long as the current ladders exists and can easily be extended whenever a new ladder is added. If nobody has it, the baseline can be imputed for everyone and it doesn't make a difference anyway.

Also agree that it doesn't provide a complete picture. That is after all simply not possible. Ladders are one of the only ways in which players play against each other in an "objective" framework with somewhat objective pairings (note the quotation marks, I'm well aware it's not all that objective, but the closest thing to it we have at the moment). It can still be gambled as we all know, but it's as close to a system we have. Other systems like clan league, clot and some tournaments could play a role as well, but from a data standpoint, their information is more difficult to be processed and it would be subjective what to include and what not.
A Method of Ranking Clan Players: 12/21/2015 11:57:05


Onoma94
Level 59
Report
Ranked 37 members in The Lost Wolves
   
1. Mitochondria: 6285 points
2. McMuller: 4284 points
3. Onoma94: 4077 points
4. [WOLF] Navigator: 2002 points
5. World Dominator: 1599 points
6. [WOLF]SuperSiezure: 1297 points
7. Semicedevine: 1062 points
8. [WOLF] Marconios: 1029 points
9. [WOLF] White Fang: 647 points
10. [WOLF] AMERICAN EMPIRE: 619 points
11. japan77: 334 points
12. [WOLF] Keyes: 331 points
13. [WOLF] Zephyrum: 253 points
14. [HOLY]WARLORD t(-_-t)[WOLF]: 168 points
15. ♤Prince Riku Lyonheart♤: 155 points
16. [Wolf] Karkat: 110 points
17. Try Again: 88 points
18. Darth player: 84 points
19. The Third Reich: 67 points
20. Hermit: 67 points
21. RICK: 60 points
22. Mister_Wolfy: 59 points
23. Kakarot: 51 points
24. You:: 44 points
25. [WOLF] Executor: 36 points
26. Major General Whiskers (aka) TEG: 34 points
27. [Wolf] Relm: 34 points
28. Femaelstrom: 33 points
29. General Gynther: 28 points
30. [WOLF]Sean Mckenna: 27 points
 
32. [WOLF]HoustonLeHammer: 26 points
33. justin: 25 points
34. Villarrealfan: 22 points
35. Snowy: 21 points
36. [Wolf] Eon Kid: 17 points
37. General Rubino: 14 points
A Method of Ranking Clan Players: 12/21/2015 14:06:17

MasterOfWarZone
Level 42
Report
1. {101st}Blowfly: 57508 points

3. Andersault: 46738 points
4. Falker: 40279 points
5. Tenshi: 33621 points
6. marc2013: 29281 points
7. Ineffable: 21073 points
8. Mirror: 17582 points

Blowfly is better than most of WG. Legit rating scale.
A Method of Ranking Clan Players: 12/21/2015 15:29:56

[wolf]japan77
Level 57
Report
Well, that seems broken

I can't possibly be the 11th best player in the clan, and who on earth is this Mitochondria person.
A Method of Ranking Clan Players: 12/21/2015 15:43:20


Min34 
Level 58
Report
Blowfly is better than most of WG. Legit rating scale.


indeed quite broken. Blowfly should be rated above ALL WG members. They don`t even come close to the great and almighty Blowfly.
A Method of Ranking Clan Players: 12/21/2015 15:47:41


Onoma94
Level 59
Report
Mitochondria = Steven Luo, he changed nickname. Still has biggest lvl in TLW, so that's possible.
A Method of Ranking Clan Players: 12/21/2015 15:56:28


ChrisCMU 
Level 60
Report
Why total wins? That just rewards volume. What if one person is like 30-2 and the other is 35-100 on the same ladder?

I think your formula needs a tweak. Personally, I would go by rating, not wins-losses as you don't know who those wins are against.

Also, why am I not on our list of WG people anyway?

Edited 12/21/2015 15:58:55
A Method of Ranking Clan Players: 12/21/2015 16:07:05


Onoma94
Level 59
Report
Also, why am I not on our list of WG people anyway?

Because of stats not being publicly visible.

Edited 12/21/2015 16:07:20
A Method of Ranking Clan Players: 12/21/2015 16:16:39


{Shredtail2}
Level 56
Report
My bootrate completely killed my score. I need to work on that.
A Method of Ranking Clan Players: 12/21/2015 16:28:18


ChrisCMU 
Level 60
Report
Because of stats not being publicly visible.


Wait, so this is not even ladder stats, but just stats in general? Wow, this is even more screwed up (sorry Ben). Win rate is easily abused in non ladder games.
A Method of Ranking Clan Players: 12/21/2015 17:00:20


Fleecemaster 
Level 59
Report
^It's true, stats are pretty meaningless

Edit: Fleecemaster: 1888 points

Haha! :D

Edited 12/21/2015 17:02:15
A Method of Ranking Clan Players: 12/21/2015 20:12:49


knyte 
Level 58
Report
@TLW: I have no idea where #31 went. I thought it was an issue with my ranking function but nope. Literally no idea what happened.
Posts 1 - 30 of 35   1  2  Next >>   

Contact | About WarLight | Play Risk Online | Multiplayer Strategy Game | Challenge Friends, Win Money | Skill Game | Terms of Service