Fortress | Sveet
▀▄▀▄▀▄▀▄▀▄▀▄▀▄▀▄▀
I never really considered him a MW resident though.... idk why
-
Welcome to Smashboards, the world's largest Super Smash Brothers community! Over 250,000 Smash Bros. fans from around the world have come to discuss these great games in over 19 million posts!
You are currently viewing our boards as a visitor. Click here to sign up right now and start on your path in the Smash community!
Chicago-Land Power Rankings (Updated 12/30/2011)
- Thread starter Fortress | Sveet
- Start date
Fortress | Sveet
▀▄▀▄▀▄▀▄▀▄▀▄▀▄▀▄▀
he was in ohio somewhere, but hes back to living in NJ i believe
instead of just asking for opinions or ranking placements, why don't we use a more active policy.
check more intervals: mms, seriouslies, round-robins, single elims.. enforce confrontation matches for ambiguous rankings (tink, mattr, deku).
really we should just do a single round-robin each tournament, best of 7, then record those as power rankings.
check more intervals: mms, seriouslies, round-robins, single elims.. enforce confrontation matches for ambiguous rankings (tink, mattr, deku).
really we should just do a single round-robin each tournament, best of 7, then record those as power rankings.
HoChiMinhTrail
Smash Master
I really like that idea, vro.
Also:
I would really like to do sort of an elo-style ranking system. Give everybody a skill number. This way each player could sort of gauge there improvement. And creating a leaderboard would be easy. (Put them in order duh.) A TO could also use it for creating balanced brackets.
I could try to do this. (Though it would be a **** ton of work.) The system I made kind of sucks balls. I'd rather do something closer to ELO ranking. The main problem is I need a ton of data. Like pool results, bracket images (perferably with win/loss record.) Considering how many brackets we've made on the back of beer cases and random pieces of paper, I don't know if it's doable.
Also:
I would really like to do sort of an elo-style ranking system. Give everybody a skill number. This way each player could sort of gauge there improvement. And creating a leaderboard would be easy. (Put them in order duh.) A TO could also use it for creating balanced brackets.
I could try to do this. (Though it would be a **** ton of work.) The system I made kind of sucks balls. I'd rather do something closer to ELO ranking. The main problem is I need a ton of data. Like pool results, bracket images (perferably with win/loss record.) Considering how many brackets we've made on the back of beer cases and random pieces of paper, I don't know if it's doable.
Cosmo!
nerf zelda's dsmash
I wrote this program that uses a custom Elo format (start at 1000, K factor is always 20).
http://www.speedrunslive.com/tools/testrecord/
If you want to give it a try.
http://www.speedrunslive.com/tools/testrecord/
If you want to give it a try.
Cosmo!
nerf zelda's dsmash
everybody starts with a number, say 1000
ratingA = 1000
ratingB = 1000
so then using this you can find the expected outcome of a game. it uses a logistic curve.
expectedA = 1 / (1 + 10^((ratingB - ratingA)/400)
expectedB = 1 / (1 + 10^((ratingA - ratingB)/400)
in this case both ratings are the same so you get
expectedA = 1 / (1 + 10^(0/400))
or
expectedA = 1 / (1 + 10^0)
or
expectedA = 1 / 2
so as you see if both ratings are the same you get an expected win of 0.5, because you have a 50% chance of winning or losing (according to the player ratings).
you take the expectedA and do this with it:
K*(actualA - expectedA)
for actual, you use 0 if you lost the match, 1 if you won the match
K is the "K factor" which is commonly used as a somewhat arbitrary way of letting people who are new to the system climb it quicker if they are good, or drop faster if they are bad. once you get higher up in ratings the K factor is supposed to drop so a loss would cost you less and a gain would gain you less as well.
because K is already arbitrary I prefer to skip the whole K factor switching and setting it to a fixed value like 20. 20 is really good when using 1000 as a starting base because the math works out really clean: check this out
lets say player A lost against player B.
20 * (0 - 0.5)
that comes out to -10
meanwhile playerB comes out to +10.
so that is the change. 1000 - 10 = 990 is the new rating for losing a match against an even opponent.
there are some flaws in this system that microsoft tried to fix with "TrueSkill" which is what they used with Xbox live. it not only keeps track of rating but also how "certain" the system is of the players skill. So a new player would have like a wider unknown range of skill. the math here is a ton more complex and probably not worth it, but the results would become more accurate at a quicker pace, while Elo takes awhile to propagate.
ratingA = 1000
ratingB = 1000
so then using this you can find the expected outcome of a game. it uses a logistic curve.
expectedA = 1 / (1 + 10^((ratingB - ratingA)/400)
expectedB = 1 / (1 + 10^((ratingA - ratingB)/400)
in this case both ratings are the same so you get
expectedA = 1 / (1 + 10^(0/400))
or
expectedA = 1 / (1 + 10^0)
or
expectedA = 1 / 2
so as you see if both ratings are the same you get an expected win of 0.5, because you have a 50% chance of winning or losing (according to the player ratings).
you take the expectedA and do this with it:
K*(actualA - expectedA)
for actual, you use 0 if you lost the match, 1 if you won the match
K is the "K factor" which is commonly used as a somewhat arbitrary way of letting people who are new to the system climb it quicker if they are good, or drop faster if they are bad. once you get higher up in ratings the K factor is supposed to drop so a loss would cost you less and a gain would gain you less as well.
because K is already arbitrary I prefer to skip the whole K factor switching and setting it to a fixed value like 20. 20 is really good when using 1000 as a starting base because the math works out really clean: check this out
lets say player A lost against player B.
20 * (0 - 0.5)
that comes out to -10
meanwhile playerB comes out to +10.
so that is the change. 1000 - 10 = 990 is the new rating for losing a match against an even opponent.
there are some flaws in this system that microsoft tried to fix with "TrueSkill" which is what they used with Xbox live. it not only keeps track of rating but also how "certain" the system is of the players skill. So a new player would have like a wider unknown range of skill. the math here is a ton more complex and probably not worth it, but the results would become more accurate at a quicker pace, while Elo takes awhile to propagate.
HoChiMinhTrail
Smash Master
wtf is goin on here
john!
Smash Hero
maybe for the tournaments that already happened we could use a higher k factor (40ish) to get everyone sorted faster. future tournies could use 20 though.
cosmo's program looks somewhat like what we would need, but it doesn't account for individual matchups, and isn't that what elo is based on? seems like it just uses placement to determine who would've beaten whom
cosmo's program looks somewhat like what we would need, but it doesn't account for individual matchups, and isn't that what elo is based on? seems like it just uses placement to determine who would've beaten whom
(which isn't always what happens, in my first melee tourney i beat sago but he placed higher than me, I'M STILL SALTY SAGO)
Cosmo!
nerf zelda's dsmash
the program I wrote was designed to handle races with multiple entrants
tournaments are different because you are directly comparing two players on a match per match basis. if you want to do some test results then simply set the "entrants" for each match to 2 and run it multiple times.
not a bad idea to use old data as a K factor of 40. (simply double the change to do this)
fwiw, here is data after 638 races for Super Mario 64, using my Elo setup:
1 sicko 1643
2 jiano 1608
3 Siglemic 1379
4 FunilaSM64 1364
5 hisu25 1306
6 jape 1302
7 SilentSlayers 1292
8 idunno 1278
9 TRV 1247
10 RaikerZ 1216
11 Gombill 1187
12 Swordless 1167
13 Runnerguy2489 1162
14 Manbearpig7 1141
15 ZFG 1113
16 CriminallyVulgar 1095
17 Axon 1076
18 ZAR 1076
19 Pidgey 1062
20 Torsak 1053
21 Losstarot 1051
22 sippu_sura 1032
23 joe 1020
24 kabu0404 1009
25 keeta 1005
26 FFGamer86 1005
27 octonion 1005
28 Tomatobird8 1000
29 Knighty 999
30 zalgo 996
31 pokey 992
32 EddieRuckus 984
33 zaccio 980
34 JMC47 979
35 Mp16 978
36 cyghfer 970
37 rex 970
38 Tranquilite 966
39 oeyvind 965
40 Perdy 965
41 Zozoken 965
42 Animeowzerz 962
43 maxx 962
44 Librari 961
45 Enterim 959
46 TheReaLildude322 958
47 Mapler90210 958
48 Chris 957
49 Venick409 956
50 toes 952
51 playe 941
52 PTS 937
53 Kostya 937
54 Allegretto 934
55 Patricklh 932
56 woawkoaw 932
57 SD2 931
58 bluemarth 931
59 Fuzzey_Ninja 931
60 Kirua 929
61 Johannes 925
62 FullRinse 925
63 GAG 914
64 EmilSM64 913
65 Cosmo 905
66 DJS 902
67 gnargl 886
68 Kyoki 879
69 Blechy 871
70 Zerst 863
71 JetSM64 861
72 Moltov 856
73 brak 856
74 ingx24 853
75 Blackoblivion 829
76 CVagts 821
77 darkeye 779
78 neskamikaze 762
79 muizzjkabani 717
80 thundrio 712
81 Elminster 706
82 zero 702
tournaments are different because you are directly comparing two players on a match per match basis. if you want to do some test results then simply set the "entrants" for each match to 2 and run it multiple times.
not a bad idea to use old data as a K factor of 40. (simply double the change to do this)
fwiw, here is data after 638 races for Super Mario 64, using my Elo setup:
1 sicko 1643
2 jiano 1608
3 Siglemic 1379
4 FunilaSM64 1364
5 hisu25 1306
6 jape 1302
7 SilentSlayers 1292
8 idunno 1278
9 TRV 1247
10 RaikerZ 1216
11 Gombill 1187
12 Swordless 1167
13 Runnerguy2489 1162
14 Manbearpig7 1141
15 ZFG 1113
16 CriminallyVulgar 1095
17 Axon 1076
18 ZAR 1076
19 Pidgey 1062
20 Torsak 1053
21 Losstarot 1051
22 sippu_sura 1032
23 joe 1020
24 kabu0404 1009
25 keeta 1005
26 FFGamer86 1005
27 octonion 1005
28 Tomatobird8 1000
29 Knighty 999
30 zalgo 996
31 pokey 992
32 EddieRuckus 984
33 zaccio 980
34 JMC47 979
35 Mp16 978
36 cyghfer 970
37 rex 970
38 Tranquilite 966
39 oeyvind 965
40 Perdy 965
41 Zozoken 965
42 Animeowzerz 962
43 maxx 962
44 Librari 961
45 Enterim 959
46 TheReaLildude322 958
47 Mapler90210 958
48 Chris 957
49 Venick409 956
50 toes 952
51 playe 941
52 PTS 937
53 Kostya 937
54 Allegretto 934
55 Patricklh 932
56 woawkoaw 932
57 SD2 931
58 bluemarth 931
59 Fuzzey_Ninja 931
60 Kirua 929
61 Johannes 925
62 FullRinse 925
63 GAG 914
64 EmilSM64 913
65 Cosmo 905
66 DJS 902
67 gnargl 886
68 Kyoki 879
69 Blechy 871
70 Zerst 863
71 JetSM64 861
72 Moltov 856
73 brak 856
74 ingx24 853
75 Blackoblivion 829
76 CVagts 821
77 darkeye 779
78 neskamikaze 762
79 muizzjkabani 717
80 thundrio 712
81 Elminster 706
82 zero 702
I've read a bit about true skill, and the system blizzard is using on B.net. ("rest" times.) Seems interesting but a bit too much work for right now.
I'm going to dive into the formula in a bit. (Gotta finish a few other things.) I'll post some calculations and thoughts once I got em.
I'm going to dive into the formula in a bit. (Gotta finish a few other things.) I'll post some calculations and thoughts once I got em.
jnolan
Smash Apprentice
Cosmo, your system seems like a nice base to start with... though it would need some editing. Just from looking at your JavaScript, it can easily be modified to do some neat things, for not just the MW PR, but for PR as a whole. I got nothing really going on tonight, so I am going to spend a while converting your system to a PHP based backend... going to include a SQL database with it, to store all the data. Aside from the basic functionality your system already has, some of the stuff I plan on adding are...
-Concrete 'Player' database. Allows the people running the PR to add/remove/edit players. Just a basic player database.
-User submitted results. Allows ANY player to submit results of a match. By default, it will be set so that the PR people have to 'approve' user submitted results. This will allow for say... PlayerA mm PlayerB to be entered by a player, but would require the PR people to then approve the match to be used in the calculations.
Another thing I am thinking about adding, is sorta a 'how bad did you loose' feature... basically, lets say PlayerA vs PlayerB. When entered into the system, you would start off by entering some 'info' about each game in the set (best out of, and then how many stocks started/ended, character choice for each match, stage for each match, etc). This will allow for a much more accurate rating system of an individual player... (for example, it will show that PlayerA is best with character X, then Y, then Z, etc, and what stage he is best on with each of those characters). This feature will not be that accurate at first (even the general player ranking wont be that accurate for 'close' players) but will of course gain accuracy as time goes on.
I will post an update, once I have something to show... feel free to post any feature you would like to see in this.
-Concrete 'Player' database. Allows the people running the PR to add/remove/edit players. Just a basic player database.
-User submitted results. Allows ANY player to submit results of a match. By default, it will be set so that the PR people have to 'approve' user submitted results. This will allow for say... PlayerA mm PlayerB to be entered by a player, but would require the PR people to then approve the match to be used in the calculations.
Another thing I am thinking about adding, is sorta a 'how bad did you loose' feature... basically, lets say PlayerA vs PlayerB. When entered into the system, you would start off by entering some 'info' about each game in the set (best out of, and then how many stocks started/ended, character choice for each match, stage for each match, etc). This will allow for a much more accurate rating system of an individual player... (for example, it will show that PlayerA is best with character X, then Y, then Z, etc, and what stage he is best on with each of those characters). This feature will not be that accurate at first (even the general player ranking wont be that accurate for 'close' players) but will of course gain accuracy as time goes on.
I will post an update, once I have something to show... feel free to post any feature you would like to see in this.
john!
Smash Hero
can it have tetris too? 
sounds neat. one thing i don't like though is the "how bad did you lose" feature. some people use the first game to find out the opponent's style and then dominate the next two games (armada did this at genesis a lot). then some people use characters that are really bad on some counterpicks, and can lose game 2 even if they are significantly better and win the other two games.
chess rankings (the most popular elo system) work the same way. they don't count the pieces after the game, because good players can sacrifice pieces to gain an advantage, or even get checkmate with less pieces than the opponent. the only thing that matters is victory or defeat.
but yeah it looks great. we would need a private server with limited access, though.
sounds neat. one thing i don't like though is the "how bad did you lose" feature. some people use the first game to find out the opponent's style and then dominate the next two games (armada did this at genesis a lot). then some people use characters that are really bad on some counterpicks, and can lose game 2 even if they are significantly better and win the other two games.
chess rankings (the most popular elo system) work the same way. they don't count the pieces after the game, because good players can sacrifice pieces to gain an advantage, or even get checkmate with less pieces than the opponent. the only thing that matters is victory or defeat.
but yeah it looks great. we would need a private server with limited access, though.
Woah that sounds cool.
Don't include how bad did you lose. Remaining stocks change from match-up to match-up. How big a lead is changes. That **** does not matter. Only do final result. Did you win the set or not?
Uh I'm collecting data right now. I need to know if this is a good list.
Tournaments (Going back about a year)
Active Series:
SGNG
SSBO
Ann Arbor/MI **** (Need names, must dig for SWEET results)
Inactive/small Series:
MAS
Vro Zones
One time events:
SMYM12
Pound 5
Trail's Livestream junk
GSS
Old ones:
Revival of WI
HDR
SCSYN3
POE3
ROM (Any MW go to this?)
APEX (Go vro, dj, lance, and I!)
Genesis 2? (Did anyone from MW go?)
Am I missing anything?
Don't include how bad did you lose. Remaining stocks change from match-up to match-up. How big a lead is changes. That **** does not matter. Only do final result. Did you win the set or not?
Uh I'm collecting data right now. I need to know if this is a good list.
Tournaments (Going back about a year)
Active Series:
SGNG
SSBO
Ann Arbor/MI **** (Need names, must dig for SWEET results)
Inactive/small Series:
MAS
Vro Zones
One time events:
SMYM12
Pound 5
Trail's Livestream junk
GSS
Old ones:
Revival of WI
HDR
SCSYN3
POE3
ROM (Any MW go to this?)
APEX (Go vro, dj, lance, and I!)
Genesis 2? (Did anyone from MW go?)
Am I missing anything?
Tomacawk
Smash Master
I can't believe how much you guys care about getting the rankings accurate
I'm going to shamelessly sit around and let everybody figure it out and then talk **** about how it's a stupid system when it's done and how I can do better
I'm going to shamelessly sit around and let everybody figure it out and then talk **** about how it's a stupid system when it's done and how I can do better
Fortress | Sveet
▀▄▀▄▀▄▀▄▀▄▀▄▀▄▀▄▀
Sounds cool. I might take some time to put that into a program so we could have an ELO system for the MW powerrankings.
Rat, when you're done digging up results, are you going to post them? If so, include bracket images when possible.
Rat, when you're done digging up results, are you going to post them? If so, include bracket images when possible.
john!
Smash Hero
Fortress | Sveet
▀▄▀▄▀▄▀▄▀▄▀▄▀▄▀▄▀
Trail is one of the biggest nerds i know. Dude spends more time on SC2 than me, and i haven't closed the window in 2 weeks.
Tomacawk
Smash Master
Has the pun brochiminhtrail been made before?
HoChiMinhTrail
Smash Master
where did this come from lol? **** you sveet im gonna come up there and whoop that *** in smash one of these days
*butt hurt*
lolll....this mathematical list is gonna fro me on placing and when it does, id like to be taken off of it :/.
also, trail+cunning will not be placed accurately in the least, when they should be very high on the list :/
lolll....this mathematical list is gonna fro me on placing and when it does, id like to be taken off of it :/.
also, trail+cunning will not be placed accurately in the least, when they should be very high on the list :/
jnolan
Smash Apprentice
Yeah, I am basing my system off the old chess ranking system I wrote for my HS years ago.
I do see your point on the 'how bad' thing, as you are right.
What do you guys think about the 'character' and 'stage' option though? I will most likely be having those be purely optional fields, however I feel it would be a neat addition.
Anyway...
So far I have the basics of the SQL database set up (the matches table, and the players table) and some of the core functionality done (adding/updating matches, adding/updating players)... of course, this is all done through functions right now, and the system has 0% GUI interface. I will probably add some crappy one after all the core is done, however don't expect it to look pretty. I am NOT an art person.
---
Also Tink, this is NOT going to eliminate the possibility of a "User generated" PR. Think about it with ANY ranking system (sports, chess, etc)... there are always situations when PlayerA has a lower rank than PlayerB, but the general populous believes that PlayerA is better.
I do see your point on the 'how bad' thing, as you are right.
What do you guys think about the 'character' and 'stage' option though? I will most likely be having those be purely optional fields, however I feel it would be a neat addition.
Anyway...
So far I have the basics of the SQL database set up (the matches table, and the players table) and some of the core functionality done (adding/updating matches, adding/updating players)... of course, this is all done through functions right now, and the system has 0% GUI interface. I will probably add some crappy one after all the core is done, however don't expect it to look pretty. I am NOT an art person.
---
Also Tink, this is NOT going to eliminate the possibility of a "User generated" PR. Think about it with ANY ranking system (sports, chess, etc)... there are always situations when PlayerA has a lower rank than PlayerB, but the general populous believes that PlayerA is better.
ya, thats why i dont like it all mostly on results....ppl like vro who just play terribly in tournament, but are soo amazing at the game are gonna get rolled :/.
edit: i also didnt read any of that wall of text cosmo rained down....so if theres a way around this, then cool.
edit: i also didnt read any of that wall of text cosmo rained down....so if theres a way around this, then cool.
jnolan
Smash Apprentice
There is, and there isn't... at least with my system. My system is going to allow for players to submit matches (friendlies, MMs, Pools, etc) not just the people running the PR. These results will not be calculated right off the bat (player submitted matches will be 'pending' and require a PR person to approve it, for it to be calculated).
So even if someone just always does bad in tourny, but does will in MMs, non-tourney serious matches, etc... all they would have to do is submit them to the system, and have them approved. Yes, it does require some work on both the players and PR people for this, but it will allow for a much more accurate rating system.
Also, based on the 'type' of match (tourney, mm, serious, friendly, etc) there could be different 'weights' assigned. But that would require some talk between the PR people, as to decide if they wanted those weights, and if so, what they would be.
Anyway, if anyone here knows PHP/SQL (or web design, for a nice, pretty, GUI) either post here, or send me a PM. This could be a VERY big project if done right (I am already thinking about allowing it to expand aside from the Chicago-Land PR... although you would have the ability to 'minimize' your range of who you were calculating for rankings and such... but that's all in the future) so any help would be nice.
So even if someone just always does bad in tourny, but does will in MMs, non-tourney serious matches, etc... all they would have to do is submit them to the system, and have them approved. Yes, it does require some work on both the players and PR people for this, but it will allow for a much more accurate rating system.
Also, based on the 'type' of match (tourney, mm, serious, friendly, etc) there could be different 'weights' assigned. But that would require some talk between the PR people, as to decide if they wanted those weights, and if so, what they would be.
Anyway, if anyone here knows PHP/SQL (or web design, for a nice, pretty, GUI) either post here, or send me a PM. This could be a VERY big project if done right (I am already thinking about allowing it to expand aside from the Chicago-Land PR... although you would have the ability to 'minimize' your range of who you were calculating for rankings and such... but that's all in the future) so any help would be nice.
Little England
Smash Master
This is becoming waaaaay too convoluted. It's hella easy to just look at brackets/pools and see who beat/lost to who and make this thing.
jnolan
Smash Apprentice
In terms of a simple ~15 person PR, with events only being added maybe once a month or so, you are right.
However, minus the fact that I am bored right now and have nothing better to do, tools such as these can be useful for the community as a whole. It allows for easy expansion (in terms of # of players being examined, # of events being calculated into it, # of matches over all) and helps get rid of some of the 'vagueness' between two very closely ranked players.
Cosmo!
nerf zelda's dsmash
i dont really see the purpose of a database. the person who is handling this (rat) could simply save the list in a txt file. setting up a system where people could submit matches or w/e then get them approved later is indeed convoluted.
jnolan
Smash Apprentice
Again, for a simple PR, there is no real use for a database, other than removal of the need to manually calculate the rankings. What I am working on is not just a simple list of the "Top X players of a region" however. The use of a database in my system is to allow for player tracking over time (overall, against specific other players, in specific match ups, etc).
As for the player-submitted matches, it isn't even that complacated, and allows for not only more data to sample from, but also data that wouldn't normally be figured in.
Using purely tournament placement results, only gives you the ranking of how well people do in tournament. It totally excludes out of tournament matches. Sure, the PR people can manually track down these results and add them to the ranking if they choose to, but a system such as this takes that task away from the PR people, and into the hands of the players.
So again, for just a simple ranking system of the top X players in a region, there is no need for anything other than a pencil, paper, and calculator. However for a much more indept 'ladder'-esk system, you can't just get by with those 3 things.
As for the player-submitted matches, it isn't even that complacated, and allows for not only more data to sample from, but also data that wouldn't normally be figured in.
Using purely tournament placement results, only gives you the ranking of how well people do in tournament. It totally excludes out of tournament matches. Sure, the PR people can manually track down these results and add them to the ranking if they choose to, but a system such as this takes that task away from the PR people, and into the hands of the players.
So again, for just a simple ranking system of the top X players in a region, there is no need for anything other than a pencil, paper, and calculator. However for a much more indept 'ladder'-esk system, you can't just get by with those 3 things.
Fortress | Sveet
▀▄▀▄▀▄▀▄▀▄▀▄▀▄▀▄▀
For some reason i doubt it roy. You always john out for one reason or another lol
Scythe
Smash Master
- Joined
- Dec 5, 2004
- Messages
- 3,875
Yeah isn't this how the PR is originally made?
i don't really remember to well but I don't recal any drama over placings that weren't related to not being from IL
john!
Smash Hero
scythe with dat triple negative
Dart!
Smash Master
Genesis 2 hasn't happened yet.
Sago
Smash Lord
http://www.youtube.com/watch?v=5y8DTH13eEw (sick new marth vid by sago)
yo im a real melee player now........lol
yo im a real melee player now........lol
Dart!
Smash Master
where's dat green greens combo?