GSTL is a massive indicator of who is good. This is where we saw MMA and DRG on the rise long before their code S and blizzard cup victories. I don't think you can have an accurate representation without some influence of the GSTL.

I know its tough to add, but it's too big of an indicator of skill to overlook.

minilance

Canada500 Posts

August 07 2012 14:24 GMT

#62

I would had a regression systeme similar the the Power rank in BW.

TeMiL

Peru545 Posts

August 07 2012 14:39 GMT

#63

maybe a ranking system point based on the participants should be more exactly but difficult to make it.

FuzzyJAM

Scotland9300 Posts

August 07 2012 14:41 GMT

#64

On August 07 2012 23:05 Metalteeth wrote:
3) To people complaining about the tournament ratings, some of it is valid, some of it is useless: "I don't like the point values!" Well that is nice, what would you change?
It seems most people think GSL should be worth more. But as someone said in here, if it gets too high, the system will overvalue mid-level Koreans who do well in GSL and nothing else. But too low, and GSL success isn't worth the difficulty that is has. What I'm thinking is bump GSL up to 4000 points for winning it. Keep MLG/IPL/Dreamhack/ASUS/IEM WC/etc (the non-GSL Grand Slams) at 2000. Drop IEM tour stops to 1500. Increase last year's Dreamhack 8 person invites to 1500. Keep everything else the same, for now.

"Overvaluing" mid-level Koreans isn't possible - if you're in Code S, you're essentially guaranteed to be better than all but one of the players not in the GSL. It's these very same "mid-level Koreans" that dominate foreign events when the top ones don't go. Violet and Puma in the top 10 rather prove that.

NightOfTheDead

Lithuania1711 Posts

August 07 2012 14:46 GMT

#65

GSL should bear double if not triple weight, no offense but Killer(Chille) was never better than Squirtle in a year.

Metalteeth

United States115 Posts

August 07 2012 14:52 GMT

#66

On August 07 2012 23:46 NightOfTheDead wrote:
GSL should bear double if not triple weight, no offense but Killer(Chille) was never better than Squirtle in a year.

That was before regression. Killer (Chile) had last years Latin American BattleNet Invite, some IEM success, then he won WCS Chile and WCS South America. Those last two were 1500 points.

Aeceus

United Kingdom1278 Posts

August 07 2012 15:00 GMT

#67

Good idea but the scoring is wrong. WCS continentals should be higher than IEMs and MLG arenas

FireFish

Denmark228 Posts

August 07 2012 15:03 GMT

#68

I personally think this looks great

X3GoldDot

Malaysia3840 Posts

August 07 2012 15:07 GMT

#69

gsl needs like 1.5x the points lol, squirtle and parting aint even in the top 30, and creator isnt even in top 100 lol. also there is absolutely no way in hell does stephano seem better than MKP, hero, taeja etc. Flaws are kinda apparent with huk and violet in the top 10 rofl. Good ranking system tho, this has the potential to be better alot tho.

Naphal

Germany2099 Posts

August 07 2012 15:16 GMT

#70

i would just take prize money as the indicator if you want to go that far, but goody is top100 according to your list so all is well.

Metalteeth

United States115 Posts

August 07 2012 15:17 GMT

#71

OK, made some changed. Added a regression (see a previous comment for the formula, go Gaussian statistics!), and changed around some of the tournament values. GSL got a boost, IEM was dropped a little. Please see OP for full everything, but Here is a revised Top 10.

1 (P)

MC 9860.60
2 (Z)

DongRaeGu 9321.96
3 (Z)

Stephano 6354.38
4 (P)

MarineKing 5889.53
5 (T)

Mvp 5561.02
6 (T)

MMA 5444.66
7 (T)

TaeJa 4881.37
8 (P)

Violet 4204.42
9 (T)

PuMa 4064.23
10 (Z)

Nerchio 3681.95

Kal_rA

United States2925 Posts

August 07 2012 15:17 GMT

#72

Something centralized like Kespa rankings from BW would be great!

This whole thread is moot anyways. We all know its the TL Power Rank that really matters.

Kovaz

Canada233 Posts

August 07 2012 15:24 GMT

#73

I think you'd almost have to separate the korean standings from the foreign standings. There are so few koreans who can afford to travel to a bunch of tournaments, and the reverse is also true: there's no more than a handful of foreigners who are able to go to korea and compete in gsl/gstl and such. You'd have to come up with some points system that includes gstl and korean weeklies and such, and then you could use the rankings you have now minus gsl as the foreigner ranking.

I just feel like the ranking loses any meaning past the top handful of players that can play in GSL and foreign tournaments. Mana is ranked higher than Nestea, even though Mana hasn't competed in gsl and nestea rarely leaves korea. The purpose is to determine who is playing objectively better at that moment, but these come down to your subjective relative weighting.

That said, I think a ranking system like this is awesome to have. It'd be really cool to have a #1 ranking that players strive for that actually means something, and tournaments could start using it for a more fair seeding system. It would be cool to see who's really the top foreigner according to the stats, and there could be interesting storylines like Roger Federer at Wimbledon a few weeks ago reclaiming his world #1 ranking. Imagine Mvp winning another GSL to get back on top after a year of struggle with his wrists. Imagine the excitement when some no-name upsets the #1 ranked player in straight sets (a la Bisu over sAviOr).

NonY

8748 Posts

August 07 2012 15:26 GMT

#74

I don't think the system could ever work because so many players don't even try so many leagues. It's not fair to a ton of non-Korean players that the GSL has any weight at all because it's not a viable tournament to play without taking up residence near its location. The GSL isn't hard just because it has good players. It's hard because you have to move to a foreign country that speaks virtually no English just to participate. And then you've got to pay more and travel longer to continue to participate in NA/EU tournaments.

I understand rankings in other sports work similarly in the sense that a player can forgo practice or rest to participate in more competitions in an attempt to bolster ranking. But the situation for SC2 is way more extreme and I believe impossible to solve.

Firebolt145

Lalalaland34490 Posts

August 07 2012 15:28 GMT

#75

Idra at 1337 points eh.

opterown

Australia54784 Posts

August 07 2012 15:30 GMT

#76

On August 08 2012 00:17 Metalteeth wrote:
Ok, made some changed. Added a regression (see a previous comment for the formula, go Gaussian statistics!), and changed around some of the tournament values. GSL got a boost, IEM was dropped a little. Please see OP for full everything, but here is a revised Top 10.

1 MC 9860.60
2 DongRaeGu 9321.96
3 Stephano 6354.38
4 MarineKing 5889.53
5 MVP 5561.02
6 MMA 5444.66
7 Taeja 4881.37
8 viOLet 4204.42
9 PuMa 4064.23
10 Nerchio 3681.95

looks a bit better than before! :D

nkr

Sweden5451 Posts

August 07 2012 15:32 GMT

#77

IdrA with 1337 points, not surprising.

lichter

1001 YEARS KESPAJAIL22272 Posts

August 07 2012 15:34 GMT

#78

There has to be a more precise/mathematical/logical basis to the way tournaments are given points.

The different tournament formats should also make a difference in the way points are allocated. For example, knockouts are usually thought to be more difficult than group stages, while the presence of a loser's bracket allows second chances (leniency). Going undefeated in a 64-man seeded knockout tournament is harder than a 2-groupstage then knockout tournament, which I think is again more difficult than a tournament with loser's brackets. Of course this is again opinion; there must be a way to determine how these tournament formats make a difference to difficulty.

Anyway, as for a better way to determine points...

These things should matter:
Difficulty of player pool (while this is subjective, giving a difficulty range should be enough to mitigate most opinion)
Size of player pool
Size of prize pool
Estimated viewership numbers (again in a range rather than specific)
Subjective Prestige (hard not to consider this)
Invite only or Qualifiers (invite only should matter less, imo)
Regional Difficulty Quotient (sorry, not all regions are equal)

There are probably more. I am typing this in the middle of a hurricane so I am too bothered to try to make it more detailed T_T

Freeze967

United States230 Posts

August 07 2012 15:35 GMT

#79

What's weird about this is that different people have different "styles" of tournaments. Players like MVP and NesTea are amazing at GSL style (knowing your opponent much in advance) so they can prepare for them. But then if you look at their success in team leagues, they crumble. They aren't nearly as good playing without tons of practice vs that one person. If there was some way to represent that, then this might work.

ZXRP

South Africa114 Posts

August 07 2012 15:36 GMT

#80

As many people have noted, the subjectivity in the assignment of tournament weights is the problem. As far as I can see, the system chess uses to rate tournaments would be the most naturally applicable. That is, calculate the average ELO for an event, and then assign a "category" value to that tournament. For example, the very strongest tournaments are category 21 events, where the average ELO ranges from 2751-2775.

This makes much more sense than the blanket assignment of weights to a tournament, especially when one considers that the same tournament could differ greatly in the strength of its participants over several seasons.

A salient point here is that, if one trusts the ELO enough to assign the tournament value based off of it, then why not simply use the ELO in the first place to determine rankings? I would argue that we should stick to the ELO system, although I believe we should amalgamate the "international" and "korea" list to have one, unambiguous reference for who is the current best player.

As a last remark, it could be argued that one should have different ELO "types" based on the type of tournament held. For instance, chess has different rankings for different time controls. The classical time control ELO differs from the rapid and blitz lists - although the intersection is clearly non-zero. The SC2 equivalent of classical time controls might be something like the GSL, where one has no less than one week to prepare for each match. The equivalent of rapid time controls could be something like an MLG, where one has hours (or indeed even minutes) to prepare.

tl;dr - The Chess system is a better model for SC2 rankings than the Tennis model.