|
On May 12 2013 02:01 bilivaxyto wrote:
Forgg being that high actually shows how bad the alligulac rating system is for sc2. also they need higher K-factor since sc2 is a game that evolves quickly. It shouldnt take 200 games and 1 year for a good player to reach the top
They used to have a higher K-factor
and then people complained about the ratings being too volatile
so they tuned it down.
|
On May 12 2013 01:35 Grovbolle wrote:Show nested quote +On May 12 2013 01:27 achan1058 wrote:On May 12 2013 00:38 hastur420 wrote:On May 09 2013 18:51 Greenei wrote:On May 05 2013 20:02 hastur420 wrote:On May 05 2013 07:02 Greenei wrote:On May 05 2013 05:46 Nerchio wrote:
If a guy that doesn't scout in more than 50% of the games gets even close to rank1 it means something is wrong with your ranking. Or maybe you should read the fucking OP... or maybe one could think even if there is an explanation for it, that the rating is worthless if such a thing happens funny how dorks jump on the opportunity to be an ass with a progamer just because they can No it's not worthless. It's the appliciants fault if he does not understand how to properly use a tool. Also obviously scouting has not much to do with a rating. See MKPs dominance. Or maybe I don't give a shit if someone is a progamer. lol what "proper use"? proper use is to produce useless numbers?  Aligulac gives predictions as well, just like ELO (even if you don't know about that). Those predictions are surprisingly accurate, as long as you do not compare cross regions. There's probably no easy answer to the ForGG problem, other than "subjective" analysis. Edit: In fact, ELO suffers from a similar, but different, problem. Comparing across time period is useless due to grading inflation. The formula behind the rating is optimized with regard to predictions, to predict as many results correctly as possible
Hey, really cool project you have going there, although I can only imagine how much these "your stats are worthless cause my favourite player isn't #1" posts must annoy you.
About the ForGG (and Polt) situation, pardon me if I'm repeating something suggested earlier, but have you considered trying to formalize the intuition we all share - that regions aren't created equal? Blizzard used to have something like that for Masters League, where divisions had different elo modifiers, although in their case they just decided on them arbitrarily.
So, maybe you could cluster your players (should result in geographical regions with some outliers), calculate the average rating for each region, and modify that for cross-region matches? So, for instance, if a Code B Korean play in an EU tournament and beats ForGG, this would not only affect their ratings, but also the average ratings of EU and KR, so the rating of every EU player would decrease slighly, and the rating of every KR player would increase slightly (I assume ratings aren't actually integers). I believe this could capture the intuition that we all have: That Koreans beating Europeans should devalue rating gained by playing in EU only, which would solve the ForGG problem.
|
While the crowds were crying in fear for the ForGGeddeon, they failed to realize the impending apocalypse happening right in front of their eyes. Yet they refused to acknowledge it, they were blind to see the other cataclysm. Behold, Poltgeddeon has arrived.
|
Hwat da heck Polt. GJ on your qualifier lol 
|
On May 12 2013 12:02 Toadvine wrote:Show nested quote +On May 12 2013 01:35 Grovbolle wrote:On May 12 2013 01:27 achan1058 wrote:On May 12 2013 00:38 hastur420 wrote:On May 09 2013 18:51 Greenei wrote:On May 05 2013 20:02 hastur420 wrote:On May 05 2013 07:02 Greenei wrote:On May 05 2013 05:46 Nerchio wrote:
If a guy that doesn't scout in more than 50% of the games gets even close to rank1 it means something is wrong with your ranking. Or maybe you should read the fucking OP... or maybe one could think even if there is an explanation for it, that the rating is worthless if such a thing happens funny how dorks jump on the opportunity to be an ass with a progamer just because they can No it's not worthless. It's the appliciants fault if he does not understand how to properly use a tool. Also obviously scouting has not much to do with a rating. See MKPs dominance. Or maybe I don't give a shit if someone is a progamer. lol what "proper use"? proper use is to produce useless numbers? Aligulac gives predictions as well, just like ELO (even if you don't know about that). Those predictions are surprisingly accurate, as long as you do not compare cross regions. There's probably no easy answer to the ForGG problem, other than "subjective" analysis. Edit: In fact, ELO suffers from a similar, but different, problem. Comparing across time period is useless due to grading inflation. The formula behind the rating is optimized with regard to predictions, to predict as many results correctly as possible Hey, really cool project you have going there, although I can only imagine how much these "your stats are worthless cause my favourite player isn't #1" posts must annoy you. About the ForGG (and Polt) situation, pardon me if I'm repeating something suggested earlier, but have you considered trying to formalize the intuition we all share - that regions aren't created equal? Blizzard used to have something like that for Masters League, where divisions had different elo modifiers, although in their case they just decided on them arbitrarily. So, maybe you could cluster your players (should result in geographical regions with some outliers), calculate the average rating for each region, and modify that for cross-region matches? So, for instance, if a Code B Korean play in an EU tournament and beats ForGG, this would not only affect their ratings, but also the average ratings of EU and KR, so the rating of every EU player would decrease slighly, and the rating of every KR player would increase slightly (I assume ratings aren't actually integers). I believe this could capture the intuition that we all have: That Koreans beating Europeans should devalue rating gained by playing in EU only, which would solve the ForGG problem.
That's a lot of fancy words.
I am not the "brain" behind the site, that would be TheBB, he is currently on vacation, but I'll make sure he sees this and gives you an answer 
Clustering players would be really hard though, Right now the biggest problem is that we are still waiting for Kespa players to have enough points transferred to their "region" to make their rating more accurate compared to their actual skill level. The Kespa players made a huge splash because a lot of great players entered with a 1000 rating, and they are obviously a lot better than their rating currently shows, because some players almost never play, so we never know how good they are.
All in all, you can never take statistics at face value, do I believe ForGG is top 3 player in the world? No, I do not. Currently TheBB is actually working on a new underlying model trying to improve the overall rating system.
|
On May 12 2013 01:03 sAsImre wrote:Show nested quote +On May 12 2013 00:38 hastur420 wrote:On May 09 2013 18:51 Greenei wrote:On May 05 2013 20:02 hastur420 wrote:On May 05 2013 07:02 Greenei wrote:On May 05 2013 05:46 Nerchio wrote:
If a guy that doesn't scout in more than 50% of the games gets even close to rank1 it means something is wrong with your ranking. Or maybe you should read the fucking OP... or maybe one could think even if there is an explanation for it, that the rating is worthless if such a thing happens funny how dorks jump on the opportunity to be an ass with a progamer just because they can No it's not worthless. It's the appliciants fault if he does not understand how to properly use a tool. Also obviously scouting has not much to do with a rating. See MKPs dominance. Or maybe I don't give a shit if someone is a progamer. lol what "proper use"? proper use is to produce useless numbers? they show something, but the method to calculate the points is not perfect (not technically, it's probably very well written, but the idea itself), hence we have a flawed list where forgg is on top. there is no proper use of this webpage to produce a better list. and btw MKP never dominated as much, and it was inevitable that he falls with such an obvious flaw And btw if you think that MKP's fall has anything to do with its scouting, you're just blatlantly showing your ignorance of the game.
ohh hyea you're so unignorant you just proved MKP's stubborn no scout into greedy builds didn't get punished by 124213432 zerg players after his "dominance" mainly causing his fall. you're like the king of arguments.
|
ah yes, the good o' cycle of "this list is dumb" vs "go learn aligulac system" ^_^
|
On May 14 2013 04:06 jinorazi wrote:
ah yes, the good o' cycle of "this list is dumb" vs "go learn aligulac system" ^_^
It's the circle of Startale Life.
|
On May 14 2013 04:20 Grovbolle wrote:Show nested quote +On May 14 2013 04:06 jinorazi wrote:
ah yes, the good o' cycle of "this list is dumb" vs "go learn aligulac system" ^_^ It's the circle of Startale Life.
It's genesis of the Nestea.
|
Casters just said "fOrGGeddon" on WCS EU stream.
|
On May 15 2013 04:16 jalstar wrote:Casters just said "fOrGGeddon" on WCS EU stream.
I have been plugging it everywhere, it is now a thing.
|
On May 12 2013 12:02 Toadvine wrote:Show nested quote +On May 12 2013 01:35 Grovbolle wrote:On May 12 2013 01:27 achan1058 wrote:On May 12 2013 00:38 hastur420 wrote:On May 09 2013 18:51 Greenei wrote:On May 05 2013 20:02 hastur420 wrote:On May 05 2013 07:02 Greenei wrote:On May 05 2013 05:46 Nerchio wrote:
If a guy that doesn't scout in more than 50% of the games gets even close to rank1 it means something is wrong with your ranking. Or maybe you should read the fucking OP... or maybe one could think even if there is an explanation for it, that the rating is worthless if such a thing happens funny how dorks jump on the opportunity to be an ass with a progamer just because they can No it's not worthless. It's the appliciants fault if he does not understand how to properly use a tool. Also obviously scouting has not much to do with a rating. See MKPs dominance. Or maybe I don't give a shit if someone is a progamer. lol what "proper use"? proper use is to produce useless numbers? Aligulac gives predictions as well, just like ELO (even if you don't know about that). Those predictions are surprisingly accurate, as long as you do not compare cross regions. There's probably no easy answer to the ForGG problem, other than "subjective" analysis. Edit: In fact, ELO suffers from a similar, but different, problem. Comparing across time period is useless due to grading inflation. The formula behind the rating is optimized with regard to predictions, to predict as many results correctly as possible Hey, really cool project you have going there, although I can only imagine how much these "your stats are worthless cause my favourite player isn't #1" posts must annoy you. About the ForGG (and Polt) situation, pardon me if I'm repeating something suggested earlier, but have you considered trying to formalize the intuition we all share - that regions aren't created equal? Blizzard used to have something like that for Masters League, where divisions had different elo modifiers, although in their case they just decided on them arbitrarily. So, maybe you could cluster your players (should result in geographical regions with some outliers), calculate the average rating for each region, and modify that for cross-region matches? So, for instance, if a Code B Korean play in an EU tournament and beats ForGG, this would not only affect their ratings, but also the average ratings of EU and KR, so the rating of every EU player would decrease slighly, and the rating of every KR player would increase slightly (I assume ratings aren't actually integers). I believe this could capture the intuition that we all have: That Koreans beating Europeans should devalue rating gained by playing in EU only, which would solve the ForGG problem.
Hey,
You're right, there are some problems related to heterogeneity in the scene. I could potentially solve it by introducing some region-specific parameters. There is just one problem. I basically can't decide what the ratings are for . Are they for predictions only? In which case I should go ahead. Or is it also a "competition" where I need to be concerned with fairness? In which case I can't.
For example, chess Elo is used for seeding in the World Championships. That would never be possible if Russians, say, were given bonus points, no matter how mathematically justified.
Of course, Aligulac isn't being used for any such thing, but I have some respect for the integrity of the system. Right now, Koreans have exactly as many points as they deserve to have, mathematically, based on the games they've won. At least that's something.
I realize this doesn't have a right answer. Just pondering...
|
47-3. at the moment. 94% win rate. can he make it a 48-3?
|
|
|
|
|
I'm scared, somebody hold me.
|
On May 15 2013 06:02 Kasaraki wrote:
I'm scared, somebody hold me.
Well, let's see if he actually gets #1 :D
|
I used a college football rating system on all of the WCS-counted games (so all of WCS and dreamhack)
It can be found at http://wilson.engr.wisc.edu/perform/ and I was just interested in a more specific alternative to aligulac.
It returned: (only looking at players with 6+ wins in events that give out WCS points)
+ Show Spoiler +
|
ForGG looking good against MMA today in WCS
|
Looks like KingKong is the Zerg version of ForGG because of his results from the insulated SEA scene. He's ranked surprisingly high. :D
|
|
|
|
|
|
EPT![[image loading]](http://puu.sh/2TwUQ.jpg)
![[image loading]](http://puu.sh/2Tx1p.jpg)
