|
Public Service AnnouncementUse the Champion threads whenever appropriate. Don't use General Discussion simply out of ease. ===== If you want to whine about server lag, use the QQ thread. We all suffer alike when Riot servers kaput. No need to make a post about it in GD. |
Russian Federation4235 Posts
On January 27 2012 02:59 Slayer91 wrote:
Chaox, Yellowpete, Entenzwerg.
Ty.
Btw, on the ELO topic, the original system uses the logistic distribution to represent player performance, it's not significantly different from normal distribution (all ratings use bell-shaped curves to represent individual skill and/or performance in fact). However, there are two things to mention here:
1) ELO is not in any way ready for individual rankings in a team-based game. It's most likely that Riot uses a home-brewn adaptation of ELO to team-based games, or even their own rating system (based on the ideas from Glicko maybe) and calling it ELO for convenience.
2) Player distribution among rating levels is an experimentally measured curve which should not in the general case be gaussian. The shape of that curve depends on the rating system used. What all ratings measure is the average of player performance over some timespan, well, it's pretty obvious that it's a combo of at least two factors - natural talent and experience in that particular game (in fact, many more factors), it's not guaranteed to be a sum of those (and you get normal distributions only in case of summing lots of independent random variables). It might be a normal distribution or it might be not, noone knows but the guys who actually see it in their databases.
|
I wonder if a reduction in elo chance from a win/loss would help spread out the mid ranges.
Like half the elo gain/loss causing people to spike up and down a lot less. Its natural for peoples elo to go up and down but a 5 game loss streak which isnt all that uncommon can spiral someones elo from 1300 to 1200 rather quick.
In a way this is also how SC2 handles it cause you need to go on significant losing streak to be demoted and ofc likewise for promoting.
Less change in elo leads to less volatile ranking and helps give people more of an idea that they belong at the point they are at?
|
On January 25 2012 06:32 jacosajh wrote:Show nested quote +On January 25 2012 06:27 Shiv. wrote:
What strikes me as odd is that they are investing resources into their bot AI while stuff like replays still hasn't been released yet. I agree. They did say that the majority of players who play often play against bots every week though. This is where stats might be extremely deceptive? I go into bots every now and then to try new jungle routes and new champions. I'm sure many do too? But I guess, like any company, they are in a position where they have to allocate resources to pleasing current customers or trying to get more customers. I would, however, like to see more features implemented like replays (where you can rewind) and better spectating (more than 2 and spectate while matchmaking). I don't want Magma Chamber, moar champions, nerfs/buffs -_- We might never see replays because of difficulty implementing with the game engine though. And spectating options are probably going to be a server nightmare. AFAIK, Riot only has one US server location on the west coast, which is why us east coast brethren have like ~80 ping minimum.
Where are you at on the West Coast? I normally get 15 ping and my friend gets about 30(we live 40mins apart, he's in LA).
|
On January 27 2012 03:09 Gorsameth wrote:
I wonder if a reduction in elo chance from a win/loss would help spread out the mid ranges.
Like half the elo gain/loss causing people to spike up and down a lot less. Its natural for peoples elo to go up and down but a 5 game loss streak which isnt all that uncommon can spiral someones elo from 1300 to 1200 rather quick.
In a way this is also how SC2 handles it cause you need to go on significant losing streak to be demoted and ofc likewise for promoting.
Less change in elo leads to less volatile ranking and helps give people more of an idea that they belong at the point they are at?
I don't think so, because another problem is that people find it very slow to climb to their "true" (stable) rating. The problem is just too much noise in the system, and that can't be solved no matter how fancy your rating system is without more data.
|
On January 27 2012 03:12 Mondeezy wrote:Show nested quote +On January 25 2012 06:32 jacosajh wrote:On January 25 2012 06:27 Shiv. wrote:
What strikes me as odd is that they are investing resources into their bot AI while stuff like replays still hasn't been released yet. I agree. They did say that the majority of players who play often play against bots every week though. This is where stats might be extremely deceptive? I go into bots every now and then to try new jungle routes and new champions. I'm sure many do too? But I guess, like any company, they are in a position where they have to allocate resources to pleasing current customers or trying to get more customers. I would, however, like to see more features implemented like replays (where you can rewind) and better spectating (more than 2 and spectate while matchmaking). I don't want Magma Chamber, moar champions, nerfs/buffs -_- We might never see replays because of difficulty implementing with the game engine though. And spectating options are probably going to be a server nightmare. AFAIK, Riot only has one US server location on the west coast, which is why us east coast brethren have like ~80 ping minimum. Where are you at on the West Coast? I normally get 15 ping and my friend gets about 30(we live 40mins apart, he's in LA).
Psh, I get average 9ping, and if I'm lucky I hit 8 ping. University internet ftw!
|
On January 27 2012 03:17 wei2coolman wrote:Show nested quote +On January 27 2012 03:12 Mondeezy wrote:On January 25 2012 06:32 jacosajh wrote:On January 25 2012 06:27 Shiv. wrote:
What strikes me as odd is that they are investing resources into their bot AI while stuff like replays still hasn't been released yet. I agree. They did say that the majority of players who play often play against bots every week though. This is where stats might be extremely deceptive? I go into bots every now and then to try new jungle routes and new champions. I'm sure many do too? But I guess, like any company, they are in a position where they have to allocate resources to pleasing current customers or trying to get more customers. I would, however, like to see more features implemented like replays (where you can rewind) and better spectating (more than 2 and spectate while matchmaking). I don't want Magma Chamber, moar champions, nerfs/buffs -_- We might never see replays because of difficulty implementing with the game engine though. And spectating options are probably going to be a server nightmare. AFAIK, Riot only has one US server location on the west coast, which is why us east coast brethren have like ~80 ping minimum. Where are you at on the West Coast? I normally get 15 ping and my friend gets about 30(we live 40mins apart, he's in LA). Psh, I get average 9ping, and if I'm lucky I hit 8 ping. University internet ftw!
I'm going to take a guess and say UCI?
Also, all this talk about ELO and having good mechanics... I feel like a lot of people have the mechanics down but it really comes down to how people play in a teamfight. Knowing when to engage and when you can be aggressive makes a lot of the fight. That or just play an AP that can gimp the other person in one burst (my favorite being Brand).
|
Two of the best EU teams playing against each other (CLG.eu vs M5) on Wickd stream : 800 viewers
Reginald playing normal games ; 7k viewers.
Wtf ?
|
On January 27 2012 03:25 wussleeQ wrote:Show nested quote +On January 27 2012 03:17 wei2coolman wrote:On January 27 2012 03:12 Mondeezy wrote:On January 25 2012 06:32 jacosajh wrote:On January 25 2012 06:27 Shiv. wrote:
What strikes me as odd is that they are investing resources into their bot AI while stuff like replays still hasn't been released yet. I agree. They did say that the majority of players who play often play against bots every week though. This is where stats might be extremely deceptive? I go into bots every now and then to try new jungle routes and new champions. I'm sure many do too? But I guess, like any company, they are in a position where they have to allocate resources to pleasing current customers or trying to get more customers. I would, however, like to see more features implemented like replays (where you can rewind) and better spectating (more than 2 and spectate while matchmaking). I don't want Magma Chamber, moar champions, nerfs/buffs -_- We might never see replays because of difficulty implementing with the game engine though. And spectating options are probably going to be a server nightmare. AFAIK, Riot only has one US server location on the west coast, which is why us east coast brethren have like ~80 ping minimum. Where are you at on the West Coast? I normally get 15 ping and my friend gets about 30(we live 40mins apart, he's in LA). Psh, I get average 9ping, and if I'm lucky I hit 8 ping. University internet ftw! I'm going to take a guess and say UCI? UCR.
|
On January 27 2012 01:39 Craton wrote:Show nested quote +On January 27 2012 01:37 Kaniol wrote:On January 26 2012 23:45 Craton wrote:
Except you have a linear rating scale when the grouping of people is parabolic. It can be displayed more effectively than with with the existing system. Same rule applies to IQ. The IQ system has a wide host of problems, too, you know.
For some reason this comparison made me laugh.
("I'm actually very smart, I'm just stuck in IQ hell")
|
On January 27 2012 03:28 Lylat wrote:
Two of the best EU teams playing against each other (CLG.eu vs M5) on Wickd stream : 800 viewers
Reginald playing normal games ; 7k viewers.
Wtf ?
I think people rather pick streams of players they know and streams that allready have alot of viewers.
on the elo thing: I find 100 elo actually alot and I definitely see less mistakes and generally better play at 1.5k than on 1.4k. The system sc2 uses is much better because you have to significantly improve your current skill to climb/fall to a different bracket. For lol it would be even better since there are factors you cannot influence. For example if you get carried a few games you can easily get up to an elo where you shouldnt play right now, same is true if you are getting trolled, or if a guy who got carried alot of games straight up feeds unintentionally. Those help and hinder you by the same amount statistically but fast climbing/descending makes it worse imo. The skill difference and understanding is thinner at higher rankings but between a 1.2k player and a 1.3k player there is such a huge gap its not even funny. We're talking about not getting smite as a jungler, building champions straight up incorrect but actually believing otherwise (not suboptimal, straight up false). Just very basic stuff. Players there actually don't know the stregths and weaknesses of the most played champions. They will tell you things like 'dont focus the tank' etc. They dont even have basic map awareness like at all.
I honestly believe that you shouldn't be able to gain 100+ elo when you don't consistently prove that you understand and play the game better than those 90% of players in your range and you shouldn't lose 100- elo if you just had a few bad luck games in a row.
Two short examples of how huge the skill difference from player to player at my elo range is (playing between 1.4-1.5). I've recently played two games as gp. I lost one and I honestly believe that both games had little to do with me but rather with my opponent and teammate respectively. The first I jungled, started red and went top to get FB. I saw the enemy lee sin bottom so i thought I should steal blue. The opponent cassio was missing around the time I got blue, which most likely ment that she smelled me there so I told ryze to walk towards me to kill cassio. He didnt, but as I expected she tried to kill me. So I tried to deal as much damage to her and juked her a few times until I died. She survived with 10% hp and ryze finally showed up after a long juking fight and about 20-30s after told him to come. So he got the kill and double buff, which he fed to her after a short time because he didnt go b to buy. The guy also rushed tear and got repeatedly 1shot by her.
The other example I played gp top and saw Lee sin starting with redbuff so I stole his blue and zoned cho for about 3 levels from exp and CS. I knew that lee would gank me so I kept myself near the brushes on the left until I built up a huge creepwave and then started attacking chogath straight up. When I saw lee comming I was prepared and dodged his Q running to the brushes. It was obv at that moment that lee had to fight me solo because of the huge creepwave and cho not being able to tank it + my Q (i was ghosted and he didnt have silence yet). But lee actually followed me feeding redbuff to me and FB. You can imagine how it went from there.
So I had a terrible and an amazing game, due to random people doing mistakes that you can't ever make at that rating. Those are players that went from a lucky winning spree into a bracket they shouldn't play at yet. If the system slowed that down then I believe we would have far less of that randomness.
|
I remember being rather annoyed with the way he had that table % set up. Anyway, it's not a curve because it's only one end of the ladder.
On January 27 2012 02:15 Gorsameth wrote:
There cant really be a comparison with IQ because if you are dumb i dont get dumber. While being a teamgame my ELo drops because you go derpmode and feed some akali.
Thats why the gap in skill of 1200 players is so high, there is a lot of players who would have a 50% winrate vs 15-1600 players in 1v1 but arnt good enough to carry there bad teammates and so lose a lot of games.
Not relevant. Given enough games you'll get to where you should be despite trolls or w/e. The myth of Elo hell is pretty much sustained by the belief that the only reason you can't get to whatever Elo is because your teammates are bad, but that's not the case.
On January 27 2012 02:31 NotSorry wrote:
if lol was like poker and I could just fold a shitty team preflop that might have some relevance
That's pretty much queue dodging.
On January 27 2012 02:33 Two_DoWn wrote:
The really surprising thing to me is how compressed the lower bits of the ladder are but how spread out the higher part is. Like I just hit 1708 and am only the 4,500th best lol player in NA (/humblebrag). But when 1500 I would jump that much by winning a game.
Yep.
On January 27 2012 02:35 NSANE.hydra wrote:
Oh, you're just looking for more distinction within the ~1100-1400 range? I think there are a multitude of problems causing that that entire 300 point range to be essentially the same skill level, while the other ranges have more clear-cut "skillcaps" to them.
But they're not the same level. A 1400 player is much better than a 1100 player. The issue tends to be trying to compare people who aren't actually that level and are only at whatever rating because they have too few games played. If someone has streaked with +40-50 wins (which is basically winning 4 games in a row early on) and are therefore 1400 rating, but their skill-level is really that of a 1100 player, it's going to skew the appearance of Elo and make 1400 the same as 1100, when in reality it isn't. You just need people to play enough to get to their real level before drawing comparisons.
Here's another interesting note: 1250 was roughly the top 25% of all players near the end of S1. New players start at 1200. Why do players start so far up in the system in the first place? If players started lower it could probably remove much of the instability of the "middle-ranges," since you wouldn't constantly be injecting players of an unknown ability into that range. I would think at most new players should start around whatever the 50% mark is.
|
I'm interested by the M5 teamcomp. Seems like if they get their ideal situation (Sona stuns with ult, vlad, brand and MF land theirs) they could straight out win any teamfight. I suppose scrims are where you try out this sort of thing.
|
Scrims usually get a high amount of views and Wicked is a pretty popular streamer 1k+ easy (he's at 1k+ now). I think this is just an expection or more people are interested in Regi because he hasn't streamed lately. Also if you look at the front page of own3d there are a lot of people watching from the M5's perspective.
M5 winning so far 8-0 or something just tuned in. Wicked counter jungling after killing Darian top lane as Trynd vs Vlad.
|
People just like his "raging" and over-the-top antics (BROBROBRO). Wickd is just more annoying than anything (omg did you see that? I'm so good).
|
On January 27 2012 03:30 wei2coolman wrote:Show nested quote +On January 27 2012 03:25 wussleeQ wrote:On January 27 2012 03:17 wei2coolman wrote:On January 27 2012 03:12 Mondeezy wrote:On January 25 2012 06:32 jacosajh wrote:On January 25 2012 06:27 Shiv. wrote:
What strikes me as odd is that they are investing resources into their bot AI while stuff like replays still hasn't been released yet. I agree. They did say that the majority of players who play often play against bots every week though. This is where stats might be extremely deceptive? I go into bots every now and then to try new jungle routes and new champions. I'm sure many do too? But I guess, like any company, they are in a position where they have to allocate resources to pleasing current customers or trying to get more customers. I would, however, like to see more features implemented like replays (where you can rewind) and better spectating (more than 2 and spectate while matchmaking). I don't want Magma Chamber, moar champions, nerfs/buffs -_- We might never see replays because of difficulty implementing with the game engine though. And spectating options are probably going to be a server nightmare. AFAIK, Riot only has one US server location on the west coast, which is why us east coast brethren have like ~80 ping minimum. Where are you at on the West Coast? I normally get 15 ping and my friend gets about 30(we live 40mins apart, he's in LA). Psh, I get average 9ping, and if I'm lucky I hit 8 ping. University internet ftw! I'm going to take a guess and say UCI? UCR.
DAMN lol. UCR was my original answer but i edited it >_>. Man, on the topic of low pings, it's so hard to change from 20ish to like 100. I know people who play at higher pings get used to it but just slight changes like these piss me off so badly... I don't know how loco does it with like 180-200 all the time while still being one of the best lol.
|
On January 27 2012 03:32 BlackMagister wrote:
Scrims usually get a high amount of views and Wicked is a pretty popular streamer 1k+ easy (he's at 1k+ now). I think this is just an expection or more people are interested in Regi because he hasn't streamed lately. Also if you look at the front page of own3d there are a lot of people watching from the M5's perspective.
M5 winning so far 8-0 or something just tuned in. Wicked counter jungling after killing Darian top lane as Trynd vs Vlad.
Anyone catch what happened? It looks like Wickd destroyed top and then the rest of the team sort of fell apart
|
Alistar countergank Yi up top at lvl 1 or 2, double killed Vlad/Yi, crazy snowball from there. 16min forfeit.
|
Saint is watching the Scrim giving his commentary, for those interested.
Edit: Games over nvm, 17/0 in favor of CLG EU
Edit 2: oh wait there doing another game
|
queue dodging penalty is too high to work like folding, if you dodged all the shit teams you would be at zero elo even if you won ever game you played out
|
Wicked saying he thinks they weren't really serious picks by M5 so the game didn't really mean much.
|
|
|
|
|
|
EPT![[image loading]](http://i.imgur.com/OHlSd.png)
