Altitude Game: Forums  

Go Back   Altitude Game: Forums > Altitude Discussion > Ladder Discussion
FAQ Community Calendar

Ladder Discussion Everything related to altitudeladder.com and the ladder servers goes here.

Reply
 
Thread Tools Display Modes
  #1  
Old 01-05-2011, 05:06 PM
CCN CCN is offline
Senior Member
 
Join Date: Jun 2009
Location: Xiang Gang
Posts: 1,992
Default Atrocious People never reach their true rating.

If you look at the bottom of TBD ladder you see Booksmart, Danielle, and Jayfourke. They are at the bottom and still dropping.

And yet, you would take those over quite a number of people rated much much higher.

They rate at between 400-600 points, and are still dropping.
Yet, Mad Nire is at 1000 and a 20% win rate. He will never play enough to get rated properly - for any realistic timeline, a ladder match with him will not be balanced.

When you are rated properly you should win/lose 50% of your games, so your % should be similar to 50%, less similar the further away from 1500 rating. But, with percentages below 40% is is fair to assume it is a misrating. Taking a figure of 42% - That encompasses 38 people on the last ladder page of 50. About 32 on the next, about 34 on the next.


So, ladder rating in practicality will not create balanced matched when people new to the scene join, as they are without exception always much below average skill, and the ladder does not let them drop far enough fast enough. If they become well skilled, and play many many many matches, maybe then.

Ladder rules, allow everyone to play because ranking will even out teams.

TL: DR Balanced matches are not a reality. Don't expect people who want good ladder games to be nice to new people.

Last edited by CCN; 01-05-2011 at 05:43 PM.
Reply With Quote
  #2  
Old 01-05-2011, 05:37 PM
Urpee Urpee is offline
Senior Member
 
Join Date: Dec 2010
Posts: 170
Default

I think there is much to be said about ladder ranking. I routinely play games with an average ranking of 1800, meaning that the average rank in the ladder is 100 or better (i.e. the top ranked 10% is playing)! This is an awfully narrow spectrum for quite a lot of games and yes if you factor in my own gashtly rating, you essentially get the top 75 or better.

That means that if you are not 1800 material (more like 1600 perhaps) you still have a chance of dropping consistently, because there is just nothing going on in that price-range. I.e. you are right that some people who should look not so bad look quite bad, and it's just the distribution of active versus inactive players.

In ladder just a few games can swing you from 1300 to 1700 because there is a huge inactive bulk in the middle.

I also have seen some funny things like that one team has 3 people below 1500 while the other has one and that is because one very high ranked person is there.

But in reality the density in the top 50 is very high and the skill gradient is not huge, so you basically get games that are quite imbalanced because of the assignments of the low-ranked folks.

And the problem is exasperated by the fact (as you noted) that especially the mid-range scores are not that representative. You can get a rank 1500 and a rank 1300 but in reality the 1500 just hasn't played that much and the 1300 is actually really close to 50% but due to the numbers played as accumulated a sizeable number of raw losses, creating the score.

This leads me to a final phenomenon, that I'd call digging a hole. If you start off poor but learn, you will end up in a loss whole. Even if you then have a 50% ratio and if you started fresh would be at 1500, you are actually stuck at whatever score you dug yourself into. Learning to be 50% in games with an average score of 1800 will keep you looked in at the score you have (you gain no wins, hence you gain no score ground).
Reply With Quote
  #3  
Old 01-05-2011, 05:41 PM
CCN CCN is offline
Senior Member
 
Join Date: Jun 2009
Location: Xiang Gang
Posts: 1,992
Default

My whole post is about rating,t he points next to your name, not ranking, the rank those points give you. Ladder balances on points, not rating.

to move from 1300 - 1700 rating is 500/25 = 20 losses.
To move from 1300 - 1700 rank is much, much less, and is completely irrelevant to my point.

If you are bouncing within 100-200 points then that is reasonably stable and thus well rated, regardless of what that does to your rank.


Edit: Thinking it through adding a decay factor is just going to exacerbate this problem, unless a ELO is held. As good people decay down they imbalance every time they come back.

Last edited by CCN; 01-05-2011 at 05:47 PM.
Reply With Quote
  #4  
Old 01-05-2011, 05:47 PM
Urpee Urpee is offline
Senior Member
 
Join Date: Dec 2010
Posts: 170
Default

Quote:
I think there is much to be said about ladder rating.
Fixed it. I responded to what you are talking about. I mostly talk about rating too and how the mechanism influences how games are balanced. That is the topic, right?
Reply With Quote
  #5  
Old 01-05-2011, 05:54 PM
Urpee Urpee is offline
Senior Member
 
Join Date: Dec 2010
Posts: 170
Default

Basically I try to point out that the argument that people land at a rating that reflects their 50% win/loss ratio is not correct.

I don't think that if your score is stable you are well-rated at all.

In fact observe ladder and compare win ratio and rating. People with a lower percentage but more games will have a better ranking than people with a higher percentage but fewer games.

You can compete at a 50% level with players whose average rating is 1800 score and keep your current score no matter what it is. It may well be way below 1800.

I wonder if selecting people for balance based on their win ratio factored by number of games played would be more reliable.
Reply With Quote
  #6  
Old 01-05-2011, 07:14 PM
shrode shrode is offline
Senior Member
 
Join Date: Jul 2010
Location: Fargo, ND
Posts: 823
Default

when new people come into ladder, they should start out moving by 100 or so for the first couple games, and working their way down to the standard 25 points after 10 games or so, thus, allowing them to get to their proper rating much faster. This type of system is used in the very popular game league of legends, and helps to combat the smurfing issue. If I create a new account in LoL, the first game I just RAPE everybody, but after I win 6 or so games in a row, I start playing with people my skill level already.

I know this would get rid of the true zero-sum thing, but who really gives a ****?
Reply With Quote
  #7  
Old 01-05-2011, 07:16 PM
tgleaf tgleaf is offline
Super Moderator
 
Join Date: Oct 2009
Location: being a video game clan CEO is not that prestigious
Posts: 2,737
Default

Quote:
Originally Posted by Urpee View Post
In fact observe ladder and compare win ratio and rating. People with a lower percentage but more games will have a better ranking than people with a higher percentage but fewer games.

I wonder if selecting people for balance based on their win ratio factored by number of games played would be more reliable.
Yes, this is why ladder ratings stop working after a while, and why people have continually suggested that there be a minimum number of games played before players are officially rated.

Win percentage means very little as long as you are just above 50% and have played a TON of games. So as an example -- assuming 25 pts per match -- if you went 110-100 (52.4% wins), you could improve your rating by (110 x 25 - 100 x 25) = 250 pts; if you went 550-500 (52.4% wins), you could improve your rating by (550 x 25 - 500 x 25) = 1250 pts. But in both cases, you'd have won the same percentage of games. I know that example is really simplified, but I think it still holds true.

Gogogo Ladder Season 2 (reset)!!!
Reply With Quote
  #8  
Old 01-05-2011, 07:34 PM
Ajuk999 Ajuk999 is offline
Senior Member
 
Join Date: Dec 2009
Location: United States
Posts: 3,823
Send a message via AIM to Ajuk999
Default

Look at me, I've played almost 200 games in ball ladder and I haven't even gotten past the 1K mark yet.

http://www.altitudeladder.net/profil...2-0ccfaefb753e
Reply With Quote
  #9  
Old 01-05-2011, 07:48 PM
tgleaf tgleaf is offline
Super Moderator
 
Join Date: Oct 2009
Location: being a video game clan CEO is not that prestigious
Posts: 2,737
Default

Quote:
Originally Posted by Ajuk999 View Post
Look at me, I've played almost 200 games in ball ladder and I haven't even gotten past the 1K mark yet.

http://www.altitudeladder.net/profil...2-0ccfaefb753e
Yes, and what is your point? Actually, a better example would be my rating (http://www.altitudeladder.net/profil...a-15150a9f431f) since I have about the same win percentage as some of the top-ranked players. I'm not for a second suggesting that I should be ranked higher... if anything, I should be kept out of the balancing at this point since I haven't played enough games and tend to skew incorrectly.

See also Maimer at #64 lol.
Reply With Quote
  #10  
Old 01-05-2011, 07:51 PM
lemon lemon is offline
Senior Member
 
Join Date: Jul 2010
Posts: 123
Default

I'm confused how this ladder ranking works in general. People say its ELO but you gain same amount every time (+-25), thats not ELO. Maybe you use team ratings against each other and because they average to the same you always get +-25 points. But that doesnt make any sense! If I am 1300 player I am playing against average 1800 oposition and I gain +-25 points, same as 2500 rated player in my team, which is a nonsence. You need to apply ELO formula of one's individual ranking against the opponents average rating. So after the game you'd have something like this (somewhat random numbers):

Winners (average 1800):
2500guy +2 (apply ELO formula for 2500 winning against 1800)
2000guy +4
2000guy +4
1500guy +10
1000guy +20

Loosers (average 1800):
2500guy -15
2000guy -10 (apply ELO formula for 2000 loosing against 1800)
2000guy -10
1500guy -5
1000guy -2
Reply With Quote
  #11  
Old 01-05-2011, 08:26 PM
elxir elxir is offline
Senior Member
 
Join Date: Feb 2010
Location: All-American
Posts: 2,687
Default

i would kill myself if i gained 2 points per ladder match

there aren't enough players to make small gains like that meaningful
Reply With Quote
  #12  
Old 01-05-2011, 08:34 PM
Stormich Stormich is offline
Senior Member
 
Join Date: Apr 2009
Location: Corporate Police State
Posts: 1,151
Default

It's modified ELO. Also from reading Microsofts versions of modified ELO (used in Halo iirc), when we were discussing how the point accumulation should look like, they said you need at least 300 games before your rating starts to resemble the truth.
Reply With Quote
  #13  
Old 01-05-2011, 08:55 PM
nobodyhome nobodyhome is offline
Senior Member
 
Join Date: Apr 2009
Posts: 1,088
Default

I've acknowledged several times in the past (and have plans to fix) the fact that the ladder rating system is bad.

The problem that CCN describes is the biggest problem with it. The best way to fix this is to introduce inflation into the system so that newer players start near the low end of the range of possible ratings as opposed to smack dab in the middle.

Shrode's solution helps alleviate the problem in that it puts people at their correct rating faster, but is best combined with inflation so that newer people not only get to their rating faster, but start nearer to their correct rating.

Urpee is correct in that the distribution of ladder players (most of the active players are ~1700-2100 in range) results in some funny stuff involving people who are lower than that range. Notably I have seen that there are three ranges of players: the top tier (1900 and above) whose ratings simply go steadily up, the mid range (~1500) whose ratings fluctuate wildly up and down, and the low range (1100 and below) whose ratings just keep going downwards. There are ways of correcting this that are being considered.

Tgleaf is also correct in that win percentage matters little to none (and alongside this, neither does your actual number of wins and losses). The only thing that factors in is your rating. A person who has gone 50-10 with a resulting rating of 2500 is the same for all intents and purposes as a person who has gone 1000-960 with a rating of 2500. The ladder system may as well never save the number of wins and losses someone has and can still work exactly the same. The reason why this is not such a problem is because the difficulty of each of your matches changes depending on how high your rating is--the higher you are rated, the more difficult your opponents are (or the crappier the players you are put with are).
Reply With Quote
  #14  
Old 01-05-2011, 09:02 PM
nobodyhome nobodyhome is offline
Senior Member
 
Join Date: Apr 2009
Posts: 1,088
Default

Quote:
Originally Posted by lemon View Post
I'm confused how this ladder ranking works in general. People say its ELO but you gain same amount every time (+-25), thats not ELO. Maybe you use team ratings against each other and because they average to the same you always get +-25 points. But that doesnt make any sense! If I am 1300 player I am playing against average 1800 oposition and I gain +-25 points, same as 2500 rated player in my team, which is a nonsence. You need to apply ELO formula of one's individual ranking against the opponents average rating. So after the game you'd have something like this (somewhat random numbers):

Winners (average 1800):
2500guy +2 (apply ELO formula for 2500 winning against 1800)
2000guy +4
2000guy +4
1500guy +10
1000guy +20

Loosers (average 1800):
2500guy -15
2000guy -10 (apply ELO formula for 2000 loosing against 1800)
2000guy -10
1500guy -5
1000guy -2
ELO is only meant for 1v1s (afaik it was developed as a rating system for chess), but when modified for teams it should not work the way you described it. The system is indeed using your team's average and not your personal rating against the entire other team's. If your team's ratings were not taken into account in scoring you, then your expected value (gain or loss) of each game you play would be dependent on only two things: your rating, and the average opponent's rating. Notice how your team's rating would be entirely discluded from your expected value, which is completely unreflective of the actual game (in which your teammates' skill levels is a huge factor in whether you win or lose).
Reply With Quote
  #15  
Old 01-05-2011, 09:07 PM
ryebone ryebone is offline
Senior Member
 
Join Date: Jan 2009
Posts: 470
Default

Quote:
Originally Posted by lemon View Post
I'm confused how this ladder ranking works in general. People say its ELO but you gain same amount every time (+-25), thats not ELO. Maybe you use team ratings against each other and because they average to the same you always get +-25 points. But that doesnt make any sense! If I am 1300 player I am playing against average 1800 oposition and I gain +-25 points, same as 2500 rated player in my team, which is a nonsence. You need to apply ELO formula of one's individual ranking against the opponents average rating. So after the game you'd have something like this (somewhat random numbers):

Winners (average 1800):
2500guy +2 (apply ELO formula for 2500 winning against 1800)
2000guy +4
2000guy +4
1500guy +10
1000guy +20

Loosers (average 1800):
2500guy -15
2000guy -10 (apply ELO formula for 2000 loosing against 1800)
2000guy -10
1500guy -5
1000guy -2



It's illogical to compare individual ratings to the average opponent ranking, because it goes AGAINST the expected amount of contribution. By your suggestion, you're saying the 2500 guy should win less and lose more than the 1000 guy, even though 99% of the time the 2500 guy contributed far more to the team while the 1000 guy just went along for the ride?
Reply With Quote
  #16  
Old 01-05-2011, 09:10 PM
lemon lemon is offline
Senior Member
 
Join Date: Jul 2010
Posts: 123
Default

Quote:
Originally Posted by elxir View Post
i would kill myself if i gained 2 points per ladder match

there aren't enough players to make small gains like that meaningful
It's just an example, you can multiply these numbers by some value if you want to. It does make a lot of sense tho, if you are at the top of the ladder with 2500 or however many points it should become harder and harder to progress. Thats not the case now as you still gain/lose 25 points each game. Big problem with this is like someone mentioned - if you lose at the start and then keep winning at 50% you'll never move up, despite playing against 1800 average rated teams. Just for the record, 1500 guy is expected to win maybe 25% against 1800 opposition using ELO model.

edit: two posts while I was making this one. Fair points, but I think my last point is still standsl. If I can consistently win 50% games against 1800 rated opposition shouldn't I be 1800 myself?

Last edited by lemon; 01-05-2011 at 09:12 PM.
Reply With Quote
  #17  
Old 01-05-2011, 09:27 PM
XX1 XX1 is offline
Senior Member
 
Join Date: Mar 2010
Location: Seattle
Posts: 727
Default

Quote:
Originally Posted by Ajuk999 View Post
Look at me, I've played almost 200 games in ball ladder and I haven't even gotten past the 1K mark yet.

http://www.altitudeladder.net/profil...2-0ccfaefb753e
lululululul you
I still think that it is easier to climb tbd ladder though
Reply With Quote
  #18  
Old 01-05-2011, 09:58 PM
ryebone ryebone is offline
Senior Member
 
Join Date: Jan 2009
Posts: 470
Default

Quote:
Originally Posted by lemon View Post
It's just an example, you can multiply these numbers by some value if you want to. It does make a lot of sense tho, if you are at the top of the ladder with 2500 or however many points it should become harder and harder to progress. Thats not the case now as you still gain/lose 25 points each game. Big problem with this is like someone mentioned - if you lose at the start and then keep winning at 50% you'll never move up, despite playing against 1800 average rated teams. Just for the record, 1500 guy is expected to win maybe 25% against 1800 opposition using ELO model.

edit: two posts while I was making this one. Fair points, but I think my last point is still standsl. If I can consistently win 50% games against 1800 rated opposition shouldn't I be 1800 myself?

Lemon, I think you might've missed the point about team autobalance, so I'll just state it again so there's no miscommunication. Both teams will have an identical TEAM average rating, composed of five INDIVIDUAL ratings. To do this, there must be a mix of high-rated people, middle-rated people, and low-rated people. The higher one person is, the lower his teammates must be to balance out according to the other team. Yes, the team's expected winning% is 50%. But as long as there is someone in the game who isn't properly rated, it will not be 50%. With that said,

Quote:
if you are at the top of the ladder with 2500 or however many points it should become harder and harder to progress. Thats not the case now as you still gain/lose 25 points each game.
It IS more difficult to progress, because as a top player your teammates are getting worse and worse while your opposition stays at 1800 (on average). Therefore, you're expected to carry your team to a greater and greater extent. True, your team still has a 50% chance of winning, but because your teammates are poor, you have to carry a larger load. If you're not good enough to do that, then you'll start losing games, and your rating will go down to a point where you can fulfill a reduced amount of contribution, because your teammates' contributions will have gone up. Don't think of each game by itself; instead, think of the individual's overall body of work, and it should make more sense.

Quote:
if you lose at the start and then keep winning at 50% you'll never move up, despite playing against 1800 average rated teams
A 50% winning ratio is the ENDPOINT in our system. If you're winning at 50%, it means that you're at a specific rating that will give you teammates of an appropriate rating in order to win 50% of your games. In order to move up, you have to win more than 50%. The quality of the opposing team is irrelevant, because your team has the same rating.

Quote:
Just for the record, 1500 guy is expected to win maybe 25% against 1800 opposition using ELO model.
In a 1v1 situation, sure. But Altitude isn't 1v1.

Quote:
If I can consistently win 50% games against 1800 rated opposition shouldn't I be 1800 myself?
Not necessarily, because if you're 1000, it's most likely your teammates doing most of the work, not you. If you're rated 1000 and the opponent average is 1800, then you'll get teammates who average 2000 (whereas if you were 1800, your teammates would average 1800).
Reply With Quote
  #19  
Old 01-05-2011, 10:08 PM
nobodyhome nobodyhome is offline
Senior Member
 
Join Date: Apr 2009
Posts: 1,088
Default

Quote:
Originally Posted by lemon View Post
It's just an example, you can multiply these numbers by some value if you want to. It does make a lot of sense tho, if you are at the top of the ladder with 2500 or however many points it should become harder and harder to progress. Thats not the case now as you still gain/lose 25 points each game. Big problem with this is like someone mentioned - if you lose at the start and then keep winning at 50% you'll never move up, despite playing against 1800 average rated teams. Just for the record, 1500 guy is expected to win maybe 25% against 1800 opposition using ELO model.

edit: two posts while I was making this one. Fair points, but I think my last point is still standsl. If I can consistently win 50% games against 1800 rated opposition shouldn't I be 1800 myself?

It does become harder and harder to progress if you are at the top of the ladder. Once again you are neglecting the factor of your team. Let's say the average rating of your entire team is 1800 and you happen to be a guy with 2500 rating. Then the rest of your team (not including you) has an average rating of 1625. You thus have a much more difficult time winning than supposing you were a guy rated 1500 on the same (average rating 1800) team, because that guy's teammates is rated 1875.

The model is hard to wrap your head around because of the balancing system, but with some examples it is easy to see that your last point does not stand either. Let's suppose again that most games are played with both teams averaging 1800. For sake of clarity let's take your example to the extreme and suppose that you are rated 0, but somehow you still win 50% of your games. Should you not be moving up? No, you shouldn't be, because there are four guys on your team rated 2250 who are basically carrying your ass.

Yes it is true that a 1500 guy is expected to win 25% against 1800 guys in the ELO model. However, once again, the 1500 guy is not playing in a vacuum 1v5 against the other team, he is playing with four teammates who together make up a 1800 entity (same as the other team, a 1800 entity).
Reply With Quote
  #20  
Old 01-05-2011, 10:11 PM
nobodyhome nobodyhome is offline
Senior Member
 
Join Date: Apr 2009
Posts: 1,088
Default

We are sort of derailing this topic (from some of the points and suggestions that CCN, shrode have put up) and are sort of focusing solely on lemon's point. Lemon, if you wish to continue this discussion please open up a new thread about it. Just FYI, the system you propose was the rating system first used when ladder was launched a year ago. We had it for about a week and then we realized it wasn't working so we changed it.
Reply With Quote
  #21  
Old 01-06-2011, 03:41 AM
CCN CCN is offline
Senior Member
 
Join Date: Jun 2009
Location: Xiang Gang
Posts: 1,992
Default

I was a bit frustrated, and wanted to bring up this point in hopes of being told by Nobo that Eso was taking it into account.

Last edited by CCN; 01-06-2011 at 05:43 AM.
Reply With Quote
  #22  
Old 01-06-2011, 08:07 PM
Jayfourke Jayfourke is offline
Senior Member
 
Join Date: Jun 2009
Location: England
Posts: 490
Default

I object to being labelled as atrocious. There are are fair few people statistically worse than me, and my w/l ratio is barely under 50%.

Mostly I think my playing style isn't quite what ladder calls there. There's an element of strategy that I totally ignore. Going in guns blazing is much more fun.

Tl;Dr, CCN, you're a jerk. </3
Reply With Quote
  #23  
Old 02-11-2011, 10:53 PM
zz- zz- is offline
Member
 
Join Date: Jan 2010
Posts: 90
Default

Its easier to win the lower your rating is, and harder to win the higher your rating is. This is why resetting ladder, although being an idea im not really opposed to if people want to do it, wouldn't make anyone any happier with their rating -- if you cant move up now, your rating wouldnt be any higher after a reset. In fact, a ladder reset would make EVERY player inaccurately rated. Which means there'd be a lot more chaos and variables in each game, making it take more games for an underrated player to move up. The more accurate everyone's rating is, the easier and quicker it is for an underrated player to move up. If you aren't moving up in ladder it's because you aren't playing better than your rating says.


Edit: CCNs point about horrible players not reaching their true rating is valid. Is it possible for a person's first 20 games, instead of going +/-25, JUST they go +/-43 and everyone on the opposing team go +/-28 and everyone on the new guys team stay at +/-25? Sorry if thats confusing but it would preserve the zero-sum nature of ladder and help newer players reach their appropriate rating faster. The change in the opposing team (+/-28) isnt much different than the current 25, and doesn't stand to benefit or punish anyone unfairly or impact the nature of the ladder in any significant way other than new players being rated accurately more quickly. Perhaps even +/- 30 for the opposing team and +/-55 for the newb.

Last edited by zz-; 02-19-2011 at 09:52 PM.
Reply With Quote
  #24  
Old 02-19-2011, 09:45 PM
zz- zz- is offline
Member
 
Join Date: Jan 2010
Posts: 90
Default

Bump because I think this is something that should be addressed sooner rather than later.

The last two days of ladder have been an absolute luckfight over dodging the overrated new players. Yesterday Bayonator went on an 11 game losing streak as he fell from 1400 to 1100. His true value is probably like 600, so he's not even half done ruining games. When every day there's a new guy or two going on 8 game losing streaks it wipes out a large portion of the chance ladder has of being competitive for 2-3 hours. As I'm typing this a game just ended that brought the combined score of the last 4 games to 24-5 in favor of the winning team.


Is the solution in my edit in the post above a viable one? Even if just temporary. It's unfortunate that ladder is getting a nice influx of new players that are being greeted with moaning and groaning from everyone (including me) when they have to play with them. It's not good for the long term health of ladder and Altitude in general as there are many of us who really just play ladder.

Last edited by zz-; 02-19-2011 at 09:48 PM.
Reply With Quote
  #25  
Old 02-19-2011, 09:57 PM
elxir elxir is offline
Senior Member
 
Join Date: Feb 2010
Location: All-American
Posts: 2,687
Default

let's use VORP
Reply With Quote
  #26  
Old 02-19-2011, 10:03 PM
zz- zz- is offline
Member
 
Join Date: Jan 2010
Posts: 90
Default

Lol it's like VORP where the RP is accidentally not valued at 0.

Lix you see I jacked your stat about the winning teams running score hope you don't mind
Reply With Quote
Reply


Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off

Forum Jump


All times are GMT. The time now is 06:09 AM.


Powered by vBulletin® Version 3.8.2
Copyright ©2000 - 2021, Jelsoft Enterprises Ltd.
2008 Nimbly Games LLC