Another Look at 2v2 Super Titan

[worldfamous]

(03-22-2013 07:43 AM)amoffett Wrote:  

(03-22-2013 07:02 AM)worldfamous Wrote:  

(03-22-2013 04:09 AM)GreatGonzales Wrote:  

(03-22-2013 04:01 AM)amoffett Wrote:  Weren't you guys arguing though that you felt you were about as good as they were? If that were true, it would only have been a matter of time.

so basically a bunch of ST-quality teams were eating each other and no one could climb high enough to reach the bar.

Well said. That's exactly what was happening.

Except for a couple of teams did reach the bar.

You can't have it both ways, you can't a) be as good as them and b) not get up to their skill rating eventually. So which is it?

It seems you're just being antagonistic. I think you're just not understanding that 2v2 and 1v1 are different worlds. Adam and Alex took a second look, realized as much, and adjusted. If it makes you feel any better, it's definitely not as satisfying reaching ST the way we did.

[amoffett11]

(03-22-2013 10:00 AM)worldfamous Wrote:  

(03-22-2013 07:43 AM)amoffett Wrote:  

(03-22-2013 07:02 AM)worldfamous Wrote:  

(03-22-2013 04:09 AM)GreatGonzales Wrote:  

(03-22-2013 04:01 AM)amoffett Wrote:  Weren't you guys arguing though that you felt you were about as good as they were? If that were true, it would only have been a matter of time.

so basically a bunch of ST-quality teams were eating each other and no one could climb high enough to reach the bar.

Well said. That's exactly what was happening.

Except for a couple of teams did reach the bar.

You can't have it both ways, you can't a) be as good as them and b) not get up to their skill rating eventually. So which is it?

It seems you're just being antagonistic. I think you're just not understanding that 2v2 and 1v1 are different worlds. Adam and Alex took a second look, realized as much, and adjusted. If it makes you feel any better, it's definitely not as satisfying reaching ST the way we did.

I'm not trying to be antagonistic. However, I really don't understand where youre coming from when you say that "I don't understand that 1v1 and 2v2 are different worlds", when that's all I've been saying since the beginning, by saying that 1v1 numbers and statistics should not be applied to 2v2.

---

(03-22-2013 08:19 AM)GreatGonzales Wrote:  

(03-22-2013 07:43 AM)amoffett Wrote:  You can't have it both ways, you can't a) be as good as them and b) not get up to their skill rating eventually. So which is it?

Neither A or B actually... Not saying we (those that are newly ST) are as good as the two pre-existing STs. Saying we are almost as good, good enough to deserve to be in the highest league. Also not saying that we are at "their" skill rating; we are ST because we have surpassed a point threshold that has been lowered, not the old threshold which was more excluding. You seem to have a misunderstanding about how ranking systems work; it's not as though people can go up in score indefinitely. There's a limit to how far one can increase in skill rating/points that is in part dependent upon the number of players in the population. The old threshold was too high for the size of the player population, so OML adjusted it and now we're ST. A good way to think of it is this: you can only climb as high as there are people below you. Not a perfect analogy but you get the idea.

But new people are playing Outwitters all the time, as well as new teams of existing players adding to the 2v2 pool. The size of the population is growing all the time. Therefore, if your analogy is correct, you would've become a supertitan by waiting it out a bit(and by continuing to win a bunch though that goes without saying).

As far as how ranking systems work, the number of super titans is independent of the number of players in the population. It depends only on the skill distribution of the players. Because of the matchmaking system, there is a maximum number of supertitans, but there is no minimum. For example, if everyone played everyone 50/50, there would no supertitans. If one team won 90% of the time, and the others all won 50% of the time, you'd have one ST. This is the case whether there's 10 players or 1000s.

Here's an interesting thought: imagine that everyone was awesome at 2v2. Or that the only people who ever played were just the top 40 teams. You'd all probably be in gifted or clever league (except for the current STs, based on their W/L they'd still probably be STs). It's not about how many players there are, but how many are below you in skill (below you in skill so that you can beat them). The reason none of you became STs as quickly as the other teams is because you all played each other so evenly. Imagine a population of 50 players. One team beats everyone else: they are a ST, because they are better than 49 teams. One team loses every game: they are fluffy, because they are better than 0 teams. The other 48 teams play each other evenly: they are all in clever (maybe not even there) because there is only one team worse than all of them. If you are even with a team, they are not worse than you. This is what happened more or less with the old threshold, so many of you were even towards the top, that there weren't enough players below you for you to advance. Now the threshold is lower, so less 'lower' players are required. This is why I doubt that OML will find a threshold to give you the percentage you're after, it's like splitting an atom, you guys are too close together.

[laYahooz]

I think there are too many STs for 2v2. I got 2 gifted teams and one master team and I definately think masters aren't as good as they should be. Not saying I want teams like GG and Rawkhawk in masters, but there are some ST teams that I think I can beat (and we suck compared to the upper teams like rawkhawk and GG for example).

Don't really want to get to in depth on this as like CombatEx said in a post, t shouldn't be too complicated on the distribution of STs. Top 1 percent should be the only STs (I reckon that's like top 20 teams?). However, with top 2v2 lists I don't think the title ST is THIS big of a deal, though I do understand why the truly talented teams would like to shorten the gap a little.

[TheQwertiest]

i dont think theres anything wrong with the number of STs in 2v2.
i agree that theres a noticeable skill gap between the top and the bottom of the STs but isnt that the same in 1v1?
the distinct level that separates master and st just isnt as noticeable in 2v2 than in 1v1 because of the nature of 2v2.
theres more factors to be considered. 2v2 games tend to be slower so people have less experience in games so results will be more variable until both pairs have more roots in how 2v2 should be played

lol wtf am i saying. sorry i really cant explain what i want to say in english properly.

[Necrocat219]

Seems fine to me to be honest. The top 200 list is a good indicator of what your exact placement is and if you can stay consistently in Super Titan it shows that you are consistent in your placement and deserve to be there. It's not like a Super Titan is the equivalent of a grand master in chess, the bottom of the ST league in 1vs1 are very considerably weaker compared to the top, and if the quality of the top 2vs2 players gradually changes over time we'll notice a shift in how members are placed.

[GreatGonzales]

(03-22-2013 12:24 PM)amoffett Wrote:  But new people are playing Outwitters all the time, as well as new teams of existing players adding to the 2v2 pool. The size of the population is growing all the time. Therefore, if your analogy is correct, you would've become a supertitan by waiting it out a bit(and by continuing to win a bunch though that goes without saying).

True, but players are also leaving the player pool all the time. Also, we would need to see many orders of magnitude greater number of players to support 10-15 STs with the old threshold. Think of it this way, suppose there are 1500 players in a player pool (approximately the number of 2v2 teams by the way), and the threshold is set at .2%, which is 3 players. Now, suppose we increase the player population by 10%, bringing the total number of players to 1650. Keeping the same .2% threshold, that would yield 3.3 players. In order to reach the lower end of your estimate, 10 players, we would need about 5000 players, which is a 333% increase. There is simply no reason to believe that this was going to happen under the old threshold.

Quote:As far as how ranking systems work, the number of super titans is independent of the number of players in the population. It depends only on the skill distribution of the players. Because of the matchmaking system, there is a maximum number of supertitans, but there is no minimum. For example, if everyone played everyone 50/50, there would no supertitans. If one team won 90% of the time, and the others all won 50% of the time, you'd have one ST. This is the case whether there's 10 players or 1000s.

I maintain that you are mistaken on this point. The number of players awarded ST status is dependent on the skill distribution, yes, but also on the number of players in the population. Think of it in the reverse way; suppose that there were 100,000 1v1 players instead of about 20,000. Would we see many more STs? Of course. You're right that if there was one player who won always won 50% of all matches, regardless of player skill, then they would not move up or down in league. But that's not a very useful hypothetical; in practice, player populations have diverse skill levels, and skill levels increase over time as players learn the game, so skill diversity within a player population emerges.

Quote:Here's an interesting thought: imagine that everyone was awesome at 2v2. Or that the only people who ever played were just the top 40 teams. You'd all probably be in gifted or clever league (except for the current STs, based on their W/L they'd still probably be STs). It's not about how many players there are, but how many are below you in skill (below you in skill so that you can beat them). The reason none of you became STs as quickly as the other teams is because you all played each other so evenly. ... This is what happened more or less with the old threshold, so many of you were even towards the top, that there weren't enough players below you for you to advance.

Again, not a very useful hypothetical, but you actually hit right on target here; you're right, players who are close together in skill would average about 50% win ratio, which means they would not increase in skill rating anymore. So, what happened was that there were a bunch of teams who were very, very good (myself, worldfamous, awpertunity/terence, burnodrod/mastercaster, sir3 and p1noyboy, etc.), and so we rose to the top of the ranking list. But because the threshold was so high, and we were roughly even in skill, we all played each other (due to matchmaking) and no one could get out ahead enough of the pack in order to reach the very limiting threshold. This was a bad system, as it did not award the very very good teams who were merely not quite as good as the top 2. If there were (many) more players in the population, this wouldn't have been such a big issue, but with only about 1500 something had to change. It's too bad that I was promoted this way, didn't feel very triumphant. I wish OML had planned the league system more appropriately, but I guess you can't fault them for not being able to see the future.

Quote:This is why I doubt that OML will find a threshold to give you the percentage you're after, it's like splitting an atom, you guys are too close together.

I agree that the top 10-15 teams are very close in skill, but beyond that it is more disparate. As for changing the threshold again, as I have said I'm not sold on the idea that it should be changed again; I created this thread to see what the community thinks. But I still don't see why OML couldn't merely pick a spot between the old threshold and new threshold and call it a day.

[GreatGonzales]

Amoffett, another example that might help. I used to be an active member of the Hero Academy community, which had a fan-based league. It was pretty cool, we had some tech-saavy members create an automated match reporting system that was based on the ELO ranking system (also used for chess). I do not think OML is using ELO, as there is some aspect of winning and losing streaks that plays into it, which I assume is a means of ensuring that a player who is borderline between two leagues isn't being promoted and demoted every other game. But probably the Outwitters system shares the same basic principles of ELO.

Anyway, in ELO everyone starts with a score of 1200. If you beat someone who has a higher score, you get more points than if the person you beat had a lower score. Likewise, if you lose to someone with a lower score than you, you would lose more points than if the person had a higher score. Pretty simple. The population was probably about the same size as the 2v2 Outwitters population. There are no leagues, just a ranking based on ELO score. But if we did assign leagues based on ELO score, we would need to keep in mind that there is such a thing as too high of a threshold. In the history of the league, I think only 1-2 people ever crossed the 1600 threshold. So, if we set the bar at 1600, or 1700, that would be too high. In chess, with a much larger population, it's possible to reach a score as high as 2500. Look at this online chess population: http://www.chess-mind.com/en/elo-system This page shows an overview of various chess levels using the ELO system. But if you look at their statistics page, you'll see that their highest rated player is only at the 1900 level. This is because the population of this website is only about 2600 registered users (though, on a second look, only about 200 active users. ELO range likely the result of a high K value, which is part of the formula. And/or, players register, "feed the population", and then leave.). So, if this site were to split the population into 5 leagues, like Outwitters, it would be inappropriate for them to set a threshold of 2000 ELO for the highest league. Suppose also, the greatest chess player in the world, having a ELO in the global community of 2900 let's say, entered into this population. Could this player reach 2900 in this smaller population? Yes, in theory, but probably not for years and years, and he better not lose at all or he'll fall in rank significantly! This is because he would be gaining so little in points with each win as he increases his rank. This is what I mean when I say there are limits to how high a player can go in rank within a given population. In theory, it's true, you can go on indefinitely. But in practice, people lose from time to time, so really there is a limit. Hope that helps.

[amoffett11]

(03-23-2013 12:57 AM)GreatGonzales Wrote:  I maintain that you are mistaken on this point. The number of players awarded ST status is dependent on the skill distribution, yes, but also on the number of players in the population. Think of it in the reverse way; suppose that there were 100,000 1v1 players instead of about 20,000. Would we see many more STs? Of course. You're right that if there was one player who won always won 50% of all matches, regardless of player skill, then they would not move up or down in league. But that's not a very useful hypothetical; in practice, player populations have diverse skill levels, and skill levels increase over time as players learn the game, so skill diversity within a player population emerges.


Again, not a very useful hypothetical, but you actually hit right on target here; you're right, players who are close together in skill would average about 50% win ratio, which means they would not increase in skill rating anymore. So, what happened was that there were a bunch of teams who were very, very good (myself, worldfamous, awpertunity/terence, burnodrod/mastercaster, sir3 and p1noyboy, etc.), and so we rose to the top of the ranking list. But because the threshold was so high, and we were roughly even in skill, we all played each other (due to matchmaking) and no one could get out ahead enough of the pack in order to reach the very limiting threshold. This was a bad system, as it did not award the very very good teams who were merely not quite as good as the top 2. If there were (many) more players in the population, this wouldn't have been such a big issue, but with only about 1500 something had to change. It's too bad that I was promoted this way, didn't feel very triumphant. I wish OML had planned the league system more appropriately, but I guess you can't fault them for not being able to see the future.

No, the number of ST's is dependent on the # of players only if you assume that their is an even skill distribution among the players (which in this case their mostly is, although I would say that in 2v2 its weighted more towards the top). I'm not sure why you talk about the threshold as if its based on a certain percentage (0.3%, etc.), the threshold isn't based on a percentage, but on a skill rating.

I think you've misunderstood my hypothetical about the teams who are all 50/50. You're saying that there's a bunch of teams who are very, very good, and I'm sure you're right. But what I was saying is that imagine that everyone was that good: is anyone actually very very good or is everyone just average. This is what I was trying to say: the reason you did not reach super titan before was because you were not *better* than enough teams. You were worse than 2 (the two STs), you were even with several other teams (the very, very good teams you listed, and perhaps others), and better than all the rest. To be a supertitan, you have to be *better* than a certain number of teams; the teams you're even with you're not better than; playing them to 50/50 won't help you reach that level.

" I still don't see why OML couldn't merely pick a spot between the old threshold and new threshold and call it a day"

If the old threshold was 10, and the new one is 9, for example, then there is no spot in between. Obviously they're further apart than that, but I think that they're closer than you think, and to pick an arbitrary spot in between to get an acceptable percentage *right now* is not a good idea, the skill distribution will change, and the percentage will continually be in flux. The trick is to be patient and let it sort itself out.

[GreatGonzales]

(03-23-2013 03:19 AM)amoffett Wrote:  No, the number of ST's is dependent on the # of players only if you assume that their is an even skill distribution among the players (which in this case their mostly is, although I would say that in 2v2 its weighted more towards the top). I'm not sure why you talk about the threshold as if its based on a certain percentage (0.3%, etc.), the threshold isn't based on a percentage, but on a skill rating.

I pretty thoroughly addressed the first point in my other ninja post about ELO. As for speaking in terms of percentages, it's really just out of a lack of a better way to communicate it. It's true, OML didn't set percentages for the leagues, but I imagine when they were picking the skill rating thresholds, they had imagined percentiles of the population that they'd like to separate. This is why I use percentages, because the skill rating thing is a black box.

Quote:I think you've misunderstood my hypothetical about the teams who are all 50/50. You're saying that there's a bunch of teams who are very, very good, and I'm sure you're right. But what I was saying is that imagine that everyone was that good: is anyone actually very very good or is everyone just average. This is what I was trying to say: the reason you did not reach super titan before was because you were not *better* than enough teams. You were worse than 2 (the two STs), you were even with several other teams (the very, very good teams you listed, and perhaps others), and better than all the rest. To be a supertitan, you have to be *better* than a certain number of teams; the teams you're even with you're not better than; playing them to 50/50 won't help you reach that level.

Yeah, I agree with this. That's exactly what was happening. But the threshold needed to be lowered to let more people in, it was too excluding. 2 person league is pointless.

Quote:If the old threshold was 10, and the new one is 9, for example, then there is no spot in between.

9.5? Anyway, I tend to agree that we should sit with the current threshold for a while and see how the population settles in. But if OML could do it over, based on my experience, I think they should have used "9.5".

[GreatGonzales]

Hi everyone still reading this thread.

I was inspired to do some number crunching using my 1v1 and 2v2 ranking trends spreadsheets, to see if we can determine the skill disparity in a player population by examining the average rank differential over time.

My reasoning is, if there is not a lot of rank shift in a particular section of the ranking list, then the player population has settled into those ranks and are therefore closely matched in terms of skill. Higher rank differential would suggest a more diverse skill disparity. Take these graphs with a few grains of salt, though: other reasons why the rank differential does not shift much include 1) The player is playing infrequently, or is not playing at all, and/or 2) Matches are more difficult so the player takes a longer time to complete matches. However, I think it's human nature to slow down on the competitive edge when you feel you have met your match. All that to say, I think it's plausible to correlate a low rank differential with low skill disparity.

tl;dr: look at the pretty graphs.

I used 1v1 ranking data from 2/1 until present. For the 1v1 population, we can see that there is very low rank differential up until about rank 45.





I used 2v2 ranking data from 3/1 until present (all that we have available to us). For the 2v2 population, we can see that there is very low rank differential up until about rank 20.





So, I'm not sure how interesting these graphs are, or how useful. But, I just felt like doing it!