PDA

View Full Version : The Ladder Team Assigner System


blln4lyf
04-24-2011, 07:26 PM
Right now it assigns teams 1 then 2/3 then 4 etc, while it used to attempt to balance the ratings between the 2 teams.

This new method causes a lot of games to have a team that has a 75% win percentage and a team that has a 25% win percentage. Also, I tracked the % of points that goes to the 75% win team if they win and if they lose(along with the 25% team ofc) and if they 75% win team wins, they get around 40% of the points pool, while if they lose they lose about 60%. A team that has a 75% chance of win shouldn't get 40% of the points pool for a win and only lose 60% when they lose. This makes it harder for people to obtain their real ranking because these games are lop-sided and not properly balanced through the point scale as well.

I am of the belief that ladder should never have games that are more than 60/40% chance of winning/losing anyway because when there is a team with a 75% of winning, there could be a player decently underrated on the 25% team and a player decently overrated on the 75% team, and both can play to their true ratings, holding all else constant, the 75% team will still win the majority of those games. It doesn't allow people to make as much as a factor to overcome the odds stacked against/with them to get to their true ratings, imo.

If your not going to go back to the old system or at least a modified system of this one that does not allow a 75/25% game and caps out around 60/40%, at least change the point scale so that the 75% win team only wins around 25-35% of the point pool for a win, and loses around 65-75% of the point pool for a loss. I like the new ladder system a lot, but the team assigner system has been downgraded since last season which is causing people to have a large delay in reaching their true ratings, imo.

elxir
04-24-2011, 10:44 PM
I see no problem with this system. It forces the team with the lower overall rated players to actually prove that they deserve to be rated higher, instead of ****ing the high rated players in the butt by handicapping them with dead weight

It feels a lot like the beginning of the old ladder, when players would be like "YES WE HAVE TMIC WE SHOULD DEFINITELY WIN THIS"

whereas later in the season when crappier players became the rule instead of the exception, it became, "**** WE HAVE TO PLAY PERFECT TO EVEN HAVE A CHANCE THANKS TO THIS ASSHOLE"

blln4lyf
04-24-2011, 10:55 PM
I see no problem with this system. It forces the team with the lower overall rated players to actually prove that they deserve to be rated higher, instead of ****ing the high rated players in the butt by handicapping them with dead weight

It feels a lot like the beginning of the old ladder, when players would be like "YES WE HAVE TMIC WE SHOULD DEFINITELY WIN THIS"

whereas later in the season when crappier players became the rule instead of the exception, it became, "**** WE HAVE TO PLAY PERFECT TO EVEN HAVE A CHANCE THANKS TO THIS ASSHOLE"

It gives a team a 75% chance of winning and doesn't penalize them for having such a superior team.

elxir
04-24-2011, 11:00 PM
well better players should win more often

bummeln
04-25-2011, 03:11 AM
My main question-mark with it is just how predictable teams are, you can easily count the good and horrible players and know what kind of team you'll get.

nobodyhome
04-25-2011, 07:20 AM
This new method causes a lot of games to have a team that has a 75% win percentage and a team that has a 25% win percentage. Also, I tracked the % of points that goes to the 75% win team if they win and if they lose(along with the 25% team ofc) and if they 75% win team wins, they get around 40% of the points pool, while if they lose they lose about 60%. A team that has a 75% chance of win shouldn't get 40% of the points pool for a win and only lose 60% when they lose. This makes it harder for people to obtain their real ranking because these games are lop-sided and not properly balanced through the point scale as well.

The problem here is that you are looking at each game indivdually, and that is not how the points are calculated. As described in the post about the rating system, the point are distributed such that a 75% chance of winning team will get only 12.5 points if they win and lose 37.5 points if they lose. The point disparity that you are seeing is a result of the multipliers (as described in the post about the ratings), which takes into account past performance.

blln4lyf
04-25-2011, 05:17 PM
The problem here is that you are looking at each game indivdually, and that is not how the points are calculated. As described in the post about the rating system, the point are distributed such that a 75% chance of winning team will get only 12.5 points if they win and lose 37.5 points if they lose. The point disparity that you are seeing is a result of the multipliers (as described in the post about the ratings), which takes into account past performance.

I understand there are other multipliers, but i don't understand why I haven't seen even one case of it being anywhere near 12.5/37.5 split and always more like 18/23.
1,4,5,7,9,12 vs 2,3,6,8,10,11 right?
Just change it back to trying to balance the teams ratings out which is a known system, and profit imo. Plus the games will on average be more competitive, nothing wrong with that.

Dark_Sage
04-25-2011, 05:37 PM
I understand there are other multipliers, but i don't understand why I haven't seen even one case of it being anywhere near 12.5/37.5 split and always more like 18/23.
1,4,5,7,9,12 vs 2,3,6,8,10,11 right?
Just change it back to trying to balance the teams ratings out which is a known system, and profit imo. Plus the games will on average be more competitive, nothing wrong with that.

Here (http://www.altitudeladder.com/match.php?id=1982&mode=ball_6v6)'s an example of losing a very small amount

But overall I agree with you ball'n, especially after what happened to me in ladder last night. There were a ton of new people, and I kept getting them on my team because none of them had achieved their true ranking. What was even worse was all of them played loopy, which is my main, so I had to go bomber or something else in order to give the team a fighting chane with 3 loopies. It's a little freakin ridiculous when you're mvp in 11 out of 13 games you lost while consistently leading the team in kills as well.

And I do agree that when the team with the better team wins (like they should) they aren't penalized enough. Why can't the good players on the losing team lose a small amount and the players on the winning team win a small amount?

nobodyhome
04-26-2011, 05:49 AM
The old system was a major part of the reason why the top few ranks of ladder were so dominated by bomb runners and ball carriers. Nowadays it's a bit more balanced and that's the way I'd like to keep it.


And I do agree that when the team with the better team wins (like they should) they aren't penalized enough. Why can't the good players on the losing team lose a small amount and the players on the winning team win a small amount?

That's exactly what happens.

zz-
05-03-2011, 07:32 PM
This new team balance system is, frankly, trash. I really love all the effort and thinking that's going on about ladder now, and I wouldn't be so blunt if it wasn't so blatantly evident that the new team-balance system is just terrible. And I also wouldn't be so upset if the previous system weren't so proven to be near-perfect.


If you're the highest ranking player on, you can simply choose who you play against. If you're #1, you can ALWAYS choose the top two opponents you play against. Of course you will never know exactly how the teams shake out, but it's painfully obvious every time I play now that it's either a "good" time to play or a "bad" time to play. And usually, the teams are simply NOT balanced. Game selection is currently the most important skill in ladder. The beauty of the old average team rating balance is:

#1] You can't NEARLY as easily predict any teammates or enemies

#2] If you're the highest rated player on, there is NO ADVANTAGE for playing when there are stronger or weaker players on, because the system ADJUSTS ACCORDINGLY. If there's lots of weaker players on, giving vipmattman the 4th,5th,8th,9th,and 12th best teammates will be way too strong for him -- usually he'd get like 4th, 7th, 9th, 10th, 12th. Look at the matches in the matchlist - the higher rated team wins too much, and most of the games are BORING.

#3] Every game is close

----

The reason you are seeing a mixed balance of players at the top is because the standings are simply SKEWED and INCORRECT right now because the current system is keeping the games ****ty (and mirandas etc cant gravitate to the top as they rightfully should since miranda is the most influential plane in the game). I don't see how there's any argument that the quality of the games has dropped considerably. Just go back to average team rating!!

zz-
05-03-2011, 07:54 PM
Sorry for double post but I started editing that one too much and wanted to clarify:

The new team balance mechanism takes EXACTLY what's good about the old system and removes it. With the old system, no matter which 12 players were in the game, it created a close game that would properly reward underrated players and punish overrated ones. The new system prevents the games from being properly balanced, forces one team to have a higher rating with the top 2-3 players on each team having a good idea which top players they'll be playing with and against.

Since simply fixing how much it rewards players for losing imbalanced games doesn't address the issue of game-selection being the most important factor in moving up the ladder (semi cheating in my opinion), or fix the issue of so many games being imbalanced (BORING), I think we really should go back to the average-team-rating balance mechanism.

shrode
05-04-2011, 08:40 PM
I was going to make a thread about this issue, but see you two have it covered already, and very eloquently, I must add.

I will stress zz's third point,as I feel that it is the most important one: Close games are more fun.

And, as the other two have already pointed out, the very best player can easily chose when to play to have the greatest chance of winning. This causes the best player to continue hitting that 'trending' multiplier, and because they can continue to select when they play, they will profit from this multiplier more than they will be harmed.

Thus, the game currently is too much "I hope i'm the 4th best rated person so that I get donk and will probably win" and not enough "alright I don't have donk but the teams are still balanced cause he has yank and prince, so I better try hard cause this game will be close."

nobodyhome
05-04-2011, 10:39 PM
Yes, it is true that the new balancing system makes games more unbalanced (in terms of ratings and win percentages) than they were in the past. However, it is untrue that the system doesn't punish people enough for playing in the advantaged team. As said before, the system is the same as before in that the side with lower ratings will get proportionately more points for winning and lose proportionately less points for losing. This is, of course, before the multipliers, which you can see by looking at the matchlist and looking at the column under "change".

Once again, this system was put into place because putting people into always balanced matchups is not the best way to accurately rate people. The problem with the old system lies in that the people who are highest rated are not necessarily the ones who are the best at the game, but simply are the ones who are most able to single-handedly drag a bunch of noobs into victory.

A point of clarification about the trending multiplier: the trending multiplier isn't based off your win%, but rather your net gain in points pre-multiplier. This is so that if you are always in the advantaged team, the trending multiplier will still reflect accurately whether you were playing better than your level or not.

Having 100% balanced games, once again, is not required for an accurate rating system, and it can be detrimental as previously stated. Closer games being more fun is also merely an opinion--I postulate that reasonably-close games are sufficient, and that once you get balanced enough so that the underdog doesn't get simply demolished, it may be fun to have games in which the underdog wins and scores more points as a result.

Ribilla
05-04-2011, 10:58 PM
I also feel that the ratings system is flawed, if only because some of the games are so unbalanced one team has almost no chance of winning. Also, if there are only 10 people in ladder and their ranks are all reasonably far apart, you get the same (unbalanced) games over and over. My last few games have been 70% expected win for other team and low and behold they have won all of them. I quite like the idea that none of the games are exact, but can I suggest that some way of limiting the unbalance is put in?

I have no idea how to implement this, but perhaps the server should randomly select a number between 40 and 60 and then balance the teams so that the win% (as close as possible) matches this number. This would give varied games without the predictability or the vastly unbalanced games.

zz-
05-04-2011, 11:41 PM
I have no idea how to implement this, but perhaps the server should randomly select a number between 40 and 60 and then balance the teams so that the win% (as close as possible) matches this number. This would give varied games without the predictability or the vastly unbalanced games.

This sounds like a great idea. Somehow balance the teams "randomly" but with a small range of average rating, or expected win %. This solves the problem of players choosing who they play against (why is this being ignored? its the most glaring problem with this system), AND eliminates the very uneven games, AND keeps the varying evenness of teams that many people seem to want.

seems pretty legit...

nobodyhome
05-05-2011, 12:14 AM
I consider the staticness of the balancing algorithm right now to be a legit problem (allowing people to choose who they play against, as well as making the same teams happen over and over). I will put in a fix for this asap but this is likely to create a greater chance of unbalanced teams. Let's reserve judgment on that though until it happens.

Ribilla
05-05-2011, 12:43 AM
Do you think my idea, or some variation on it, is workable?

elxir
05-05-2011, 02:43 AM
i like the new system because the best player generally gets the best chance of winning...the ****tier players shouldn't get a better chance just because they are ****tier

Ribilla
05-05-2011, 10:34 AM
i like the new system because the best player generally gets the best chance of winning...the ****tier players shouldn't get a better chance just because they are ****tier

Maybe, but in this case the second best player doesn't get the second best chance of winning, he gets ganked having the worst player.

shrode
05-06-2011, 04:34 PM
The only probably with the old matchmaking system was that it took forever for scrubs to get down to where they needed to, so they often ruined game balance. However, the new ladder points system already addressed that, so no need to fix a non-existent problem and change the matchmaking too

Ribilla
05-06-2011, 04:53 PM
Found another problem:

Playing ladder last night there were only about 10 people on, so everyone played everygame. Blln was the highest rank and I was second (initially anyway, I dropped a lot of points), now blln was the only primary runner playing those games so it stood to reason that his team would probably win. Problem is that I would never be put on the same team, so despite playing well most of the games I still lost all of them.

I know that I should learn to run, especially if I want to be ranked highly, but I feel this system punishes people in a 10 player situation by not modifying the teams enough, so if one team has a player who is clearly better than everyone else, the same team will lose every time. This was not the case with old ladder.

I just want to say that I think the points calculating algorithm and everything else this season is great, it's just the balancing system I have a problem with.

elxir
05-06-2011, 06:10 PM
sounds like it's working as intended

if you can't muster up someone from five ladder players who is capable of bomb running then that is just a sign of a lack of players who understand TBD and who also don't understand how to protect a bomb runner and work with timing.

You don't need fancy moves in a miranda to be an effective bomb runner. You need knowledge of spam lanes, timing, proper running lanes, and the like. Your teammates need these same things.

These same skills will allow you to defend against a runner such as blln. Each plane needs to know their role, and your team as a group needs to understand how to shut down an attack and muster a counterattack using the game's built in timing.

This new system is simply exposing those who undestand the strategic underplay of the game modes, and those who don't.

Ribilla
05-06-2011, 08:56 PM
sounds like it's working as intended

if you can't muster up someone from five ladder players who is capable of bomb running then that is just a sign of a lack of players who understand TBD and who also don't understand how to protect a bomb runner and work with timing.

You don't need fancy moves in a miranda to be an effective bomb runner. You need knowledge of spam lanes, timing, proper running lanes, and the like. Your teammates need these same things.

These same skills will allow you to defend against a runner such as blln. Each plane needs to know their role, and your team as a group needs to understand how to shut down an attack and muster a counterattack using the game's built in timing.

This new system is simply exposing those who undestand the strategic underplay of the game modes, and those who don't.

I did do counter attacks and I can defend against runners, better than anyone else, if you would care to look on the tbd stats page. I even scored some hits on the neutral maps. However I am not ever going to be as good a runner as blln, if only because both accurate plane control and use of TA are nigh on impossible using a pure mouse control scheme. What I think is unfair is continually pitting the same set of players against each other when it is obvious one team will triumph.

MajorPayne257
05-06-2011, 09:55 PM
Yeah I've been running into this problem every time both ball ladders start up. I went on a 5 game losing streak in the original server because the teams were the exact same (maybe one switch here and there), and then I went over to the second server and won 4 games because those teams were similar.

Tekn0
05-07-2011, 02:36 AM
Yeah I've been running into this problem every time both ball ladders start up. I went on a 5 game losing streak in the original server because the teams were the exact same (maybe one switch here and there), and then I went over to the second server and won 4 games because those teams were similar.

+1.
Same.

chars.

Duck Duck Pwn
05-07-2011, 03:27 AM
Honestly, i feel as though the only static conditions should be that the 12th and 11th rated players are on separate teams, because having the two worst players even if you have other players to make up for it blows. Maybe 1 and 2 separate as well but that i care less about. I'm fine with the rest having some form of fluctuation.

elxir
05-07-2011, 05:28 AM
Honestly, i feel as though the only static conditions should be that the 12th and 11th rated players are on separate teams, because having the two worst players even if you have other players to make up for it blows. Maybe 1 and 2 separate as well but that i care less about. I'm fine with the rest having some form of fluctuation.

1 vs 2/3, split 11/12, and 4-10 randomized based on average rankings could work in theory, but a problem arises when the 2/3 get the 4 also and the 1 is kind of stranded on an island

bummeln
05-07-2011, 08:29 AM
What about adding some different matchmaking methonds? Put a selector before setting teams with something like 40% chance of the current system, 40% old system 15% experimental (like setting teams by plane composition) and a rare chance of craziness like full random or if there's a mechanic in place captain games :)

This could also provide a decent way to compare which methods are enjoyed more even though it would probably end with a consensus that everything is horrible just like every map.

Ribilla
05-07-2011, 12:27 PM
What about adding some different matchmaking methonds? Put a selector before setting teams with something like 40% chance of the current system, 40% old system 15% experimental (like setting teams by plane composition) and a rare chance of craziness like full random or if there's a mechanic in place captain games :)

This could also provide a decent way to compare which methods are enjoyed more even though it would probably end with a consensus that everything is horrible just like every map.

I think this would go badly. At least there is consensus that no one (bar lix) likes the current system.

blln4lyf
05-07-2011, 04:19 PM
Oh another thing, you could add something so if there are 2 new ppl they get put on separate teams.

shrode
05-07-2011, 06:17 PM
just do old system, no real need for complicating it with static splits.

blln4lyf
05-19-2011, 09:21 PM
Basically, with the ratings not being equal or close to equal between the 2 teams, it causes uneven games. This is obvious, and it is addressed by giving less points to the higher rated team, etc. but that doesn't fully fix the issue imo.

Lets say player 1 is ranked 3k and player 2 is ranked 2000. Lets say player 11 is ranked 1.2k and player 12 is ranked 700. Last season this would be addressed to cause the ratings to be equal between the 2 teams anyway, but now, it's possible for a team to get player 1 and 11 while the other gets 2 and 12, causing team 1 to have a 80% or higher win %. I can't prove that this system is flawed though, but I personally think that it isn't working. Closer games will allow people to make more or less of a difference in every game to properly reach their correct rating, while lopsided games could have an underrated rated player play well above their rating, but still lose almost every time just because the teams aren't balanced. Unbalanced teams, imo, cause outside factors(team balance mainly) to come into play more, which lets people have less of an effect on if they win or lose than if there is balance.

The old system was a major part of the reason why the top few ranks of ladder were so dominated by bomb runners and ball carriers. Nowadays it's a bit more balanced and that's the way I'd like to keep it.



However, the new system did nothing to change this at all, the reason those players are the highest is because they have the most impact on the game, it has nothing to do with the way the team assigner is set up at all.



This system was put into place because putting people into always balanced matchups is not the best way to accurately rate people. The problem with the old system lies in that the people who are highest rated are not necessarily the ones who are the best at the game, but simply are the ones who are most able to single-handedly drag a bunch of noobs into victory.

Having 100% balanced games, once again, is not required for an accurate rating system, and it can be detrimental as previously stated. Closer games being more fun is also merely an opinion--I postulate that reasonably-close games are sufficient, and that once you get balanced enough so that the underdog doesn't get simply demolished, it may be fun to have games in which the underdog wins and scores more points as a result.

I personally, love the new system, minus the team assigner.

I know that close games being more fun is merely an opinion, but its one that most people believe, so why stray away from what the public wants in that regard? And lets be real. Who likes playing in games that are not close compared to games that are close when playing competitively?

As for the problem with the old system that you stated, "people who are the highest rated simply are the ones who are most able to single-handedly drag a bunch of noobs into victory" instead of being the best players, it takes being one of the best players to be able to drag a bunch of noobs to victory. You giving this problem is your opinion and I personally do not think it is a problem at all. If you look at the top players in both tbd and ball ladder, the majority are the same from last season. This doesn't prove that this issue you stated isn't an issue, but it sure dispels it a bunch.



I think the public wants more balanced games, so even something that sets the upper limit to 60% or even 65%(though I'd much rather see 60%) for a game would create much better games, and allow people to make more of a difference in the game, like I described near the start of this extremely long post. Ideally though, I'd change the team assigner to what it was last season, it makes the ratings as close as possible for the 12 people who are in.


Oh, and a totally random point, if there are two relatively new players(less than X amount of games played), they should probably be put on separate teams.

[Y]
05-19-2011, 10:58 PM
Losing less points per unbalanced game doesn't make up for the fact that it's unbalanced imo. I'd much rather play games balanced around the total ratings.

shrode
05-20-2011, 12:47 AM
I think the public wants more balanced games.

Refer to my poll thread if you need evidence.

Carlos98
05-20-2011, 01:46 AM
I think the public wants more balanced games
this.

I'm ok with these changes with the team assigner to find a good system. To be frank S1 wasn't perfect (takes forever to get to true ranking), neither was S2 version1 (people can "pick" their games, teams are predictable), and definitely not S2 version2 (absolutely no balance).

This S2v2 (whatever it is currently) is absolutely terrible. Unfortunately I really don't have any ideas as to a better system (other than S1 or S2v1). But keep implementing others if people have more ideas.

zz-
05-20-2011, 06:52 PM
I agree fully with balln's post. The notion that the new system somehow rewards a different approach to the game is completely unfounded in math and logic - no matter what method you use to balance teams, the highest rated players will ALWAYS be the ones who are able to "drag the most newbs to victory" - this is just a clever/biased way of saying "win the most games with any mixture of players."

The only possible benefit to NOT balancing the teams according to average rating, is to manually insert more variation in the "winnability" of each game for each player - to see if players can avoid losing as a favorite, or scrape together more wins as an underdog. However, balancing teams according to average rating already contains a large enough amount of variation in REAL overall match expected percentage, because of things that are decided completely randomly (i.e., team 1 gets 6 players who all usually play loopy while team 2 gets a good balance, team 1 has an overrated player and a drunk player and team 2 doesnt - there is no shortage of variables in ladder.

As balln said - by creating games that are not balanced according to average rating, you're forcing the game to be decided by factors outside of the player's control. With the old system, correction to the ladder standings occurred at the margin of winning and losing --- that is, every game was ~50% for either team - you fought to move up in the standings by WINNING the games you played. Winning and losing is now WAY more contextual and in most games, the margin by which you can typically outplay/underperform your rating does not swing your team from a win to a loss or vice versa - that is, your grandmom could play for you in a lot of your wins and you'd still win. And a top 10 player could play for you in a lot your losses and you'd still lose. It is this disassociation of cause-and-effect of playing well -> winning games that has players frustrated. Especially when you just waited 12 minutes to play, got one automatic loss, and now have to spec another game.

I realize that's present in any system, but the new system manually forces an unnecessary extra amount into it. However, most players (myself included) DO see the attractiveness of some variation in average team rating. So I think the idea of capping the expected win% for one team to a certain number is a good one. The consensus appears to be ~60%, which for me would be the absolute max I would consider. As the ladder standings become more and more accurate, the unfairness of 60/40 games will increase. To me, 55-57% would be an ideal cap.

And as a final note: The second version of the team balancer this season is actually worse than the original. It addressed ONE problem: Now the highest rated player only knows the #2 rated player he'll play against, instead of #2 and #3. It is still exploitable in every other way the first one was. HOWEVER, this new system creates teams that are always either equally or more unfair. Using any cookie-cutter method to split 12 players into teams/pairs based on their ranking 1-12 will give each two teams a total sum of the ranks of their players. At first this season, it was locked at 1-4-5-8-9-12 vs. 2-3-6-7-10-11, which is 39 v 39. Now it splits 1v2, 3v4, 5v6 etc and assigns one of each pair to a team - so now it can be anywhere from 36 vs. 42 to 42. vs 36 -- The best case scenario for fairness, hilariously, is when it randomly chooses the method we used at the beginning of the season.

Either way, both team balance mechanisms put into use are inferior to the one used in season 1 - just thought I'd make it evident that the newest one traded a fraction of its exploitability for an increased tendency to imbalanced games. I think just about everybody agrees that balancing teams by average rating (slightly less strictly than s1) is the best approach for creating balanced, fun, fair games and eradicating the opportunity for players exploiting the method by which teams are built.

Edited for clearness. What a long (and brilliant) post.

elxir
05-20-2011, 08:11 PM
IMO, increased ladder size is in part the culprit.

Due to how static teams are when players do not change RANKING after a game, a full ladder server in essence has two "groups" of players cycling in and out of the game. In effect, you are stuck with the same 7-8 players every game you play, until they leave and are replaced by someone else.

One of the biggest problems of the old season was teams being comprised of the same players, over and over and over. This ****s **** up and is not fun, especially for the losing team.

There needs to be a way to diversify teams more - I think the balancer is pretty good, but the problem is that it is only balancing half of the server in a given game and either you get lucky your first game, or you don't and you're stuck with the same peeps for an hour.

zz-
05-20-2011, 08:15 PM
Also I wanted to add without editing again :

I consider the staticness of the balancing algorithm right now to be a legit problem (allowing people to choose who they play against, as well as making the same teams happen over and over). I will put in a fix for this asap but this is likely to create a greater chance of unbalanced teams. Let's reserve judgment on that though until it happens.

Looks like nobo knew about the last part of my post before he put the system in. Forgot about this post. Also overlooked that the new system put a dent in the problem of the same teams happening over and over again. Still, using average team rating would fix this even further and create better games.

Edit: Just saw elixir's post. Are you serious? You realize that under the old system, there were 924 possible combinations of 6v6, and the system chose the one with the closest two average ratings. Thus, after every game, every player had shifted 25pts in a direction and it was extremely unlikely that the same 6v6 would occur. With the new system, there's only 32 possible combinations of 6v6, and it chooses one of them. Which do you think is more likely to make more similar teams? This is a MUUUUUUCH larger problem in s2 than s1.

Re-edit: d'oh, i misinterpreted your post. apologies. anyhow, this thread should be over, everyone basically agrees that teams should be balanced according to average team rating / team expected win%, with a little wiggle room. All that needs to happen is nobo to find a way to implement it....

p.s. 2nd double post in this thread gg

sunshineduck
05-20-2011, 09:59 PM
i think that you all are horribly misjudging the purpose of ladder. it really does not matter at all how balanced the teams are as long as the formula is accounting for the expected win% when distributing points, which it most certainly does. if the heavily favored team won literally every time it would actually be problematic, but that's not the case.

the system implemented in season 2 is intended to expedite the process of reaching each individual player's "true" rating, which it does much better than any other formula used before. that's the entire point of ladder - to accurately rate players based on their positive impact to the teams they are on. that's it.

blln4lyf
05-20-2011, 10:15 PM
I don't think we have, SSD. S2 works better because of the changes to streaking etc, instead of it just being static. We are arguing that making the win % of the teams close to even will actually make it even better in terms of correctly rating people because they it allows people to more easily effect games, for reasons stated in posts above by me and zz.

Also, even if you say it won't make the system better(which I'd strongly disagree with obviously) and just keeps it the same, what is wrong with keeping the rating system's effectiveness the same while creating better games. Either way its a win imo.

zz-
05-20-2011, 10:18 PM
You're missing the point - the current system works great, everyone agrees on that. We're not complaining about the RANKING system, just the method by which the ladder balances teams. The point is, why have so many games where the outcome is uninteresting and unrelated to how well you play, when you could just as easily have good, fun, interesting games.

By your theory, a ladder that creates nothing but 99%/1% games but accurately rates the players is no worse than a ladder that creates close games much more frequently, and rates players equally accurately. This thread is full of people that disagree with that theory.

Rainmaker
05-20-2011, 10:41 PM
Regarding ZZ- post:

There is something to have in mind; the current system feeds itself from data generated from the system.
That being said, if the system is severely flawed the, the outcoming data (data2) will most probably be flawed too, due to being based on incoming flawed data (data1).

Never the less, I think; that some small changes could be applied to team balancing, as introducing forcing team composition based on planes played, player characteristics (bomb runner, goal scorer, goal blocker/bomb blocker, support's ,etc)

Carlos98
05-20-2011, 10:56 PM
i think that you all are horribly misjudging the purpose of ladder. it really does not matter at all how balanced the teams are as lon....

I think you are horribly misjudging half the purpose of ladder.


The idea behind the ladder was to give players a chance to rank their competitive ability and then challenge players of approximately equal skill


Want to have a real way of determining individual skill? Itching to play competitive games?


Stuff I put in bold no longer applies.

Seriously how do you guys get off to just a numeric representation of skill? Shouldn't the fun be had in the actual games, ones that make you earn the win?

Huge upsets are fun sure. But most of the time it is like one team of people versus 6 bots. I won 3 straight games last night playing bip (I'm god awful at bip) because I got really lucky with unbalanced teams. Honestly I can't even get warmed up most nights anymore because the other team is either tremendously terrible or I'm on the **** end of the teams.

yankinlk
05-20-2011, 11:27 PM
Regarding ZZ- post:



Never the less, I think; that some small changes could be applied to team balancing, as introducing forcing team composition based on planes played, player characteristics (bomb runner, goal scorer, goal blocker/bomb blocker, support's ,etc)

OOOh OHHH what about something as simples as that? For Ball, pick teams based on goals scored? Lately some matches there isn't even one ball person willing to take the ball at start - not even one competent randa.

Jayfourke
05-21-2011, 02:19 AM
According to Danielle and Ball'n, the team thingy is catastrophically broken. Personally, I couldn't give a flying frak.

(I mention this solely because I'm bored and they were complaining incessantly during a game earlier. Thank me later, you two.)

elxir
05-21-2011, 07:37 AM
Regarding ZZ- post:

There is something to have in mind; the current system feeds itself from data generated from the system.
That being said, if the system is severely flawed the, the outcoming data (data2) will most probably be flawed too, due to being based on incoming flawed data (data1).

Never the less, I think; that some small changes could be applied to team balancing, as introducing forcing team composition based on planes played, player characteristics (bomb runner, goal scorer, goal blocker/bomb blocker, support's ,etc)

well, the same players are at the top as were there last season

just sayin

MajorPayne257
05-21-2011, 10:07 AM
well, the same players are at the top as were there last season

just sayin

and last season's games were more fun

just sayin

Rainmaker
05-22-2011, 04:21 PM
well, the same players are at the top as were there last season

just sayin

Logic 101:

Season 1 was fun
Top 5 for Season 1 were:


Top 5 for Season 2 are:
ergo, Season 2 is Fun.


That is a falacy sir.

EDIT: sry this was for maxpayne.

elxir
05-22-2011, 06:06 PM
ahh, gotcha

MajorPayne257
05-22-2011, 09:39 PM
Umm, what?

zz-
05-22-2011, 10:23 PM
I believe what he is trying to say is that he is crazy

Urpee
05-22-2011, 10:35 PM
I think the balance does have issues simply because quite atrocious imbalanced are possible. Here is an example:

http://www.altitudeladder.com/match.php?id=5066&mode=ball_6v6

Notice how the higher scored team has the best player but not the worst. Their best outscores the opponents best by some 900 points and their worse outscores the other teams worst by some 400 points. Winning expectancy of 78%.

I don't think such a team sorting should be possible. Clearly the worst player here should have been sorted with the best to bring the average score of the teams at least somewhat in line.

Also games of 100 points differential are more common than is good. I think it has gotten worse since the recent changes adding randomization.

yankinlk
05-23-2011, 12:04 AM
sigh.

Some of the same poeple and the reverse scenario happens, go figure.
http://www.altitudeladder.com/match.php?id=5070&mode=ball_6v6

Any team can beat any team at any time. The highest ranked player still loses roughly ~37% of his games - it happens ya know.

nobodyhome
05-29-2011, 09:04 AM
Tweaked the balancer so that it attempts to balance within at most 40/60 win probability spread. Post here if there are any more issues.

Ribilla
05-29-2011, 11:26 AM
Tweaked the balancer so that it attempts to balance within at most 40/60 win probability spread. Post here if there are any more issues.

I love you.

Carlos98
05-29-2011, 06:00 PM
Tweaked the balancer so that it attempts to balance within at most 40/60 win probability spread. Post here if there are any more issues.

Awesome! thanks

blln4lyf
05-29-2011, 06:41 PM
Awesome! thanks

FYI in case you didn't see the other thread, this fix is not working currently, but nobo is going to look into it and fix it asap.

Rainmaker
05-30-2011, 01:08 AM
Tweaked the balancer so that it attempts to balance within at most 40/60 win probability spread. Post here if there are any more issues.

In other words:
- Team 1 has 70 more average free points than Team 2;
- Team 1 has a total of 350 free points.
- Sum of all ratings in Team 1 = Sum of all ratings in Team 2 + 350 points.

Doesn't seem harmful 350 points, but 60% vs 40% seems a little too much?

Ribilla
05-30-2011, 01:20 AM
In other words:
- Team 1 has 70 more average free points than Team 2;
- Team 1 has a total of 350 free points.
- Sum of all ratings in Team 1 = Sum of all ratings in Team 2 + 350 points.

Doesn't seem harmful 350 points, but 60% vs 40% seems a little too much?

IMO at 60-40 you still have a decent chance of winning, it's fairly unlikely, but it's definitely not a shutout like the 72-25 games. Having a little variation is a good thing.

Pieface
05-30-2011, 02:24 AM
definitely not a shutout like the 72-25 games

Just thought I'd leave this here:
http://www.altitudeladder.com/match.php?id=5718&mode=ball_6v6

Also, you have to be a special kind of bad to win 180 points.

Ribilla
05-30-2011, 02:52 AM
Just thought I'd leave this here:
http://www.altitudeladder.com/match.php?id=5718&mode=ball_6v6

Also, you have to be a special kind of bad to win 180 points.

yeah but wins like that don't happen 27% of the time. Also, looks like a lot of those players are badly rated, IMO.

Carlos98
05-30-2011, 02:39 PM
Just thought I'd leave this here:
http://www.altitudeladder.com/match.php?id=5718&mode=ball_6v6

Also, you have to be a special kind of bad to win 180 points.

You aren't even listed in the stats column underneath. This is all kinds of glitch.

MajorPayne257
05-30-2011, 10:47 PM
Gasp Pieface deserves a ban for the leave.

naethy
06-01-2011, 06:17 PM
2 quick notes:

First, I have seen teams with as low as a 25% chance of victory pull off a 6-0 win in ball. I would guess that this has more to do with team comp than anything else; for example, me and Goal played a match last night and lost badly even though we had a ~50% win chance. The reason is that we both play our best as randa/loopy, but I had to whale because the rest of our team simply couldn't. At that point I had a ~2000 rating, but I was playing whale like someone with a 15 or 1600. Side note: if someone could teach me to whale I would be eternally grateful.

Second, having one person in ladder over 500 points better than the next person causes major imbalances. A few days ago I was playing with Dragon and a few low-ranked people and I was always 2nd or 3rd rated playing. As you can guess, we did pretty poorly until Dragon's internet exploded.

Finally, I really LIKE the fact that good players don't always get **** teams. I'm not one of those players, but I think they deserve to play with good people and not always play "watch the n00b". I also think the balancer is working pretty well as-is, at least for me (I haven't had any crazy streaks).

I guess the point of this post is teach me how to whale.

-- Slight

Kuja900
06-20-2011, 05:08 AM
Please bring the old system back. It wasn't great, but it wasn't nearly this bad.

XX1
06-20-2011, 05:33 AM
Please bring the old system back. It wasn't great, but it wasn't nearly this bad.
agreed especially with the stuff in color

+1

elxir
06-20-2011, 06:12 AM
disagree..

Ribilla
06-20-2011, 12:48 PM
I think the most recent update has made the balance pretty good again. People just need to be less **** is all.

Fartface
06-20-2011, 06:58 PM
I don't like how you get extra points just by playing more games... there are a few people who have played so many games that they have a high rating even though they lose more than they win

Kuja900
06-20-2011, 09:04 PM
My biggest problem with it lies in its attempt to adjust to the inactivity of participants creating like 100 point swings in single games. I think while an interesting idea, it is causing a lot of frustration among some players.

naethy
06-20-2011, 10:45 PM
I don't like how you get extra points just by playing more games... there are a few people who have played so many games that they have a high rating even though they lose more than they win


Ummm, this does not happen. The extra few points you gain will be taken away just as quickly if you don't play like your new rating. The extra points for a win simply increase the TOTAL number of points in the SYSTEM, causing EVERYONE's rating to go up slowly over time (if they play a decent amount). If someone leaves for a significant amount of time, their relative rating will decrease, but playing a decent amount won't increase your rating directly except to keep it in line with "inflation".

Carlos98
06-21-2011, 01:06 AM
The extra points for a win simply increase the TOTAL number of points in the SYSTEM, causing EVERYONE's rating to go up slowly

The top 10 is increasing at a rate of about 100pts/week easily.
Top 25 at least 75 a week. (FYI we are 11 weeks in, do the math if you disagree). There is no basis for consistent ratings across different sets of players. I don't see why it has to trend up at all. The way it is now, has late joiners start lower to make it harder to get to the top. Yet there is the trending multiplier to get people to their "true" rating quicker. How is that not conflicting?

Ribilla
06-21-2011, 02:23 AM
Lol you cannot introduce more points by playing more games, because all the games are zero sum. The total number of points per player is always 1500.

The top players are going up in points because the balancer is based on ranks rather than points so aside from earning less points when your rating is high, the best and worst players are completely unbounded. This has the knock on effect of making everyone slowly move away (above or below) from 1500. I did point this out when this balancer was created, but it doesn't really matter. While the points go up, the ranks stay the same, so the balancer will produce the same teams. New players tend very quickly with their high uncertainty, so it doesn't effect them much either.

Basically playing more doesn't get you more points, but if EVERYONE plays more and the ranks stay the same, then everyone moves further from 1500 by a factor of the orginal point difference. What I mean is, if a sample of four players looked like this in week 3:

1. 2000
2. 1600
3. 1400
4. 1000

Then after, say, 10 weeks it might look like this:

1. 2250
2. 1650
3. 1350
4. 750

Obivously this is in a perfect world with super accurate ratings.

Also, whoever said "playing more gets you more points". Maybe this is because people who play more get better, and so achieve a better ranking. I think this is probably just ladder working as intended.

Ribilla
06-21-2011, 02:24 AM
You can see this effect with ball, everyone plays more games, so the extremities have much higher and lower points.

andy
06-21-2011, 02:30 AM
New system isnt zero sum, but that IMO isnt a problem.
Ill try to explain it as fast as possible:
- if you play loads of games and gain 'free points' from mass gaming your ladder rating will be above your true rating therefore you will start getting unbalanced teams and eventually you will reach your true rating again.
- if you dont play for some time and all the other players are boosted by inflation you will just be underrated and get some easy wins before reaching your true rating.

sunshineduck
06-21-2011, 03:04 AM
I don't like how you get extra points just by playing more games... there are a few people who have played so many games that they have a high rating even though they lose more than they win

i'm going to assume you mean me, since i'm the only one in either ladder that is in the top 25 with a winrate lower than 50%.

i'm also pretty sure that it's because i tend to play a lot against players with obscenely high ratings (mled, TA's) and consequently lose a lot of games in which the ratings are stacked in favor of the opposition. conversely, i win when the teams are more even or if my team is underrated/his team is overrated enough to where i can win the game. or i'm just super lucky.

ed: afaict the new system is far better in getting players to their true ratings in a timely fashion and also at having those true ratings be more accurate, but is also worse at creating balanced and competitive games. whether you think competitive games are more important than having accurate ratings is of no consequence - it literally does not matter what you think. i always hear whine about majority this and majority that, but the majority didn't create ladder and has no bearing on how ladder is run. if you don't like it, go make your own ladder, i'll gladly play!

Urpee
06-21-2011, 05:19 PM
The current system is more complex/less intuitive, but I have no problem with it.

It's important to understand two things, as Andy already hinted at:

1) It's not zero-sum. That's more an underlying mechanism and not that visible for most.
2) You can have above 50% ratings that are locked in more than before and you can have below 50% ratings that are locked in more. Or in other ways, the system no longer strongly converges to a 50% win/lose ratio given that you do not get the same number of points each game.

Because that second point hasn't been explained as much let me give an example.

If you play 50 matches, 30 of which you play against a weaker team (you should win) and 20 against a stronger team (you should lose) and that indeed happens. Let's assume that the differential is the same (X points) then you have gained 10*X points simply because you had more weak teams to play than strong ones. So the luck of which teams you are on will impact your score more than a 50% balance. This is made stronger if you lose against weaks but win against strong. You get more points than the win % reflects, so you can outpace the the 50% balance. we see this in top players whose win 50% is higher than in ladder1.

In fact there is a further effect, already mentioned above, in that a person can have a high rating but a below 50% win ratio. That's achievable if you have more weak games that you lose and few strong games that you win. Because you win the strong games you gain more points and need fewer games to achieve the rating. The rating will no longer help you stay close to a 50% win ratio. That's fine though. The indicator of performance is now the rating, not the win ratio, though of course the win ratio does is a steering factor in it all. But it's non-linear because the point gains are non-linear.

I actually think this is good. Ladder does display pretty good rate-locking behavior for quite a few people. I.e. many people do seem to level of and stay in a rating range, which is a desirable property. Adjustments do happen quickly (within just a few games), which is also good.

So overall I see the point system to be a clear improvement over ladder1. My main point of future improvement would be in the balancer at times it still allows too big a discrepancy for my taste. But I find the current system to be overall quite nice.

mikesol
06-21-2011, 06:27 PM
Lol you cannot introduce more points by playing more games, because all the games are zero sum. The total number of points per player is always 1500.

The top players are going up in points because the balancer is based on ranks rather than points so aside from earning less points when your rating is high, the best and worst players are completely unbounded. This has the knock on effect of making everyone slowly move away (above or below) from 1500. I did point this out when this balancer was created, but it doesn't really matter. While the points go up, the ranks stay the same, so the balancer will produce the same teams. New players tend very quickly with their high uncertainty, so it doesn't effect them much either.

Basically playing more doesn't get you more points, but if EVERYONE plays more and the ranks stay the same, then everyone moves further from 1500 by a factor of the orginal point difference. What I mean is, if a sample of four players looked like this in week 3:

1. 2000
2. 1600
3. 1400
4. 1000

Then after, say, 10 weeks it might look like this:

1. 2250
2. 1650
3. 1350
4. 750

Obivously this is in a perfect world with super accurate ratings.

Also, whoever said "playing more gets you more points". Maybe this is because people who play more get better, and so achieve a better ranking. I think this is probably just ladder working as intended.

Perhaps you're still thinking of season 1, whereas everyone else is talking about season 2. Urpee and Andy already pointed out that this game is not zero sum (and explained the intricacies of this system fairly well) - so the arguments you are making are mostly void. Furthermore, if you play the same people against each other over and over - it will not always result in the same teams because of the randomness introduced into the system.

Also - there are indeed points being added. If you want to see how the rating system works just read this (http://altitudegame.com/forums/showpost.php?p=116069&postcount=1) post. At the very bottom nobo points out that a few points are added each game.

Just wanted to clarify that :)

Kuja900
06-21-2011, 07:18 PM
The current system is a whole lot more complex than the previous, yet less effective.

sunshineduck
06-21-2011, 07:56 PM
The current system is a whole lot more complex than the previous, yet less effective.

less effective in what way?

elxir
06-21-2011, 08:30 PM
Could ladder conceivably do this...

Let's say you have 20 players in a tbd server.

Ranked 2000, 1990, 1980, 1970, 1960, 1950, 1940, 1930, 1920, 1910, 1900, 1890, 1880, 1870, 1860, 1850, 1840, 1830, 1820, 1810.

Why not have ladder just take from this pool, the ten highest rated players. Thus, 2000-1910 play in game 1, sorted by the current method.

In game 2, 1900-1810 would play, due to the "play every other game" rule, sorted as usual.

In game 3, once again the ten high ranked players would play. Repeat ad infinitum.

Using this method, these two groups of players would always face off against their same opponents, with the only changes being which players are assigned to which teams based on (a) new ranks and (b) the randomizer for 3/4, 5/6, etc.

The only way these two groups of players (higher ranked, lower ranked) would ever interact is via the introduction of a new variable, e.g someone leaves or joins the server. In the former instance, a leave would then promote the next highest ranked player (or low ranked) into the other group, due to the need for an additional player.

In the latter instance, the newly introduced player would automatically replace the player in his "group" who is closest in rank, and the randomizer for team selection (not player pool selection) would then do its work.

The effects of this system are simple: you are placed with players most similar to your skill, and you play with that group until either (a) someone leaves and changes the system, or (b) someone joins and changes the system.

This would result, in my opinion, in what everyone wants: games that are within their skill level.

The original idea for an "elite" ladder server was squashed due to a lacking player base. This system would not be such a thing. Instead, it would look at all the players available in the server, and pick the group that are closest in skill.

Now, minor problems would arise (e.g. it would be the same as it is now) when the server was not at double-game capacity, but for the most part I believe that sorting players in a more skill appropriate set would be of great value to ladder.

sunshineduck
06-21-2011, 08:45 PM
what do you do when the server isn't at 20+ players?

and this wouldn't work as well as in your theorycraft because the odds of having a group of players in that server whose ratings are that close are extremely low

nobodyhome
06-22-2011, 02:54 AM
A lot of the misconceptions that people have about ladder is due to the fact that people think your rating should be based off of your entire history of play. It isn't and it should not be--ideally, your rating is representative of how well you perform in the here and now (more practically, how well you performed in your past 20 games). If you play like crap for the first two months of ladder, dropping to a rating of 1000, and then suddenly you start playing like a 3500+ rated player, your rating should shoot up to 3500, once again ideally instantaneously but more practically in a few games.

This may sound funky but with some thought it should make sense. You should never be held back by your previous play--you should be instantly rewarded for improving play and instantly punished for crapping it up. This also serves to make things like balancing work, because people are rated accurately as to their current level of performance.

This is what the rating system strives to achieve as of today, and has been what it has always striven to achieve even in season 1. As a consequence of this, things like winning percentages as well as number of wins and losses are completely irrelevant to the discussion of the rating system. Unfortunately, people tend to gravitate toward those kinds of numbers but they really cannot be used to argue to whether the system is working or not. I would not worry in the least if it happened to be that the top 100 players all had a winning percentage of 10% while the bottom 100 players all had a winning percentage of 50%. Conversely, if there was an exact correlation between rating and win percentage I would never point to that as an argument for the system working 100% as intended.

Rainmaker
06-24-2011, 04:42 PM
^this; seems a perfect explanation of how it should (and most probably does) work the ranking system.

Win % isn't correlated as "how good you are"; analogy to "how good you are" to the kill's scoreboard.

Fartface
06-27-2011, 10:40 PM
Why are the ratings in ball ladder so much higher than the ones in tbd? Just wondering, didn't want to clutter the forums with a new thread

nobodyhome
06-27-2011, 10:46 PM
There are a lot more games played on ball ladder than in tbd, thus points get injected into the system at a much faster rate.