39 replies to this topic

[TruePoindexter]

IceSerpent, on 13 May 2013 - 07:22 AM, said:

Unfortunately, that's not how it seems to work. What happens is that there is a group with Elo around 1345 that's been sitting in the queue for a while (so their acceptable Elo range is huge) and a solo player with, let's say, 2500 Elo clicks launch button, that player is instantly assigned to that group. So, from that player's perspective, MM doesn't even try to find a suitable match for him/her.

*nods* Also it's entirely possible that a 2500 high ELO player is actually on that team already thanks to grouping. Consider a premade 3 man of a 2500 player and two 1000 players - the average ELO would be 1500 (assuming an arithmetic mean) but all of the players in the group are well outside their appropriate ELO. The fact that pre made groups exist throws a healthy wrench into the ELO system.

[Rick Rawlings]

HarlekinEO, on 13 May 2013 - 03:46 AM, said:

Where do you find your current ELO and how do you gain it.

If heard from ELO in my Guild. They told me that it affects matchmaking.

Also, if ELO-Matchmaking exist it would be terrible, because in MWO your results are effected by your Teammates. When I go Solo, I often play a Skirmish Role. Means flanking the enemy and drawing attention onto me. This shall help my team to get a superior number on Battle field. But thats theory. In reallity it happens often, that my Teammates died just before I could charge into battle. The result is, that my player points aint good (ELO drop?). So you can play with more scarried chickens, that frightened to fight. Especially in the current situation, where Sniper/2bl-Hit PPC/Gauss/... Mechs dominate the Battlefield and your ELO drops and drops and drops... and you wont have a chance to play with decent players.

You're on the internet, son. Everyone is in the top Elo bracket and wins 85% of their matches with a 5-6 positive K/D ratio. Oh, and they almost exclusively pug. And everyone should L2P and quit complaining about poptard PPC boats. Even though they'd never run one, personally.

[Levi Porphyrogenitus]

Currenlty Elo is tracked separately for each player in each weight class. When the MM makes a team, it tries to match the overall average Elo of the players on each team, with preference for matching overall similar weight class composition. As the MM fails to find a team, it loosens the weight class matching criteria, and if you get really unlucky you might find that it throws it out altogether.

What this does is preferentially favors the fair to middling players, the ones who are competent but not brilliant. It also favors those who run in coordinated 3-4 man premades. In the former case, if your personal Elo is about average then you'll be well-matched in general. For every enemy player better than you, you can expect another one who is worse than you. In the latter case, your group will climb the Elo ladder and wind up either dropping with a bunch of scrubs and basically 4v8ing against an average enemy team, or you'll get matched with people who are really good and wind up having a a really fun game that ends up being very close.

The second outcome for the skilled premade should be the goal of Elo. You want teams composed of people with roughly similar skill against people of roughly similar skill, since that makes for the best game play. Unfortunately, it's uncertain that you'll get such games since there remains a good chance that your skilled premade will roll right over an average team that isn't on voice comms, carrying the scrubs who dropped your team's average Elo.

The fix is to do Elo tiers. Every 100 Elo gets divided into a MM subgroup, preferentially matching players from that group against each other. If a match is taking too long to launch, then like the weight class restriction the tier restriction could be loosened, but it'd still put you in against somebody +- 100ish Elo from what yours is.

Do well and you advance up the Elo tiers and find yourself consistently matched against players of comparable skill. Do badly and you fall through the Elo tiers until you find yourself being matched against people who are more fun for you to play against. Nobody likes to get rolled, and not many people like to roll everyone without challenge (at least, not very often; the occasional massacre can be fun).

The risk here is that the community might be split. The response is that splitting the player base will make it so that a new player in a Trial mech doesn't find himself in a match with a bunch of CB veterans in custom rides just so he can lower their average Elo. Splitting a community is only really dangerous if it is a permanent, unbridgeable divide (separate and distinct servers with no transfers are a far bigger risk to the community, for instance). If there is mobility among the MM subgroups, then the imperative to get people fun games overrides the risk of having some people never have the chance to play against some others.

In fact, Elo tiers would probably foster community. If you are dropping with some of the same people with frequency then you get to know them, even just a little. This kind of thing can lead to all manner of good things for a gaming community. Solo players might get invites to units, people who drop together frequently might decide to form a premade with voice comms (PGI's TS server is still up, yes?), and people who play with or against each other regularly can form healthy rivalries that can lead to some incredible game play opportunities.

[Phaesphoros]

Hauser, on 13 May 2013 - 09:50 AM, said:

No. Only once the team is assemble the match maker takes the average elo of both teams to figure out how likely they are to win or lose (which is then used to adjust our elo rank after the match).

Average elo also is used to match one team against another - that's the whole point of elo MM: It'll try to match teams such that the probability to win:lose is quite even (50:50).

Quote

While assembling the team the match maker starts with an elo value in mind and pulls players in that are close to it. As explained here: http://mwomercs.com/...43#entry1935843

It's not clear from that quote if the target value is fixed or changes for different groups.

• If it's fixed, then MMer will begin with pulling in moderate players, but I'd guess it'll also pull in players with extraordinary elo if they've already waited very long. The rest of the process isn't described; one could guess "target value" is a target value for the team's average elo, such that for every good player, a bad player is pulled in to keep the balance (i.e. team's average elo close to the "target value"). But as I said, it's not described and only a guess.
  
• If it's not fixed, there could be an effect leading to "elo tiers". Imagine MM starts with a good player and raises the target value to 2000. Then good players will be pulled into that team and it'll be matched against a similar team - that's at least one way to read that quote.

Maybe we could get some more "official" info on this?

Edited by Phaesphoros, 13 May 2013 - 11:20 AM.

[Phaesphoros]

Levi Porphyrogenitus, on 13 May 2013 - 10:58 AM, said:

Currenlty Elo is tracked separately for each player in each weight class. When the MM makes a team, it tries to match the overall average Elo of the players on each team, with preference for matching overall similar weight class composition. As the MM fails to find a team, it loosens the weight class matching criteria, and if you get really unlucky you might find that it throws it out altogether.

i.e. it computes the average elo of team A and (separately) of team B, compares them, and if they're similar, matches them?
That doesn't describe how a team is formed, but how a match is formed from two already existing teams.

Levi Porphyrogenitus, on 13 May 2013 - 10:58 AM, said:

In the latter case, your group will climb the Elo ladder and wind up either dropping with a bunch of scrubs and basically 4v8ing against an average enemy team, or you'll get matched with people who are really good and wind up having a a really fun game that ends up being very close.

? how? get matched with people in your own team or in the opposing team?

Levi Porphyrogenitus, on 13 May 2013 - 10:58 AM, said:

The second outcome for the skilled premade should be the goal of Elo. You want teams composed of people with roughly similar skill against people of roughly similar skill, since that makes for the best game play.

I.e. you think it'd be the most fun if all players in one team had about the same elo (if they're matched against a similar team)? I'd disagree, but that's beside the point of how elo MM forms a team.

Edited by Phaesphoros, 13 May 2013 - 11:19 AM.

[Levi Porphyrogenitus]

Phaesphoros, on 13 May 2013 - 11:18 AM, said:

i.e. it computes the average elo of team A and (separately) of team B, compares them, and if they're similar, matches them?
That doesn't describe how a team is formed, but how a match is formed from two already existing teams.

I don't know the process, but I have two guesses. The first is that it makes a team, then figures out the average Elo for that team and starts to pull in players in an effort to make the second team's average match that of the first team. I suspect this might be how it works on account of often seeing one team loaded instantly and the other with players trickling in. This is not a consistent behavior, though, so that brings me to method two.

My second theory for Elo team creation is that the MM picks a player and targets his Elo. At that point it starts to draw in players for both teams in such a way as to create two teams with matching average Elo values.

Again, I don't actually know how the teams are formed, what the order of operations are for the MM, etc. All I do know is that the end result is supposed to be 2 teams with matching average Elo ratings. If either of my two theories is correct, then this process does determine how a team is formed, not just what teams get matched against each other.

Phaesphoros, on 13 May 2013 - 11:18 AM, said:

? how? get matched with people in your own team or in the opposing team?

Both. The ideal for me would be to have both teams filled with people of similar skill level, since that would promote the most challenging game play for all players on both teams. Currently the only two situations where that seems likely to happen are with skilled premades who get matched with and against high-Elo players and with average players who get matched with and against other average players.

Phaesphoros, on 13 May 2013 - 11:18 AM, said:

I.e. you think it'd be the most fun if all players in one team had about the same elo (if they're matched against a similar team)? I'd disagree, but that's beside the point of how elo MM forms a team.

Yes, I do think that it would be most fun if all players on both teams had similar Elo ratings. This would promote challenging and fun gameplay, and reduce the occurrence of one-sided massacres.

[A banana in the tailpipe]

Kanatta Jing, on 12 May 2013 - 06:58 AM, said:

Well I suppose that could answer some things.





Look at the middle hump here. What can be described as an ELO neutral band. If your in that band then the majority of the people you get teamed up with will also be from that band. Because there are just so many of them. They manage to support the team enough not be a detriment but they don't pull ahead enough to win games.


But if your up or low from the hump you get mirrored as MM needs two slightly better players and 3 slightly lower players to make the math work out anywhere near to zero. Abnormal players attracting abnormal players.


So there is exactly 2 ELO brackets. Normal and Abnormal. in the hump and outside of the hump.

According to the logic here then, for every 3 terrible players there are 2 aces on a team. All that really matters then is if those aces decided to communicate and work together while the scrubs grind eachother down. It doesn't really make the meta any more interesting. All it does is add a random element of chance that your match might be completely screwed from the start if the compotent players decided to do something about it, and all the "hump" players in the world can't stop them.

[Khanahar]

I personally wish we had the option to know where we stood. Even if it was something as vague as the Starcraft League system, where you were ranked according to your position.

Grandmaster: 99.8-100
Master: 96-99.8
Diamond: 88-96
Platinum: 74-88
Gold: 51-74
Silver: 20-51
Bronze 0-20

Gave a sort of overall feel. Bronze means brand new, Silver and Gold is where casuals hang out, and Grandmaster is where pros and semi-pros and smurfs hang out. (Conceptually... GM is not always the actual top tier due to a quirk in how it is assigned.)

[Levi Porphyrogenitus]

Khanahar, on 13 May 2013 - 12:43 PM, said:

I personally wish we had the option to know where we stood. Even if it was something as vague as the Starcraft League system, where you were ranked according to your position.

Grandmaster: 99.8-100
Master: 96-99.8
Diamond: 88-96
Platinum: 74-88
Gold: 51-74
Silver: 20-51
Bronze 0-20

Gave a sort of overall feel. Bronze means brand new, Silver and Gold is where casuals hang out, and Grandmaster is where pros and semi-pros and smurfs hang out. (Conceptually... GM is not always the actual top tier due to a quirk in how it is assigned.)

With Community Warfare, we should get something vaguely similar to this. It won't be Elo-based, but rather Loyalty Points. One assumes, though, that people who do well earn LP faster. LP will likely unlock skins, decals, and ranks, among other things, and the ranks will probably be the closest to a public skill ranking system outside of the various tournament PGI runs that we're going to get.

See somebody at Sergeant or Corporal? Probably low-to-middling LP. Captain? He's got a decent number. Colonel? He's probably got quite a few LP.

Edit: Aside from this, though, I'd love for something like the LoL Ranked Tiers to show up in MWO.

Edited by Levi Porphyrogenitus, 13 May 2013 - 12:49 PM.

[Kanatta Jing]

TruePoindexter, on 13 May 2013 - 10:36 AM, said:

*nods* Also it's entirely possible that a 2500 high ELO player is actually on that team already thanks to grouping. Consider a premade 3 man of a 2500 player and two 1000 players - the average ELO would be 1500 (assuming an arithmetic mean) but all of the players in the group are well outside their appropriate ELO. The fact that pre made groups exist throws a healthy wrench into the ELO system.

You have 5 ELO scores. One for each of the four weight classes and one for when you're standing in a 4 man.

lockwoodx, on 13 May 2013 - 12:30 PM, said:

According to the logic here then, for every 3 terrible players there are 2 aces on a team. All that really matters then is if those aces decided to communicate and work together while the scrubs grind eachother down. It doesn't really make the meta any more interesting. All it does is add a random element of chance that your match might be completely screwed from the start if the compotent players decided to do something about it, and all the "hump" players in the world can't stop them.

The idea is that in a balanced game the team that has more people giving it their all and personal best wins.

[Hauser]

@Phaesphoros,

you're seriously over thinking this. It's not that hard.

You can summarize the whole thing like this

• The match maker picks an Elo value. Say 1300.
  
• It goes through the queue looking for a player near that value.
  
• When found this player1 is put into team A.
  
• The match maker goes through the queue again looking for a player around 1300 elo and with a weight near that of team A.
  
• When found this player2 is put into team B.
  
• If nobody was found relax the target elo and weight values a bit.
  
• Repeat from 2 until teams are full.
  
• Start match putting team A vs team B.

Elo bracketing might occur but only if the brackets are of significant size.

[TruePoindexter]

Kanatta Jing, on 13 May 2013 - 01:41 PM, said:

You have 5 ELO scores. One for each of the four weight classes and one for when you're standing in a 4 man.

Yes and what does that have to do with my example? You don't always play with the same 4 man or run the same mech. It just picks the rating appropriate at the time.

[Phaesphoros]

@Hauser

Hauser, on 13 May 2013 - 03:05 PM, said:

@Phaesphoros,

you're seriously over thinking this. It's not that hard.

You can summarize the whole thing like this
[algorithm]
Elo bracketing might occur but only if the brackets are of significant size.

I appreciate your explanation, thank you. But some questions remain:

• Did the devs explain this somewhere? All I can find is a vague hint to steps 1 and 2.
  
• I think this algorithm is missing an adjustment for extraordinary elos. As with your algorithm, they're only assigned to a team if there's no other player with a mediocre elo currently in the queue.
  
• The question remains whether the initial elo value the MM picks is constant or not. As I've already noted:

Phaesphoros, on 13 May 2013 - 11:07 AM, said:

• If it's fixed, then MMer will begin with pulling in moderate players, but I'd guess it'll also pull in players with extraordinary elo if they've already waited very long. The rest of the process isn't described; one could guess "target value" is a target value for the team's average elo, such that for every good player, a bad player is pulled in to keep the balance (i.e. team's average elo close to the "target value"). But as I said, it's not described and only a guess.
  
• If it's not fixed, there could be an effect leading to "elo tiers". Imagine MM starts with a good player and raises the target value to 2000. Then good players will be pulled into that team and it'll be matched against a similar team - that's at least one way to read that quote.

[Hauser]

1. Step 4 isn't explained anywhere but if you think about the problem for two minutes you'll figure out that constructing both teams at the same time is the only feasible solution.

2. In this instance yes, players that are far from 1300 won't get in this match. That is the whole point.

3. You seem to be confusing a few things here. The match maker starts with a value in mind. While composing a team for this value it does not change (it just doesn't make sense to do so). What can change however is how far the match maker will stray from this value. The average elo of the teams under construction does not come into play untill they have been created to calculate the odds of winning, this chance is used to adjust players elo based on the outcome.


The match maker also creates matches for different elo levels, e.g. it picks different start values (there would be no point to match making otherwise). I don't know how these values are picked. An easy way would be to use the value of the first player in the queue (this avoids having to make predictions about how many players of which elo will be online at any given time).

Edited by Hauser, 13 May 2013 - 03:52 PM.

[Phaesphoros]

Hauser, on 13 May 2013 - 03:50 PM, said:

1. Step 4 isn't explained anywhere but if you think about the problem for two minutes you'll figure out that constructing both teams at the same time is the only feasible solution.

I think there are more solutions which are feasible, e.g. fetch 16 players off the pool (using a target value) and afterwards assign them to the teams such that the team average elos are similar.

Hauser, on 13 May 2013 - 03:50 PM, said:

3. You seem to be confusing a few things here. The match maker starts with a value in mind. While composing a team for this value it does not change (it just doesn't make sense to do so). What can change however is how far the match maker will stray from this value. The average elo of the teams under construction does not come into play untill they have been created to calculate the odds of winning, this chance is used to adjust players elo based on the outcome.

IMO it does make very much sense to adjust the target value according to the average elo of the players you've already assigned to a team; especially if an extraordinary elo players has been assigned. This way you ensure while composing the teams that both teams end up with a similar elo value - which is IMO the whole point of elo team MM.

Hauser, on 13 May 2013 - 03:50 PM, said:

The match maker also creates matches for different elo levels, e.g. it picks different start values (there would be no point to match making otherwise). I don't know how these values are picked. An easy way would be to use the value of the first player in the queue (this avoids having to make predictions about how many players of which elo will be online at any given time).

If there are different target values, then there are "soft tiers", i.e. high elo players would more likely be assigned to matches with other high elo players.

Still, I think there is a point to MM even if the target value is fixed; even if it's not adjusted in the way I mentioned above:

AFAIK the purpose of MM is to match two teams in a way that it's a hard game for both sides (no steamrolling). That is, to match them such that the probability/chances to win are even for both teams.

Elo is used in an algorithm to determine how probable it is you'll win against another player. It looks like PGI have extended this algorithm to compute the probability of a team winning against another team based on the elo values of the individual players. Furthermore it looks like this extension is simply to use the PvP algorithm on the elo average of the teams.

Therefore, it looks to me like MM is trying to match teams with similar average elo. If you just pull in players in the way you described, from a bracket around a target value (or with a certain distribution around this target value) you'll end up with two teams with different elo values. They'll be somewhat similar, but nowhere near as similar as possible. This wouldn't be IMO a good way to achieve the aforementioned purpose.

I imagine you could use a priority queue to order players according to their elo's deviation from 1300, where players that wait long (having an extraordinary elo) will get prioritized eventually. This way, you'll eventually pull in a player with an extraordinary elo even if the target value is fixed. By now using the team average elo to level the two teams, you'll still end up with what your prediction says will be a "good match".

Another idea would be to adjust the target value according to the average elo of players currently in the queue. It wouldn't change much in short-term, but it wouldn't be fixed as well.

Now talk about over-thinking

Edited by Phaesphoros, 13 May 2013 - 04:25 PM.

[Rekoro]

There is only one way an elo system should work

Using an elo tier system like LoL or any popular game that has an functional elo system in place
low vs low
mid vs mid
high vs high
... you get the idea....

exeptions are needed of course, but it should never mix high and low players unless they are premade and even that would suck. This is how i see it.
The mm is in place to create even matches while the elo system is used to match you with ppl of your skill. and if thats right, there is no elo system in place atm because you get mixed around with low high and mid players alike.

i posted about this a while ago, others did aswell. but i dont bother, i quit the game and won´t return untill they actually do something about gamebalance etc. said it before, as soon as they launched the ingamestore the "beta" excuse is gone. So far they take money but dont provide services. They are not as big as ea, but they suck as much as ea does.

Edited by Rekoro, 13 May 2013 - 05:47 PM.

[Nik Reaper]

Rekoro, on 13 May 2013 - 05:45 PM, said:

There is only one way an elo system should work

Using an elo tier system like LoL or any popular game that has an functional elo system in place
low vs low
mid vs mid
high vs high
... you get the idea....

exeptions are needed of course, but it should never mix high and low players unless they are premade and even that would suck. This is how i see it.
The mm is in place to create even matches while the elo system is used to match you with ppl of your skill. and if thats right, there is no elo system in place atm because you get mixed around with low high and mid players alike.

i posted about this a while ago, others did aswell. but i dont bother, i quit the game and won´t return untill they actually do something about gamebalance etc. said it before, as soon as they launched the ingamestore the "beta" excuse is gone. So far they take money but dont provide services. They are not as big as ea, but they suck as much as ea does.

This is optimal , this describes brackets , ether hard or soft but it has a requirement that lol fulfills and MWO does not, and that being a huge mass of available players at every given moment, seriously lol has such a large number of players of every elo at all times it can afford to split players in to brackets , much like starcraft 2 does and still get away with relatively short waiting times , but do not forget , to find a match in lol sometimes at certain elos at certain times of the day it can take up to 15-20 min.

The big question being asked here is how much do we need a hard balance in elo values for us to have fun games?

At this point MW:O is nether competitive nor "balanced" (still beta and all) in many ways, so it goes to reason that until the game has a large player base and one that needs precise team vs team balancing they can't afford long waiting times, as at this point all you can lose is lowering your ELO, k/d and w/l ratio, which at this point is pointless.

I personally think that when CW is introduced and playing for your faction waiting for 15 min to find a balanced match is preferable to being steam rolled and losing territory doe to a need to have a match more quickly, but also think that a more lenient MM can be acceptable in quick matches, much like the ones we have now, up to the point that we have enough players at all times to make it more tight and bracketed.

[Khanahar]

Levi Porphyrogenitus, on 13 May 2013 - 12:48 PM, said:

With Community Warfare, we should get something vaguely similar to this. It won't be Elo-based, but rather Loyalty Points. One assumes, though, that people who do well earn LP faster. LP will likely unlock skins, decals, and ranks, among other things, and the ranks will probably be the closest to a public skill ranking system outside of the various tournament PGI runs that we're going to get.

See somebody at Sergeant or Corporal? Probably low-to-middling LP. Captain? He's got a decent number. Colonel? He's probably got quite a few LP.

Edit: Aside from this, though, I'd love for something like the LoL Ranked Tiers to show up in MWO.

Not quite. You will presumably be able to grind up LP. Better players will earn them faster, as they already do with C-Bills (unless they play cap 'mechs). But they are grindable. Elo is not grindable, except in the sense that you can practice and get better. Starcraft's League system is not grindable. If you have an Elo that is stable in the bottom 20%, you will be in bronze league. If you have a stable one as the top player around, you will be Grandmaster. If your Elo is not stable, you will stay in wherever you were last time it was until you stabilize out. If you get better, you are promoted.

NB: Starcraft actually does have a part of the system to encourage grinding WHICH IS COMPLETELY INDEPENDENT OF ELO OR LEAGUE RANK AND IS MERELY COSMETIC, because your rank within your league (or actually within a group of 99 randomly selected leaguemates) is partly dependent on grinding, up to a point (past 73 points, or 10-40 games per week). But this has no effect on actual Elo or league placement. It is a facade to promote players playing, as they will see their rank going up with games. A clever system, I've always thought.

[HarlekinEO]

Rick Rawlings, on 13 May 2013 - 10:42 AM, said:

You're on the internet, son. Everyone is in the top Elo bracket and wins 85% of their matches with a 5-6 positive K/D ratio. Oh, and they almost exclusively pug. And everyone should L2P and quit complaining about poptard PPC boats. Even though they'd never run one, personally.

I didnt complained about any Spec.
The Problem is: If you have people, which running straight into them and you have to fight 8 vs 4, then the game isnt fun. Its also no fun if you are on opposite side.
I hope you got my point now. How can your Elo determined correctly, if your Stats are 100% defined by your Team. If you dont seek kills (skirmish, you know. Or if you play none Damage Related Roles)

I also play an other MMO. In the siege times you can choose your enemies. We mostly go for better ones, to try us on there Def. Even if your are less and are less good equipped you can create Strategies to beat them. You can learn nothing when you break a Def of Low-Skilled Players. You just run through them nearly unharmed.
Maybe faction wars will bring some balance to this.

Edited by HarlekinEO, 14 May 2013 - 01:07 AM.

[Rekoro]

@Nik

While mwo does not have the numbers to support such a system, I am willing to wait some time to have a better game experience. you can adjust the system like mid vs high is possible to ensure a larger player pool. but as i said, as long as their elo system is matching high vs low the system can be shut down for good because it serves no pupose at this point.

you asked how much hard balance do we need to have fun ?  I would be glad if ppl that instantly die, die without dealing ~100 dmg  or with stupid loadouts (which is an indicator for beeing bad) stay out of my matches (meaning not with or against me) i ld like to duke it out with players at the skill level. i pick that over winning or losing because one team has more ******* in it than the other.
-> http://mwomercs.com/...02#entry2307202

Edited by Rekoro, 14 May 2013 - 03:59 AM.