Three-Sided Deathmatch

moo_2021 · July 3

Moonspawn wrote: »

Hundreds of matches in the span of 7 days? How did you manage that?

Schedule vacation to play BGs

Haki_7 · July 3

@DestroyerPewnack wrote:

In my view, the issue is not 2-sided versus 3-sided that's causing the lopsided problem. It's the fact that MMR is reset every month, forcing the newbies and the sweats to end up in the same lobbies.
Make MMR last for the entire season, (or better yet, permanent per character,) and let people play against players they can handle.

@Weesacs wrote:

Im sure it isn't that difficult to implement an algorithm that makes the game fairer and more competitive.

From what we've seen, there's a good chance the matchmaking has been working as intended this entire time.

Deathmatch 1, no way to win. Green-2 ditched us to try and spawncamp newcomers. First and third critical flaws of 2-sided:

Deathmatch 2, no way to lose:

Domination, no way to lose. Green-4 stop fighting to reminisce about how they miss 3-sided:

Crazy King, no way to lose. Ventured into 4v4 once again to marvel at how it magnifies the problems of 8v8:

Chaosball 1 & 2, no way to lose:

Relic, no way to lose:

Destruction of Battlegrounds Chapter 223: Waiting 41 minutes for a lopsided match (Solo 8v8 PC/NA)

https://youtu.be/_duVdyNJi5E

Moonspawn · July 4

Front page has been updated with the questions only Zenimax can answer.

Now, regarding @imPDA 's explanation:

imPDA wrote: »

If this system works as described, it does 2 things:
1) it tries to make as many people of the same MMR as possible into one game
2) it tries to make teams as equal MMR-wise as possible

So, two results of Elo are:
1) if number of players is big, it can successfully pick 16 players of the same MMR A+-a (like, 1500+-100 for example). It is possible in big MOBA games.
2) if number of players is not that big, it can pick any players to make groups look equal. Balancer can pick 6 "1500 MMR" players, one "200 MMR" player and one "2800 MMR" player, so team will be balanced at 1500. Another team can have all "1500 MMR" players.

You usually see result 2, I suppose, because from my feeling, there are not many players in PvP. If so, it is hard to make two equal groups out of equal players, so it picks some overskilled players and team them with low MMR players, and another team can have the same situation, that is why you usually see 2-4 players with positive KD and some damage, big score, and all other players are constantly dead, 0 damage, 0 score.

It is very rough estimation, and it will not be true in many cases, but you can estimate, that average winrate of you team is near 50%. If your winrate is 80%, winrates of 6 other players are 45 55 45 55 45 55, then winrate of last player should be close to 20% Once again, it is very rough estimation.

Is there any other way for any matchmaking system to function?

Haki_7 · July 5

@moo_2021 wrote:

@Moonspawn wrote:

Hundreds of matches in the span of 7 days? How did you manage that?

Schedule vacation to play BGs

Instantaneous queues into matches that ended quickly because half the players had no intention of ever going after the chaosball.

Destruction of Battlegrounds Chapter 224: Waiting 20 minutes for a lopsided match (Solo 8v8 PC/EU)

https://youtu.be/xpLjjm9kwAM

Chaosball 1 and 2, no possibility of losing:

Crazy King, no possibility of losing. Desperately trying to end the lopsided snoozefest while my teammates spawncamp. Third critical flaw of 2-sided. Orange-6 ditched his team to go around in stealth looking for newcomers:

Relic, no possibility of winning. Team with healer somehow losing to team without healer:

Domination, no possibility of losing:

Deathmatch 1 and 2, no possibility of losing. Sacrificed myself at the beginning to steal their damage sigil. There was no need. The outcome of these matches had already been predetermined:

Moonspawn · July 5

Haki_7 wrote: »

@Demonwolff wrote:

Haki_7 wrote: »

What is the meaning of ''zero damage ball carries'' and ''base jumping''?

Zero Damage ball carries are where the ball bugs out and fails to damage the carrier at all. Base jumping is where the Ball Carrier is able to "magically" return to their spawn. I didn't mention the one where the ball disappears and you cant see the person carrying it lol

In the hundreds of matches I played during the 3-sided chaosball event, I never once saw this happen. Did you?

Maybe it was shadow patched somewhere along the way?

Haki_7 · July 6

Moonspawn wrote: »

Front page has been updated with the questions only Zenimax can answer.

Now, regarding @imPDA 's explanation:

imPDA wrote: »

If this system works as described, it does 2 things:
1) it tries to make as many people of the same MMR as possible into one game
2) it tries to make teams as equal MMR-wise as possible

So, two results of Elo are:
1) if number of players is big, it can successfully pick 16 players of the same MMR A+-a (like, 1500+-100 for example). It is possible in big MOBA games.
2) if number of players is not that big, it can pick any players to make groups look equal. Balancer can pick 6 "1500 MMR" players, one "200 MMR" player and one "2800 MMR" player, so team will be balanced at 1500. Another team can have all "1500 MMR" players.

You usually see result 2, I suppose, because from my feeling, there are not many players in PvP. If so, it is hard to make two equal groups out of equal players, so it picks some overskilled players and team them with low MMR players, and another team can have the same situation, that is why you usually see 2-4 players with positive KD and some damage, big score, and all other players are constantly dead, 0 damage, 0 score.

It is very rough estimation, and it will not be true in many cases, but you can estimate, that average winrate of you team is near 50%. If your winrate is 80%, winrates of 6 other players are 45 55 45 55 45 55, then winrate of last player should be close to 20% Once again, it is very rough estimation.

Is there any other way for any matchmaking system to function?

I don't think so, but someone else might know.

@Solantris is there another way?

Domination, inevitable victory. The scoreboards today are ghastly once again 😢.

Crazy King, inevitable victory. All their points came from backcapping empty flags while the format itself encourages my teammates to spawncamp. Third critical flaw of 2-sided. Ending the match as soon as possible is the only way I know to try and save them from themselves:

Deathmatch 1 and 2, inevitable victories. Throwing in the towel after one death. Wouldn't have happened in 3-sided. Fourth flaw:

Destruction of Battlegrounds Chapter 225: Waiting 23 minutes for a lopsided match (Solo 8v8 PC/EU)

https://youtu.be/tMk23On4qhI

Chaosball 1 and 2, inevitable victories:

Relic, inevitable defeat. Successfully extended the duration of the horrid nightmare to the maximum 15 minutes. Mission accomplished:

Solantris · July 6

I'm with impda option 2 generally. But I have a slightly different perspective on this problem as a game designer myself.

I'll be frank, the algorithm itself has never been the real problem with balancing BGs. There's a lot of player debate on this topic lately, but its a symptom, not the cause. The real issue is that ESO lacks the systems and constraints that usually underpin how an mmr algorithm actually works.

To put it succinctly, our current BG ecosystem does not present a consistent definition of success for an algorithm to measure. As a result, any single MMR algorithm will struggle to produce a reliable estimate of player skill, because a win means different things.

Deathmatch primarily rewards combat dominance. Domination primarily rewards space control and speed. Crazy King primarily rewards rotations and timing. CTR and ball boy primarily reward movement, escorting and survival. In most of them, any kind of prolonged fight is directly detrimental to the objective, and each of them asks players to optimise for something completely different.

Somebody with a high win rate at domination probably does perform equally well in crazy king, but that means absolutely nothing in a deathmatch context. Because in ESO, the design of our objective modes does not fundamentally require combat for effectiveness. Mechanically, we can often bypass combat and still be really effective.

In games like MOBAs, macro decisions (map rotations, objectives, vision) are constrained by combat competency (micro decisions). If you lose every fight, you will generally lose map control. Avoiding fights will put you at an objective disadvantage, but so will fighting uneccesarily: they're designed in harmony, interdependent. But In ESO, our objective mechanics are much kess dependent on combat performance, and the most efficient thing can often be to slap on some running shoes and avoid combat completely. Objective sucess does not follow or require combat sucess. Meaning, from an algorithmic perspective, two players can have identical win rates while possessing incredibly different actual combat ability. The wins are not comparable.

This is especially the case with deathmatch. It meassures something substantially different from the other modes, but has the same impact on a winrate.

Largely, most of us believe our current mmr algorithm has something to do with win rate, given public comments. In most games this is often one of the most reliable indicators, because a win is comparable to other wins. But in ESO, it's just not.

Regardless of implementation, an MMR algorithm fundamentally asks 'on average, which players consistently outperform others?' The problem we have is that BG wins don't represent consistent or comparable performance. Sometimes it reflects combat dominance, sometimes objective play, sometimes rotation or escorting, sometimes building like a brick and holding block on an objective.

When the definition of success changes from match to match, a single rating just cant converge on a single measure of skill. And thats why the mmr system feels so chaotic and unbalanced to players. Mathematically, I imagine it actually is creating a balanced match on paper: it's just that the data it bases that decision on doesnt actually mean anything specific in our context.

I'm not wading too far into the 3 team debate, but I will say from my perspective, its not mechanically any less imbalanced than 2 teams. Nothing really changed under the hood from a balance perspective. I do think it *feels* better to players, and that does have meaning. But thats largely the impact of increased variables and the way that feels to players, not because its fundamentally any more balanced from a mechanical perspective

Tldr: MMR systems estimate skill from outcomes. For that to be meaningful, outcomes need to reflect reasonable performance consistency. Our bg modes reward different things, meaning our wins represent different things, and thus, we have no consistency. Regardless of the actual formula used, this means our mmr systems struggle to mathematically represent meaningful player brackets

moo_2021 · July 7

Solantris wrote: »

Tldr: MMR systems estimate skill from outcomes. For that to be meaningful, outcomes need to reflect reasonable performance consistency. Our bg modes reward different things, meaning our wins represent different things, and thus, we have no consistency. Regardless of the actual formula used, this means our mmr systems struggle to mathematically represent meaningful player brackets

It doesn't have to measure wins. A simple K/A/D based ranking with total healing output for healers should be more than enough to fix the game.

Teams would always be unfair and there would always be members who don't go after objectives. As long as the combat performances of both teams are similar, they can have meaningful and impactful gameplay. It's what matters most.

Ace_SiN · July 7

Readding 3 way PvP is irrelevant to the issue. First, there will never be a system that is perfect and there were plenty of stompings happening in the original BG design. Regardless, the core issue with ESO BGs has ALWAYS been the content itself and maps. The game modes primarily promote "stand your ground" tactics so either your team has a meta healer and high damage aoe builds (with enough personal defense) or your team is collectively on the lesser end of the meta and get smashed. This is why currently having more DKs on your team is a good sign of steamrolling the other.

BGs should promote constant movement and outplay potential without head on blob collisions being the sole way of play.

For example, Capture the Flag is the most balanced Game Mode of any. If done well, there's much less of a damage or defense check to grab the flag and score it. Even if my team is trash, if they can just hold on a bit and be a good enough distraction then we can potentially win. However, one key problem with ESO's capture the relic is how long it takes to grab it so you can sometimes end in a loop where you can't kill the 1 or 2 guarding it or sneak and grab it because there's plenty of time to be countered by the enemy. There shouldnt be such a long channel time. Take a page from New World's CTF. Short channel and disabling of speed boosts so the enemy can still react/defend.

The other game modes are just blob matches that are math checks. There's no built in mechanics that allows a losing team to catch up or outplay to make up the gear/player difference. You either collectively have the gear/skill on your side or you get smoked without a chance.

Lastly, the other issue here is the length of the BGs. They should be sped up. Getting squashed is unavoidable. There is no way to have perfectly balanced pug BGs. However, there should be checks within the system or content design should be in a way that creates a mercy rule (like controlling multiple flags should boost gains after x amount of time).

moo_2021 · July 7

Ace_SiN wrote: »

Readding 3 way PvP is irrelevant to the issue.

two teams make CTF attack and defense too predictable. When it's predictable, a weak and losing team has no chance to revert the situation.

Solantris · July 7

moo_2021 wrote: »

It doesn't have to measure wins. A simple K/A/D based ranking with total healing output for healers should be more than enough to fix the game.

Teams would always be unfair and there would always be members who don't go after objectives. As long as the combat performances of both teams are similar, they can have meaningful and impactful gameplay. It's what matters most.

Fundamentally disagree. This doesn't address the lack of underlying systems and it isnt suited to a game with such diverse playstyles. It would be no more of a consistent measure of sucess in our current environment than win loss currently is.

This definition of meaningful and impactful gameplay is narrow, systemically. Players optimise for what is rewarded and this would would reward ignoring team-mates or group play in favour of single target burst windows to optimise one's own KDA

And whats more, this would be especially punishing to good support players. The better the support, the worse they'd perform, because healing output doesnt equal support quality.

The better a support player is, the more effective they are at enabling pressure, creating opportunities, controlling space, and generally creating favourable conditions. Healing is part of that toolkit, but not all of it. How much self-healing do they enable? How much damage? How much pressure do they alleviate? How much heals, how much sustain? How many buffs? Debuffs? Interrupts? Stuns? Setup? Shields? Peel? Body block? Disruption? None of that is captured. The more they actually provided, the lower their pure healing output would be, and the more theyd be punished.

This system would mathematically reward bird / remembrance spammers over optimised, experienced, active supports.

Its not as easy as just saying kda + heals is a measure of good combat performance. Very few successful competitive games base MMR primarily or exclusively on KDA. In most games its an estimation of the probability a player increases their team's chance of winning a game with a fixed objective, under relatively constrained conditions.

We dont have constrained conditions.

Solantris · July 7

For the record. If there's no ability to change on the horizon, and we had to have exactly what we have with no structural changes...

My vote is for 3 team objectives, 2 team deathmatch, with a separate dm queue and mmr (even if it is just the current jank system)

Moonspawn · July 8

1) Unless someone can come up with an alternative, the best explanation we have for how BG matchmaking works is that it uses the average MMR of the entire team.

First it tries to make as many people of the same MMR as possible into one game
When it fails, it tries to make teams as equal MMR-wise as possible

So, two results of Elo are:
1) if number of players is big, it can successfully pick 16 players of the same MMR A+-a (like, 1500+-100 for example). It is possible in big MOBA games.
2) if number of players is not that big, it can pick any players to make groups look equal. Balancer can pick 6 "1500 MMR" players, one "200 MMR" player and one "2800 MMR" player, so team will be balanced at 1500. Another team can have all "1500 MMR" players.

You usually see result 2, I suppose, because from my feeling, there are not many players in PvP. If so, it is hard to make two equal groups out of equal players, so it picks some overskilled players and team them with low MMR players, and another team can have the same situation, that is why you usually see 2-4 players with positive KD and some damage, big score, and all other players are constantly dead, 0 damage, 0 score.

It is very rough estimation, and it will not be true in many cases, but you can estimate, that average winrate of you team is near 50%. If your winrate is 80%, winrates of 6 other players are 45 55 45 55 45 55, then winrate of last player should be close to 20%

Once again, it is very rough estimation.

2) According to ZOS, the MMR is based on wins and losses.

Source:
www.pcgamesn.com/the-elder-scrolls-online/mmr-battlegrounds-interview

ESO dev lifts lid on the MMO's MMR system saying "it's nothing new"
After huge battleground changes recently hit the Elder Scrolls Online recently, players have become obsessed with MMR and how it all works.

Updated: Nov 1, 2024
The Elder Scrolls Online
PvP has received a renewed focus in the Elder Scrolls Online thanks to the most recent update which completely overhauled the game's battlegrounds. Instead of the traditional version found previously in the Tamrielic game which consisted of three teams facing off against each other, things have become simplified and streamlined. Now, it's just your team versus another one, and as a result of this new playing field fans are looking at the MMO's matchmaking rating system and wondering what effects it has and how to navigate it.
We recently got the chance to speak to Elder Scrolls Online PvP designer Brian Wheeler about all the big changes hitting the MMORPG right now. One query that came up relates to the community's current obsession with MMR, especially looking at how one goes up or down in rank against other players, and where the current system came from.
"What's funny is it's nothing new. We've had that. We've had that MMR system in since update 20," Wheeler laughs. "When we launched battlegrounds with Morrowind, we had a pretty basic MMR that was based on your win and loss, and that's it. What we explored after that was looking at Elo and TrueSkill and various things like that, because what it's supposed to do is take into account your expected win rate or not, which is basically the MMR of your enemies."
"So let's say we're in a group and the system gets the MMR average of our group and then it goes, what's the group? What's the enemy's MMR, what's their ranking and their average?" he continues. "If theirs is higher than yours and you lose, then you don't take as big of a hit. But if you win, you get a bigger gain. The reverse goes for the other team. If they were fighting us and we're lower and we win, they're going to take a bigger hit because they lost to a crappier team."
"The only calculation change we had to make was whether it's looking at two other teams or just one other team in terms of are you going to win, or are you not going to win," he says. "The calculations have always been based on win/loss, and it's been very interesting seeing how much more people are getting interested in that because as a PvP and gamer nerd, I like looking at that stuff."
That said, there's a limit to how much the developer will share - as it might become open to abuse if the full MMR calculations ever became known. "We do have some help definitions that explain a little bit how it functions but as with many things in the game, we're not going to tell you the exact calculations," he lets us know. "That's the secret in the sauce that we have across the board for the entire game."
If you'd like to check out what's new in this latest update which completely changed how battlegrounds work in ESO, head over to the full patch notes to get the lowdown.

Why, then, does the matchmaking seem so random?

Artisian0001 · July 8

Moonspawn wrote: »

1) Unless someone can come up with an alternative, the best explanation we have for how BG matchmaking works is that it uses the average MMR of the entire team.
First it tries to make as many people of the same MMR as possible into one game
When it fails, it tries to make teams as equal MMR-wise as possible

So, two results of Elo are:
1) if number of players is big, it can successfully pick 16 players of the same MMR A+-a (like, 1500+-100 for example). It is possible in big MOBA games.
2) if number of players is not that big, it can pick any players to make groups look equal. Balancer can pick 6 "1500 MMR" players, one "200 MMR" player and one "2800 MMR" player, so team will be balanced at 1500. Another team can have all "1500 MMR" players.

You usually see result 2, I suppose, because from my feeling, there are not many players in PvP. If so, it is hard to make two equal groups out of equal players, so it picks some overskilled players and team them with low MMR players, and another team can have the same situation, that is why you usually see 2-4 players with positive KD and some damage, big score, and all other players are constantly dead, 0 damage, 0 score.

It is very rough estimation, and it will not be true in many cases, but you can estimate, that average winrate of you team is near 50%. If your winrate is 80%, winrates of 6 other players are 45 55 45 55 45 55, then winrate of last player should be close to 20% Once again, it is very rough estimation.

2) According to ZOS, the MMR is based on wins and losses.
Source:
www.pcgamesn.com/the-elder-scrolls-online/mmr-battlegrounds-interview

ESO dev lifts lid on the MMO's MMR system saying "it's nothing new"
After huge battleground changes recently hit the Elder Scrolls Online recently, players have become obsessed with MMR and how it all works.

Updated: Nov 1, 2024
The Elder Scrolls Online
PvP has received a renewed focus in the Elder Scrolls Online thanks to the most recent update which completely overhauled the game's battlegrounds. Instead of the traditional version found previously in the Tamrielic game which consisted of three teams facing off against each other, things have become simplified and streamlined. Now, it's just your team versus another one, and as a result of this new playing field fans are looking at the MMO's matchmaking rating system and wondering what effects it has and how to navigate it.
We recently got the chance to speak to Elder Scrolls Online PvP designer Brian Wheeler about all the big changes hitting the MMORPG right now. One query that came up relates to the community's current obsession with MMR, especially looking at how one goes up or down in rank against other players, and where the current system came from.
"What's funny is it's nothing new. We've had that. We've had that MMR system in since update 20," Wheeler laughs. "When we launched battlegrounds with Morrowind, we had a pretty basic MMR that was based on your win and loss, and that's it. What we explored after that was looking at Elo and TrueSkill and various things like that, because what it's supposed to do is take into account your expected win rate or not, which is basically the MMR of your enemies."
"So let's say we're in a group and the system gets the MMR average of our group and then it goes, what's the group? What's the enemy's MMR, what's their ranking and their average?" he continues. "If theirs is higher than yours and you lose, then you don't take as big of a hit. But if you win, you get a bigger gain. The reverse goes for the other team. If they were fighting us and we're lower and we win, they're going to take a bigger hit because they lost to a crappier team."
"The only calculation change we had to make was whether it's looking at two other teams or just one other team in terms of are you going to win, or are you not going to win," he says. "The calculations have always been based on win/loss, and it's been very interesting seeing how much more people are getting interested in that because as a PvP and gamer nerd, I like looking at that stuff."
That said, there's a limit to how much the developer will share - as it might become open to abuse if the full MMR calculations ever became known. "We do have some help definitions that explain a little bit how it functions but as with many things in the game, we're not going to tell you the exact calculations," he lets us know. "That's the secret in the sauce that we have across the board for the entire game."
If you'd like to check out what's new in this latest update which completely changed how battlegrounds work in ESO, head over to the full patch notes to get the lowdown.

Why, then, does the matchmaking seem so random?

There is a ZERO percent change. ZERO. That MMR is based on win/loss. How do I know? I specifically went in game with friends, not on a new character, on a BRAND NEW ACCOUNT, with the express intent to lose every match. After around 30-40 matches, even though we had 1 win in the entire history of the 40 or so games we played, we were matched up against the same players we always saw in high MMR. I don't know what MMR they use, but they aren't truthful about it. I don't know why, I don't know if the range they use is just way too high, but it is not based on W/L at all. We should have been in the depths of the lowest MMR, not the highest.

Solantris · July 8

Moonspawn wrote: »

1)

Why, then, does the matchmaking seem so random?

Because whatever matchmaking systems that can exist are fed randomised datasets. Randomised roles, randomised players, randomised game modes. There's no constraints, meaning, no constants, meaning, no ability to measure anything accurately

Haki_7 · July 9

Moonspawn wrote: »

1) Unless someone can come up with an alternative, the best explanation we have for how BG matchmaking works is that it uses the average MMR of the entire team.
First it tries to make as many people of the same MMR as possible into one game
When it fails, it tries to make teams as equal MMR-wise as possible

So, two results of Elo are:
1) if number of players is big, it can successfully pick 16 players of the same MMR A+-a (like, 1500+-100 for example). It is possible in big MOBA games.
2) if number of players is not that big, it can pick any players to make groups look equal. Balancer can pick 6 "1500 MMR" players, one "200 MMR" player and one "2800 MMR" player, so team will be balanced at 1500. Another team can have all "1500 MMR" players.

You usually see result 2, I suppose, because from my feeling, there are not many players in PvP. If so, it is hard to make two equal groups out of equal players, so it picks some overskilled players and team them with low MMR players, and another team can have the same situation, that is why you usually see 2-4 players with positive KD and some damage, big score, and all other players are constantly dead, 0 damage, 0 score.

It is very rough estimation, and it will not be true in many cases, but you can estimate, that average winrate of you team is near 50%. If your winrate is 80%, winrates of 6 other players are 45 55 45 55 45 55, then winrate of last player should be close to 20% Once again, it is very rough estimation.

2) According to ZOS, the MMR is based on wins and losses.
Source:
www.pcgamesn.com/the-elder-scrolls-online/mmr-battlegrounds-interview

ESO dev lifts lid on the MMO's MMR system saying "it's nothing new"
After huge battleground changes recently hit the Elder Scrolls Online recently, players have become obsessed with MMR and how it all works.

Updated: Nov 1, 2024
The Elder Scrolls Online
PvP has received a renewed focus in the Elder Scrolls Online thanks to the most recent update which completely overhauled the game's battlegrounds. Instead of the traditional version found previously in the Tamrielic game which consisted of three teams facing off against each other, things have become simplified and streamlined. Now, it's just your team versus another one, and as a result of this new playing field fans are looking at the MMO's matchmaking rating system and wondering what effects it has and how to navigate it.
We recently got the chance to speak to Elder Scrolls Online PvP designer Brian Wheeler about all the big changes hitting the MMORPG right now. One query that came up relates to the community's current obsession with MMR, especially looking at how one goes up or down in rank against other players, and where the current system came from.
"What's funny is it's nothing new. We've had that. We've had that MMR system in since update 20," Wheeler laughs. "When we launched battlegrounds with Morrowind, we had a pretty basic MMR that was based on your win and loss, and that's it. What we explored after that was looking at Elo and TrueSkill and various things like that, because what it's supposed to do is take into account your expected win rate or not, which is basically the MMR of your enemies."
"So let's say we're in a group and the system gets the MMR average of our group and then it goes, what's the group? What's the enemy's MMR, what's their ranking and their average?" he continues. "If theirs is higher than yours and you lose, then you don't take as big of a hit. But if you win, you get a bigger gain. The reverse goes for the other team. If they were fighting us and we're lower and we win, they're going to take a bigger hit because they lost to a crappier team."
"The only calculation change we had to make was whether it's looking at two other teams or just one other team in terms of are you going to win, or are you not going to win," he says. "The calculations have always been based on win/loss, and it's been very interesting seeing how much more people are getting interested in that because as a PvP and gamer nerd, I like looking at that stuff."
That said, there's a limit to how much the developer will share - as it might become open to abuse if the full MMR calculations ever became known. "We do have some help definitions that explain a little bit how it functions but as with many things in the game, we're not going to tell you the exact calculations," he lets us know. "That's the secret in the sauce that we have across the board for the entire game."
If you'd like to check out what's new in this latest update which completely changed how battlegrounds work in ESO, head over to the full patch notes to get the lowdown.

Why, then, does the matchmaking seem so random?

I think a better question would be: How can a matchmaking system with these two characteristics be made to appear completely random?

Moonspawn · July 9

Haki_7 wrote: »

Moonspawn wrote: »

1) Unless someone can come up with an alternative, the best explanation we have for how BG matchmaking works is that it uses the average MMR of the entire team.
First it tries to make as many people of the same MMR as possible into one game
When it fails, it tries to make teams as equal MMR-wise as possible

So, two results of Elo are:
1) if number of players is big, it can successfully pick 16 players of the same MMR A+-a (like, 1500+-100 for example). It is possible in big MOBA games.
2) if number of players is not that big, it can pick any players to make groups look equal. Balancer can pick 6 "1500 MMR" players, one "200 MMR" player and one "2800 MMR" player, so team will be balanced at 1500. Another team can have all "1500 MMR" players.

You usually see result 2, I suppose, because from my feeling, there are not many players in PvP. If so, it is hard to make two equal groups out of equal players, so it picks some overskilled players and team them with low MMR players, and another team can have the same situation, that is why you usually see 2-4 players with positive KD and some damage, big score, and all other players are constantly dead, 0 damage, 0 score.

It is very rough estimation, and it will not be true in many cases, but you can estimate, that average winrate of you team is near 50%. If your winrate is 80%, winrates of 6 other players are 45 55 45 55 45 55, then winrate of last player should be close to 20% Once again, it is very rough estimation.

2) According to ZOS, the MMR is based on wins and losses.
Source:
www.pcgamesn.com/the-elder-scrolls-online/mmr-battlegrounds-interview

ESO dev lifts lid on the MMO's MMR system saying "it's nothing new"
After huge battleground changes recently hit the Elder Scrolls Online recently, players have become obsessed with MMR and how it all works.

Updated: Nov 1, 2024
The Elder Scrolls Online
PvP has received a renewed focus in the Elder Scrolls Online thanks to the most recent update which completely overhauled the game's battlegrounds. Instead of the traditional version found previously in the Tamrielic game which consisted of three teams facing off against each other, things have become simplified and streamlined. Now, it's just your team versus another one, and as a result of this new playing field fans are looking at the MMO's matchmaking rating system and wondering what effects it has and how to navigate it.
We recently got the chance to speak to Elder Scrolls Online PvP designer Brian Wheeler about all the big changes hitting the MMORPG right now. One query that came up relates to the community's current obsession with MMR, especially looking at how one goes up or down in rank against other players, and where the current system came from.
"What's funny is it's nothing new. We've had that. We've had that MMR system in since update 20," Wheeler laughs. "When we launched battlegrounds with Morrowind, we had a pretty basic MMR that was based on your win and loss, and that's it. What we explored after that was looking at Elo and TrueSkill and various things like that, because what it's supposed to do is take into account your expected win rate or not, which is basically the MMR of your enemies."
"So let's say we're in a group and the system gets the MMR average of our group and then it goes, what's the group? What's the enemy's MMR, what's their ranking and their average?" he continues. "If theirs is higher than yours and you lose, then you don't take as big of a hit. But if you win, you get a bigger gain. The reverse goes for the other team. If they were fighting us and we're lower and we win, they're going to take a bigger hit because they lost to a crappier team."
"The only calculation change we had to make was whether it's looking at two other teams or just one other team in terms of are you going to win, or are you not going to win," he says. "The calculations have always been based on win/loss, and it's been very interesting seeing how much more people are getting interested in that because as a PvP and gamer nerd, I like looking at that stuff."
That said, there's a limit to how much the developer will share - as it might become open to abuse if the full MMR calculations ever became known. "We do have some help definitions that explain a little bit how it functions but as with many things in the game, we're not going to tell you the exact calculations," he lets us know. "That's the secret in the sauce that we have across the board for the entire game."
If you'd like to check out what's new in this latest update which completely changed how battlegrounds work in ESO, head over to the full patch notes to get the lowdown.

Why, then, does the matchmaking seem so random?

I think a better question would be: How can a matchmaking system with these two characteristics be made to appear completely random?

I don't understand the question.

Haki_7 · July 10

Moonspawn wrote: »

I don't understand the question.

I'll work on rephrasing.

Solantris wrote: »

For the record. If there's no ability to change on the horizon, and we had to have exactly what we have with no structural changes...

Indeed. Moving forward we'll have to consider how the layoffs will affect Battlegrounds 😢

Crazy King, 100% guaranteed to lose. We could barely reach the flags:

Deathmatch 1 & 2, 100% guaranteed to win:

Chaosball 1 & 2, 100% guaranteed to win:

Relic, 100% guaranteed to win. Promptly giving up after losing one relic, four more to go. Fourth critical flaw of 2-sided:

Domination, 100% guaranteed to win. Running from one flag to the next trying to end the nightmare while my teammates spawncamp. Third flaw:

Destruction of Battlegrounds Chapter 226: Waiting 17 minutes for a lopsided match (Solo 8v8 PC/EU)

https://youtu.be/9MbGM3TirPo

Moonspawn · July 11

Haki_7 wrote: »

Solantris wrote: »

For the record. If there's no ability to change on the horizon, and we had to have exactly what we have with no structural changes...

Indeed. Moving forward we'll have to consider how the layoffs will affect Battlegrounds 😢

Pumping out non-repeatable content ad infinitum may no longer be viable.

Haki_7 · July 11

Let us assume the matchmaking system is bound by two main rules:

First) It uses the average MMR of the entire team.

Unless someone can come up with an alternative, the best explanation we have for how BG matchmaking works is that it uses the average MMR of the entire team. First it tries to make as many people of the same MMR as possible into one game. When it fails, it tries to make teams as equal MMR-wise as possible.

So, two results of Elo are:
1) if number of players is big, it can successfully pick 16 players of the same MMR A+-a (like, 1500+-100 for example). It is possible in big MOBA games.
2) if number of players is not that big, it can pick any players to make groups look equal. Balancer can pick 6 "1500 MMR" players, one "200 MMR" player and one "2800 MMR" player, so team will be balanced at 1500. Another team can have all "1500 MMR" players.

You usually see result 2, I suppose, because from my feeling, there are not many players in PvP. If so, it is hard to make two equal groups out of equal players, so it picks some overskilled players and team them with low MMR players, and another team can have the same situation, that is why you usually see 2-4 players with positive KD and some damage, big score, and all other players are constantly dead, 0 damage, 0 score.

It is very rough estimation, and it will not be true in many cases, but you can estimate, that average winrate of you team is near 50%. If your winrate is 80%, winrates of 6 other players are 45 55 45 55 45 55, then winrate of last player should be close to 20%

Once again, it is very rough estimation.

Second) The MMR is based on wins and losses.

According to ZOS, the MMR is based on wins and losses.
Source:
www.pcgamesn.com/the-elder-scrolls-online/mmr-battlegrounds-interview

ESO dev lifts lid on the MMO's MMR system saying "it's nothing new"
After huge battleground changes recently hit the Elder Scrolls Online recently, players have become obsessed with MMR and how it all works.

Updated: Nov 1, 2024
The Elder Scrolls Online
PvP has received a renewed focus in the Elder Scrolls Online thanks to the most recent update which completely overhauled the game's battlegrounds. Instead of the traditional version found previously in the Tamrielic game which consisted of three teams facing off against each other, things have become simplified and streamlined. Now, it's just your team versus another one, and as a result of this new playing field fans are looking at the MMO's matchmaking rating system and wondering what effects it has and how to navigate it.
We recently got the chance to speak to Elder Scrolls Online PvP designer Brian Wheeler about all the big changes hitting the MMORPG right now. One query that came up relates to the community's current obsession with MMR, especially looking at how one goes up or down in rank against other players, and where the current system came from.
"What's funny is it's nothing new. We've had that. We've had that MMR system in since update 20," Wheeler laughs. "When we launched battlegrounds with Morrowind, we had a pretty basic MMR that was based on your win and loss, and that's it. What we explored after that was looking at Elo and TrueSkill and various things like that, because what it's supposed to do is take into account your expected win rate or not, which is basically the MMR of your enemies."
"So let's say we're in a group and the system gets the MMR average of our group and then it goes, what's the group? What's the enemy's MMR, what's their ranking and their average?" he continues. "If theirs is higher than yours and you lose, then you don't take as big of a hit. But if you win, you get a bigger gain. The reverse goes for the other team. If they were fighting us and we're lower and we win, they're going to take a bigger hit because they lost to a crappier team."
"The only calculation change we had to make was whether it's looking at two other teams or just one other team in terms of are you going to win, or are you not going to win," he says. "The calculations have always been based on win/loss, and it's been very interesting seeing how much more people are getting interested in that because as a PvP and gamer nerd, I like looking at that stuff."
That said, there's a limit to how much the developer will share - as it might become open to abuse if the full MMR calculations ever became known. "We do have some help definitions that explain a little bit how it functions but as with many things in the game, we're not going to tell you the exact calculations," he lets us know. "That's the secret in the sauce that we have across the board for the entire game."
If you'd like to check out what's new in this latest update which completely changed how battlegrounds work in ESO, head over to the full patch notes to get the lowdown.

If the first two rules are still in effect, what could possibly be going wrong? What additional rule could be applied to this system to make it appear completely random?

Solantris · July 11

Haki_7 wrote: »

If the first two rules are still in effect, what could possibly be going wrong? What additional rule could be applied to this system to make it appear completely random?

No rule needs to be applied to this system to make it appear random. It appears random because winrate itself is impacted too much by randomness

Any MMR system based on winrate relies on the assumption that a win in one match is directly comparable to a win in another. Games achieve this by controlling objectives, roles and implementing strong player constraints, leaving skill as the dominant influence on win condition. It's the thing that's different game to game, therefore, you can measure it.

But in ESO, lots of things are different game to game. Lots of things influence win conditions.

Teams are highly inconsistent

Unlike many competitive games, ESO does not have fixed roles or controlled team compositions
A single support player can completely change the outcome of a match
Two teams with the same average MMR can have very different levels of synergy, composition, and experience
The objective-to-deathmatch player split is a Venn diagram, not a circle. Players often optimize for their definition of their success.

Individual players can have massively disproportionate impacts on outcomes

Fluctuating game balance means individual players can blowout matches (See: pulse ganking).
ESO has a high amount of skill variance in one match compared to most competitive games (See: 10year veterans matched with first time BG goers)
Low-interaction objective mechanics often allow individual players to create enormous value without necessarily demonstrating combat skill or even requiring impact from their team (See: guy with running shoes).
Build matchups and class imbalance are a thing

Different game modes measure different skills

Deathmatch primarily rewards combat dominance and good micro decisionmaking
Objective modes primarily reward macro decisionmaking. Different modes have different balance, but still
A player can have a high win rate in objective modes entirely through avoidance, which doesn't translate into combat performance in Deathmatch. A win in one environment is not directly comparable to a win in another.
Randomness of mode select means a player can reach the same rating through completely different skill expression depending on RNG

Population and sample size skew matchmaking

MMR systems are more accurate the more data they have. They require a large number of games to converge toward a reliable estimate.
Resets reduce the amount of information the system has available, but are sadly necessary due to the tiny population.
The system must pull from smaller pools of players with wider experience differences to keep queues functioning at all

This ecosystem means teams could be perfectly balanced on paper using these two rules, while still feeling completely random to players. The wins that created a rating were achieved through different conditions, behaviours, teams, roles, builds and skill sets.

If the conditions that determine a win are largely undefined, then the outcome is effectively random. Winrate doesn't currently reflect a level of true pattern or consistency... it mostly reflects how lucky somebody has gotten. Thus, the system can feel entirely random, while not being.

edit to add, I'm not saying player skill doesn't matter to winrate at all, or that it's totally random. I'm just saying that it has too many uncontrolled variables to isolate individual skill well.

Moonspawn · July 11

Haki_7 wrote: »

Let us assume the matchmaking system is bound by two main rules:

First) It uses the average MMR of the entire team.
Unless someone can come up with an alternative, the best explanation we have for how BG matchmaking works is that it uses the average MMR of the entire team. First it tries to make as many people of the same MMR as possible into one game. When it fails, it tries to make teams as equal MMR-wise as possible.

So, two results of Elo are:
1) if number of players is big, it can successfully pick 16 players of the same MMR A+-a (like, 1500+-100 for example). It is possible in big MOBA games.
2) if number of players is not that big, it can pick any players to make groups look equal. Balancer can pick 6 "1500 MMR" players, one "200 MMR" player and one "2800 MMR" player, so team will be balanced at 1500. Another team can have all "1500 MMR" players.

You usually see result 2, I suppose, because from my feeling, there are not many players in PvP. If so, it is hard to make two equal groups out of equal players, so it picks some overskilled players and team them with low MMR players, and another team can have the same situation, that is why you usually see 2-4 players with positive KD and some damage, big score, and all other players are constantly dead, 0 damage, 0 score.

It is very rough estimation, and it will not be true in many cases, but you can estimate, that average winrate of you team is near 50%. If your winrate is 80%, winrates of 6 other players are 45 55 45 55 45 55, then winrate of last player should be close to 20% Once again, it is very rough estimation.

Second) The MMR is based on wins and losses.
According to ZOS, the MMR is based on wins and losses.
Source:
www.pcgamesn.com/the-elder-scrolls-online/mmr-battlegrounds-interview

ESO dev lifts lid on the MMO's MMR system saying "it's nothing new"
After huge battleground changes recently hit the Elder Scrolls Online recently, players have become obsessed with MMR and how it all works.

Updated: Nov 1, 2024
The Elder Scrolls Online
PvP has received a renewed focus in the Elder Scrolls Online thanks to the most recent update which completely overhauled the game's battlegrounds. Instead of the traditional version found previously in the Tamrielic game which consisted of three teams facing off against each other, things have become simplified and streamlined. Now, it's just your team versus another one, and as a result of this new playing field fans are looking at the MMO's matchmaking rating system and wondering what effects it has and how to navigate it.
We recently got the chance to speak to Elder Scrolls Online PvP designer Brian Wheeler about all the big changes hitting the MMORPG right now. One query that came up relates to the community's current obsession with MMR, especially looking at how one goes up or down in rank against other players, and where the current system came from.
"What's funny is it's nothing new. We've had that. We've had that MMR system in since update 20," Wheeler laughs. "When we launched battlegrounds with Morrowind, we had a pretty basic MMR that was based on your win and loss, and that's it. What we explored after that was looking at Elo and TrueSkill and various things like that, because what it's supposed to do is take into account your expected win rate or not, which is basically the MMR of your enemies."
"So let's say we're in a group and the system gets the MMR average of our group and then it goes, what's the group? What's the enemy's MMR, what's their ranking and their average?" he continues. "If theirs is higher than yours and you lose, then you don't take as big of a hit. But if you win, you get a bigger gain. The reverse goes for the other team. If they were fighting us and we're lower and we win, they're going to take a bigger hit because they lost to a crappier team."
"The only calculation change we had to make was whether it's looking at two other teams or just one other team in terms of are you going to win, or are you not going to win," he says. "The calculations have always been based on win/loss, and it's been very interesting seeing how much more people are getting interested in that because as a PvP and gamer nerd, I like looking at that stuff."
That said, there's a limit to how much the developer will share - as it might become open to abuse if the full MMR calculations ever became known. "We do have some help definitions that explain a little bit how it functions but as with many things in the game, we're not going to tell you the exact calculations," he lets us know. "That's the secret in the sauce that we have across the board for the entire game."
If you'd like to check out what's new in this latest update which completely changed how battlegrounds work in ESO, head over to the full patch notes to get the lowdown.

If the first two rules are still in effect, what could possibly be going wrong?

First thing that comes to mind is that it may be expanding too much, as described by @Stamicka :

Stamicka wrote: »

A lot of MMR systems start narrow to match players as close to skill levels as possible. Then if no matching players are found it’ll expand and expand until it’s basically covering all people queued just to get them in a match.

NxJoeyD · July 11

Solantris wrote: »

[*] Build matchups and class imbalance are a thing

All of that but this more than anything.

The entirety of core combat mechanics are below poor. The simply truth of it is that, as far as the PvP portion of this game goes, the Devs didn't do a good job; they didn't deliver a quality product that players feel broadly compelled to engage with; hence the population decline.

I've said this same thing so many times in other threads. Sure, there are some "band aid" ideas that Bethesda could apply to try and reduce "some" of the pain points of BGs .. but .. the reality is that so long as the Devs continue to ignore feedback and simply deploy updates that break combat mechanical balance in PvP then no amount of MMR is going to "solve" the problem; it's, at best, only going to reduce it slightly .. and even then.

There is far too much randomness in PvP combat and that was only made worse with Subclassing. They have to fix the logic behind core combat otherwise we will always have threads like this.

BG outcomes aren't as much about skill checks anymore as they are down to the randomness of which skill and combat mechanics are going to present in a particular match. You can't MMR that.

Haki_7 · July 12

Moonspawn wrote: »

First thing that comes to mind is that it may be expanding too much, as described by @Stamicka :

Stamicka wrote: »

A lot of MMR systems start narrow to match players as close to skill levels as possible. Then if no matching players are found it’ll expand and expand until it’s basically covering all people queued just to get them in a match.

But the alternative would be never expanding enough to create matches.

Solantris · July 12

Haki_7 wrote: »

But the alternative would be never expanding enough to create matches.

I think you'll find that's why they haven't dipped their toe into it. It's a complex problem

Haki_7 · July 13

@Solantris wrote:

Haki_7 wrote: »

If the first two rules are still in effect, what could possibly be going wrong? What additional rule could be applied to this system to make it appear completely random?

Different game modes measure different skills
Deathmatch primarily rewards combat dominance and good micro decisionmaking

Objective modes primarily reward macro decisionmaking. Different modes have different balance, but still

A player can have a high win rate in objective modes entirely through avoidance, which doesn't translate into combat performance in Deathmatch. A win in one environment is not directly comparable to a win in another.

Randomness of mode select means a player can reach the same rating through completely different skill expression depending on RNG

There is one rule that may account for what you're talking about. It makes sense in theory, and may already be in effect:

Third) Every character has five different matchmaking ratings, one for each of the game modes.

Solantris · July 13

Haki_7 wrote: »

Third) Every character has five different matchmaking ratings, one for each of the game modes.

We can test this through battleboard. Filter by game mode and check your average queue times.

(I borrowed a friends data since I dont have enough recorded matches)

8v8 games
DM, 16 games, avg queue 2:31, 75%
CTR, 8 games, avg queue 4:22, 88%
Chaos, 30 games, avg queue 6:44, 70%
Crazy, 14 games, avg queue 4:12, 79%
Dom, 7 games, avg queue 2:27, 43%

4v4 games
DM, 28 games, avg queue 4:43, 79%
Crazy, 17 games, avg queue 4:18, 71%
Dom, 27 games, 2:27, 89%

If game modes did have their own mmr based on win rate, you'd expect to see modes with a higher winrate consistently having a longer average queue time. From what I can see, that isnt the case.

Moonspawn · July 13

Haki_7 wrote: »

@Solantris wrote:

Haki_7 wrote: »

If the first two rules are still in effect, what could possibly be going wrong? What additional rule could be applied to this system to make it appear completely random?

Different game modes measure different skills
Deathmatch primarily rewards combat dominance and good micro decisionmaking

Objective modes primarily reward macro decisionmaking. Different modes have different balance, but still

A player can have a high win rate in objective modes entirely through avoidance, which doesn't translate into combat performance in Deathmatch. A win in one environment is not directly comparable to a win in another.

Randomness of mode select means a player can reach the same rating through completely different skill expression depending on RNG

There is one rule that may account for what you're talking about. It makes sense in theory, and may already be in effect:

Third) Every character has five different matchmaking ratings, one for each of the game modes.

It does make sense in theory. In practice it would only render the matchmaking even more erratic, because the vast majority of high (damage/KDR) players rarely compete for the objectives, so they could be anywhere on the MMR ladder.

Haki_7 · July 14

In a magical world where the players who dislike the objective modes have their own Deathmatch queue, would a separate MMR for each of the five game modes help to make them all competitive?

Destruction of Battlegrounds Chapter 227: Waiting 25 minutes for a lopsided match (Solo 8v8 PC/EU)

https://youtu.be/24pSqvqZlXg

Crazy King, unavoidable victory. Running around trying to end the lopsided snoozefest while the third critical flaw of 2-sided drives people away from Battlegrounds forever:

Domination, unavoidable victory. Mercifully quick, thank the divines:

Relic, unavoidable defeat. Spawncamping from start to finish. They even held the last relic for a few minutes to farm us some more:

Deathmatch 1 and 2, unavoidable victories:

Chaosball 1 and 2, unavoidable victories:

NxJoeyD · July 14

Haki_7 wrote: »

In a magical world where the players who dislike the objective modes have their own Deathmatch queue, would a separate MMR for each of the five game modes help to make them all competitive?

The short answer is no, not right now.

With the extreme degree of imbalance we have given that we're in the middle of a major class refresh project there is ZERO way that any MMR or any "matchmaking-wannabe-like-system" could ever deliver "competitively" made teams.

Unfortunately, there's no way to do that, unless, maybe, one were to consider grouping Subclassed players, Werewolf players, & DK players all as one "pool" and then separately grouping all other players into a different pool to matchmake.

There is zero chance of that because that would be a huge demand on the server and a long time to design and implement; and also because the population, at least on PC, is too small to ever have that work. I play on both PC NA and XB NA and I could wait forever to get a BG queue to proc on PC but, on average, I only wait a couple of minutes on XB for a BG queue; and that's consistently. I'm never sitting and waiting for a BG on XB.

Now, this tells us that more players on console play ESO than do on PC. Why that is a trend, that's anyone's guess and I think we all have our theories.

But as for any sort of "balancing" MMR, no, put that on the back burner until at least after the class refreshes are done because this is a waste of time otherwise. It's simply not possible right now with how disheveled everything is.