devtrackers.gg → NextdoorMMR
NextdoorMMR

NextdoorMMR



01 Nov

Autofilled in Legendary Queue = Free Loss

Comment

Originally posted by Aquaburnz

queue times would easily get quadrupled for popular roles and yes people care more than they think

Yes, in fact, much worse than quadrupled. If you do a quick simulation on paper you'll see that if, for example, there's only 1 JG for every 10 players, wait time lengths go like this:

2x 3x 4x 5x 6x ... 100000000x infinite

Basically, without autofill, or at least some similar way of players choosing to not play their preferred role, getting games at all grinds to a complete halt and no one gets to play.

It's an unfortunate part of having pre-chosen preferences that aren't equal.


04 Aug

Legendary q entry requirements should have been higher

Comment

Originally posted by Sgrinfio

It's not supposed to be the elite, it's supposed to be fair for everyone who wants to play solo. If they are that bad, beat them and climb.

That's exactly right. The Diamond 4 requirement is more to ensure a decent level of experience first, not to only allow elites.


27 Jun

16 game losing streak from 100+points Master to 1 point Diamond 1

Comment

Originally posted by manurift

From nextdoormmr own words:

When putting a match together, all the matchmaker does is:
1)Grabs 10 players whose rank and MMR are close enough to each other.
2)Makes sure it can assign each player to a team and role such that the total MMR between the two teams is within an allowed MMR gap and role preferences work.

So you won't get those massive gap difference you're talking about, that's would break the whole ELO system.
When you get into a losing strike, it means you are at your peak performance and the mmr is matching you with similary good enemies, so you have to give the best of you to actually win. That's a good thing, it means the mmr works, you just can't stomp your way to challenger.

Yeah, pretty much this.

The matchmaker does have some flexibility to expand the initial lobby creation MMR gap based on current population, as long as it stays within the allowed rank gap (usually +/- 1 major rank from each party's average rank).


20 Apr

Rankedsystem Solo/Flex Que

Comment

Low win rates can be deceiving because the system can find your MMR very quickly and it uses MMR in matchmaking. Some players get lucky and the system can't find opponents good enough early on and they end up with a positive win rate until they hit Master. The win rate is positive because their team MMR gap (their team - other team) has averaged positive for most of their matches.

Other players end up unlucky and the system is able to find similar and sometimes harder opponents to them and they end up with much harder matches and a negative team MMR gap on average.

You can have two equally good players with a pretty wide range of win rate (e.g. 45-55%). Because MMR is shown to be more accurate at predicting match outcomes, if you have 2 players with the same MMR and different win rates, it means they just had harder / easier match histories.

We can tell this in the data by:

  1. Find two players with the same MMR but one with a lower win rate and on...
Read more

16 Apr

This game feels like it's deciding who the winners and losers are.

Comment

Originally posted by TerribleTeemoTime

If match making is going to concern itself with player experience at all, can you explain why you would not default to giving a duo queue with adc/support as their top picks the duo lane over all other combinations except for another duo with the same preference?

Why on earth would there not be an extremely high priority on giving a duo queue that prefers the duo lane the duo lane? The only thing that would make sense is if neither of us got the duo lane because another duo also wanted it, but most often we get split up. Sometimes this causes my friends to ragequit the match or even the game entirely.

Most of the time you will get your positions, but there are exceptions that happen due to the random nature of all the factors matchmaking has to balance: wait time, ping, rank, MMR, team balance, and position. Sometimes you get another player in there who just hasn't been able to play their position recently, whereas you have, and it's just their turn so to speak.


15 Apr

This game feels like it's deciding who the winners and losers are.

Comment

Originally posted by pwaves13

While I understand that, taking potential players out of the pool of options, still I think that when my average queue time is like. 20s or so it's not that big of a deal to wait a bit when regular league I'm waiting a few minutes. Maybe it's a regional difference.

And I'm plat 1 soloqueue. Saw it last for sure in Plat 2 so...

If you saw it solo queue in Plat 2, that means the Gold was in a party with significantly higher-ranked teammate, pulling their party average into high-plat.

Since their party was considered Plat, the feature you are talking about wouldn't make a difference for you.

This game feels like it's deciding who the winners and losers are.

Comment

Originally posted by pwaves13

So out of curiosity why not allow an option to let myself have longer queue times and match with my rank?

Like I still get in with mid golds in plat1

Mostly because that type of choice doesn't happen in a vacuum. One player waiting longer makes other players wait longer because now they can't match with that player. This can cascade into everyone waiting longer, even those who don't want or need to.

Seeing some Golds as a Plat is pretty normal though. If you are both solo queue, that means you are low Plat. Once your party average (or solo) is farther into Plat, you won't see parties that average below Plat in rank.


12 Apr

This game feels like it's deciding who the winners and losers are.

Comment

Originally posted by griddle1234

To compensate for this, we try and make sure higher MMR players get extra Ranked Fortitude or LP to help them climb even if their win% stays near 50%.

This is an elegant system when taking all you've said into context. I would say for LP it should reward it based on the disparity each game rather than on their general MMR.

So let's say a 3000 MMR was paired with other 3000 MMR Players. They should not get the same high +15 LP gains compared to a 3000 MMR players that were paired with 2500 MMR players if that makes sense.

So the same system could be more dynamic and reflect the variation each game and would help people to be more motivated knowing they won a game where they were the carry.

But im sure there's been lots of testing on designs which work best but the LP principle makes a lot of sense.

Yes, I agree, and that's how the same system works on, e.g. League PC. A combination of what the teams looked like and where everyone should be going.

Wild Rift currently assumes every match is fair, which means in some cases the LP gains/losses don't seem to line up with what happened in the match. In the long run, it gets folks to the right place since most matches are fair, but it can result in some awkwardness. We are aware of this and are evaluating potentially changing it to be closer to PC in that sense at some point.

It's basically a simplifying assumption that carries some baggage but still works.

This game feels like it's deciding who the winners and losers are.

Comment

Originally posted by griddle1234

Firstly thank you for the detailed response to the above poster.

  1. Grabs 10 players whose rank and MMR are close enough to each other.

This in principle is good but I do find it hard to believe that's its a tight criteria. There are games where I've seen such huge disparity in skill level. This must be a lot looser than the 2nd point.

  1. Makes sure it can assign each player to a team and role such that the total MMR between the two teams is within an allowed MMR gap and role preferences work.

This is far easier to achieve and what I suspect is creating the 50/50 games which is obviously the aim. This is fair for a fun gaming experience but not fair from a ranked progression system.

You are basically saying that regardless of your skill/MMR level you will find it the same difficulty to win games and rank up.

I think the more point 1 works the more point 2 fe...

Read more

Yeah, you've pretty much hit on the major tricky design pivots when it comes to matchmaking and ranking.

If we focus on matchmaking, then players get closer to 50/50 games when some feel like they deserve a higher win rate. To compensate for this, we try and make sure higher MMR players get extra Ranked Fortitude or LP to help them climb even if their win% stays near 50%.

If we don't focus on matchmaking, then it's possible to just let people climb, but then you run into the problem you brought up which is it results in a lot more players getting into matches they can't win or into matches where they feel like they aren't good enough to even play the game. In my experience, this type of experience leads to a larger negative impact than when good players have to lose half their matches.

But, yeah, you have touched on exactly the types of topics we have to balance.

EDIT: To your comment about tight criteria, you are right. Rank is the only hard thres...

Read more

This game feels like it's deciding who the winners and losers are.

Comment

Originally posted by caesetic

I ask because last season I hit diamond with a 55% win rate and have subsequently been 38% this season in Emerald/Plat. I haven’t changed my play patterns in any way and the champ I main hasn’t been impacted by the balance team recently. I also have a 77% WR in normals.

This is the second time that I have had seasons with opposite results which results in some serious negativity from a play perspective. Perhaps I’ll rebound sometime in the near term, but the feelings remain.

Outside of just random luck, there's also when you play. If you got excited to play early this season because of how well last season went, you may have run into higher MMR players before they worked their way up to where they wouldn't match you. Sometimes this can result in less fair matches early on, though it's a bit rare and still likely just random luck.

Another factor is the time of day you play. If you play closer to prime time in your region, you're more likely to get teammates closer to your own MMR just because there are more players around to match with.

This game feels like it's deciding who the winners and losers are.

Comment

Originally posted by caesetic

What about situations that result in accounts having net 40% WR over a season or 60%? Are you viewing the goal for 50 as a lifetime for an account?

If so, while that is mathematically 50, I’d worry about significant retention issues as player perception of the system.

If not, how do you account for those accounts? After a certain time, shouldn’t the MMR adjust?

That can happen for a number of reasons, and is not ideal. We don't currently look at a 40% account and try to fix it with extra wins, nor at a 60% account and try to fix it with extra losses.

Most common reasons I've seen in real data:

  • High MMR players coming over from League PC and just starting out will play way better than anyone they are allowed to match with by Rank until they get up to Diamond+. These players get high win% because our rank gap restrictions prevent us from putting them an actual fair 50/50 matches. They end up in a lot of 60/40 matches until they hit ranks that have enough high MMR players to balance them out.
  • Likewise with really low MMR players. Sometimes we just can't make teams good enough to win if a player is just that bad. This is pretty rare though.
  • Similarly, the very very top 5-10 players will likely play so good they go positive because the best we can find is around 60/40 fairness. They make so large of an i...
Read more

This game feels like it's deciding who the winners and losers are.

Comment

Originally posted by Vorcia

Awesome, thanks for the detailed reply. If you can answer, how do you feel about the fairness of matches?

The devs on PC League published an article recently about how the matches in Ranked Solo Queue were extremely close in MMR except when duos were involved, which makes me believe that a lot of the unfairness in matchmaking that people in this community perceive might be caused by the ability to duo and even trio queue which increases the randomness compared to PC League. In that same article they mention that 99% of matches had all players within 2 divisions of one another, what would be the equivalent range for Wild Rift?

Do you think there might be any evidence to confirm or deny that the ranked queue in this game mode feels more random than PC League?

I can't really speak to how we compare with PC since I don't have their raw data in front of me.

But as far as what can lead to increased MMR gap, in WR it's more likely to be trying to get everyone their position. duo/trio doesn't have a huge impact there (average duos and trios still only win 50% of their matches). 50/50 matches are still the most common, but it can push up to 60/40 in some worse cases.

As for a feeling of randomness, that can also come from the actual skill gap within your team, which doesn't have as strict of a control. The team vs. team MMR gap has a value above which the matchmaker will never allow. The within-team gap, however, resizes based on real-time population.

Basically, it will find the closest MMRs possible within the allowed Rank gap. The rank gap also can't be loosened. But it is possible during slower times of the day to get players whose MMR is much lower than your own despite having the same Rank as you. When that happens...

Read more

This game feels like it's deciding who the winners and losers are.

Comment

Originally posted by Vorcia

Tbh, there's no proof either way but I've ranked up to Diamond+ in League, Wild Rift, and TFT, and at least for the other two games I can personally attest that the matchmaking and ranking systems have been very fair and accurate over a decently large number of games, like 150-250 per season.

That really wants to make me give Riot the benefit of the doubt for Wild Rift, although even then I have to admit that the fortitude system is really sus to me and leaves a bad taste in my mouth because it's the only system in the 3 games that rewards you for playing selfishly rather than playing to win the game. And on top of that, somewhat minor but there's even a daily bonus to fortitude (according to the wiki at least) which I think is really bad if true. It makes it so that ranked only really starts at Diamond when you lose the fortitude system but even then it makes me kinda paranoid like what if a similar metric system exists to control your VP gains.

Just personally, ...

Read more

Thanks for giving us the benefit of the doubt!

You are correct. We try our best to give each team the same MMR to lead to a fair match. We don't look at your win% at all, nor your recent Win/Loss pattern.

When putting a match together, all the matchmaker does is:

  1. Grabs 10 players whose rank and MMR are close enough to each other.
  2. Makes sure it can assign each player to a team and role such that the total MMR between the two teams is within an allowed MMR gap and role preferences work.

If that all goes through, the match fires up.

That's about it.

This leads to matches being winnable by both teams. It doesn't mean there aren't outliers, or matches that can snowball, and there are a lot of other factors that can show up between team assignment and actual match play. But that's how it works.


05 Apr

Dev Update Regarding which Ranks can be Matched with Eachother

Comment

Originally posted by robotbeatrally

Plat players are not supposed to see diamonds anymore. You either get matched up or down but not both. I'm emerald and I'm mostly matching with emeralds but I noticed myself matching with all platinum players ones and all diamond players once. Which was weird. I don't think the game is hard enforcing the rules with premades though. I'm sure you can get unusual match ups still but they are just less common.

It does enforce the rules with premades (outside of a full 5q) by using the average rank of the premade. If the average rank of a premade is Emerald III (e.g., Emerald IV with Diamond IV), then they could match with Plats if needed, so in that case you can have a Diamond in with Plats. It will still balance the teams on MMR though, so both teams will have an equal shot at winning.

Dev Update Regarding which Ranks can be Matched with Eachother

Comment

Originally posted by pradohood

Hi Josh, do you guys factor in a player's historical rank into matchmaking? At emerald me and my friends often run into opponents who are currently diamond but historically master/gm/challenger

A certain amount of past rank is baked into their current rank because it determines how far back we will roll a player when a new season starts. Outside of that, no, because we have found historical rank to be much less accurate than a combination of current rank and MMR.

It's possible to run into situations like yours if the system needs to expand the MMR gap enough to find matches, however the team balancer will always make sure that your team has just as much MMR as the other team does.

Dev Update Regarding which Ranks can be Matched with Eachother

Comment

Originally posted by Elegastt

Hey Josh, thanx already to answer so many questions, especially given the controversial topic.

I got 2 questions regarding what i believe is the biggest frustration atm: the difference between actual rank and actual skill:

  • any plans to implement ELO for plat and emerald? This way it's easy to give everyone fair matches while still allow people to climb if their skill is higher than its current rank.
  • don't you think it would be better of there are more people in the lower ranks? I know this would be a controversial thing, but if people are more evenly spread across ranks it would reduce skill difference within ranks, no? That way you don't need to hard constraint rank difference

Plat and Emerald still have a back end MMR that is used to matchmake and balance teams. The gap that is allowed within a team for that MMR is dependent on who is available when you play. The closer to peak hours you play, the better your teammates will be.

Dev Update Regarding which Ranks can be Matched with Eachother

Comment

Originally posted by fedekun

Is there an MMR for PVP or is it just random? I swear (Master MMR in Ranked) get matched with people that just installed yesterday when I play PVP with friends.

PVP uses MMR only, however that MMR can expand as needed to get you into a match quickly.

Dev Update Regarding which Ranks can be Matched with Eachother

Comment

Originally posted by Euphoric_Software481

Hello Josh. Thank you for taking the time to do this. I would like to hear your opinion on a small thing I have in mind.

How hard do you think it will be to implement a system where if one role in your team is an auto fill, the enemy in that same role will also be an auto fill. We desperately need a system like that this because I can not express how infuriating and sad it is to see your E4 Lux main JG auto fill get smacked all over the map by an Ex GM D4 JG main who lives and breathes objectives. It simply doesn't seem fair. You just cannot expect an auto fill JG to match the macro and micro of a JG main. And just because of this one simple mismatch, your whole team pays the price.

Maybe it can never be implemented for all roles since it would be too complex or increase Q times by a lot, but can we have this for JG role at least? It would be a very-appreciated change for the community.

Also, ignore the trolls and negativity that you probably keep encounter...

Read more

We are actually investigating this exact type of thing. I can't speak to the difficulty of the implementation yet though. It would also have wait time implications.

We actually have measured exactly how much of an impact having a mismatched autofill JG has on winning, and while it does have a measurable impact, it is orders of magnitude less than the overall skill difference between teams. Which means in practice, despite the noted disadvantage, it is consistently overcome by teammate skill in the data.

That said, like I said, we are still looking into it because we agree the feel of it is disproportionate with the actual outcomes.

Dev Update Regarding which Ranks can be Matched with Eachother

Comment

Originally posted by UnemployedWeebTrash

Hi, I am currently GM 270 points, I'm just wondering how do you classify how much points a player gets? All of my friends receive +20 after reaching master till they reach challenger while I receive +15 or +16 since reaching master.

Thanks in advance!

It means they have a higher MMR, which in turn means they have been more likely to win matches against harder opponents.

Dev Update Regarding which Ranks can be Matched with Eachother

Comment

Originally posted by airstrike

Have ranks gotten easier to climb? I’m just as bad as I’ve always been and struggled to get to top levels of Emerald in season 2 and now breezed through to make it to Diamond (though hard to get out of Diamond IV)

There were a few updates we announced last season that helped players get to ranks they were supposed to get to before, but weren't.