Opt-in matchmaking to improve the quality of PVP

[White Kitsunee.4620]

@Trevor Boyer.6524 said:A few things to note:

1. Most plat 1+ players almost entirely unanimously agree that'd they'd be willing to wait 20+ minutes for high quality games. This is what I have been hearing from everyone, whether in forum or in game or in discord.

Hard to believe I'd be in the minority here but I suppose I am. I would dislike that greatly.

_"I would rather play only 2 or 3 games in the span of 2 hours that had very precise match making, than play 8 games in 2 hours that felt like dice roll.

I suppose this is a difference in philosophy between me and everyone else. And I'll go into that after I list this.Other notes about que time:

Pros/potential pros1.Heavily discourages que spamming as a method of climbing.2.Higher chance of players being invested in individual matches.3.Reason to work on improving skill level since you cant que spam.4.Lower chance of leavers or afk players.5.Good for fostering the competetive community.

Cons1.Doesnt guarentee good games, just makes them more likely.2.Punishes off-meta experimentation2A. Meta is more likely to become even more centralized around strongest classes.3.Gettint a afk is gonna feel 3x as bad(I still wouldnt recommend a surrender feature though)4.Potenial for community to be more toxic around weaker players.(potential for them to become better as well, just less so.)5.Awful for casuals like myself.6.Could drive a large number of players away.

I think this ultimately has to do with your philosophy surrounded GW2 PvP.In theory this should be an very healthy change for the competetive community, mostly because of Pro. 2.Players being invested in each individual game will be a huge improvement alone.

However if your not in the competetive community this is awful, and could really weed out the casual PvP community.

@Trevor Boyer.6524 said:For those of you who ONLY played unranked/ranked, I'm not so sure you guys recognize what you're missing out on, concerning how fun this game actually is when the match is balanced. Because I'll tell you what, after a good night of scrims where two teams are very evenly matched, it can be hard to go back and stomach ranked play sometimes.

Maybe your right and drastically improved match quality could harbor a strong core competetive audience, maybe people will realise that the game is actually a competetive masterwork and a chain reaction will be started.

But for me that honestly sounds awful. I just play PvP because I really enjoy its combat, not because I enjoy 5v5.

And in fact after the mini 2v2 season, I would reword it to I enjoy PvP inspite of 5v5.

[SardonyxDragoon.8213]

I love guild wars 2 spvp, I want to be able to que and not feel punished for doing so. I've been playing for several years, mostly just pvp. Many games i'm in have 100% obvious new players in them who do not understand the game mode AT ALL. Or, players are they're for 1) fast dailys, 2) fast rewards from tracks 3) whatever gear piece they are grinding for. They are not there to look at the rank, they are not there to learn the game, they are not their to learn from others on how to improve, they are not there to change from the same DH trapper build they've been using since HoT came out, they are not there to read map chat, they are not there to make plays on the map to win games, they are not there to contribute anything meaningful to the team, the game, or the community, they are there to get one of the 3 things I mentioned above. Either that, or they are just genuinely new and have no idea the difference between what is pointless suicide on the map and what is an effective decision, let alone that if they did would they have the knowledge to effectively execute that decision taking into consideration their own class, build, or the team composition overall against the enemy, or even more so that the rest of the team isn't making that correct decision entirely pointless as they squander the other parts of the game.

The differences and gaps here are large, not small, certainly not representative by the rank that is arbitrarily gained by the dice roll game of chance that is the 10 qualifier matches. This isn't a new problem, threads like this aren't new either. I had uninstalled the game last year because the return on playing gw2 just didn't justify the pain of playing because of these issues, seeing the new rollouts to balance renewed my hopes and i'm playing again this season. There's problems, its okay, discussing them is the best thing we can do as a community imo, I don't want to just uninstall and write it off as hopeless.

I think that if we separated the rewards from rank we might be better on some of the matchmaking: que looks like this: Que for Rewards/Loot (contains no ranking system(but can still use the same matchmaking system, just no assigned visual for rank) = progress reward tracks, make progress through dailies, obtain parts for gear etc) and a separate: one: Que for Glory! (obtain a starting rank, and have the ranking tiers etc, does not count for dailies, differentiated rewards/remove rewards you get from que for loot, the reward is being able to que in quality games and climb rank in a viable way with other people who desire to do the same).

In order to qualify for Que for Glory new boundaries and systems must be put into place about teaching the game mode to new players. There needs to be instanced tutorials with npc targets on nodes, there needs to be hud arrows telling you to run here and fight in the point, like the silly dodge circle pve players get in starting zones, this level of tutorial needs to be a thing for pvp, and new players need to have their hand held and be told that hey, if you run alone to far and try to fight there 3v1, your probably going to die and your team is going to suffer, and if you keep doing that over and over again the whole game then your team is probably going to lose, so that's when the training hud arrow points and says "Help your team!" and shows them how to run back to home or mid, and gives them a little training course like that.A written guide can be put in place for players to read explaining that different classes are more likely to do certain things than others, that class and build choices matter and so do skills and equipment.

I wouldn't be saying this if I wasn't in mid gold all season because I keep having matches with players that are not making good decisions, and are stating that they are new. This is non issue if gamed with relatively similar players, and the issue of low population keeps being pointed out. I would also agree that I would rather hit the que button, be told that it could be upwards of an hour long wait if any games get matched at all, then go about actually playing guild wars 2's other options while waiting for que to pop, then just sitting through another mentally wrecking que spam 6 loss streak to one freak of nature team match up to another, with the only win of the night being a complete steam roll 74-500.

The alternative is to get good and carry, on a build that allows that, with all the other people doing that, that are all just as salty as I am, only to get another team with some team member or another that makes it more than difficult to carry, against a team that may or may not have a full team that knows how to play.

If rank could be determined by decision making that would be amazing, like, if there were a way to see a players patterns of pointless suicide and count that as they need to be placed in a lower match than making quality decisions/plays, but this is just fantasizing, this is all just fantasizing. Prove me wrong anet. Sorry for anyone who read this, I only go on this much because I actually like the game, and I guess I just needed to vent.

[Exedore.6320]

This design of optional more precise matchmaking has severe flaws. It's extra development work which at best accomplishes nothing and at worst penalizes people who don't opt-in.

First, you have to realize that In order for the opt-in condition to be met for a single player, all 10 players must be within a certain rating difference when the matchmaker tries to make a game. If the population is high - which it is for most skill levels most times of the day - there's no difference between normal and opt-in queues. So this will only matter when the population is low - very off peak hours and at the rating extremes. Keep in mind that where you fall on rating doesn't really matter an analysis of this idea's flaws; rating only affects how often you experience the low population cases.

Next is how the matchmaker works, which has the framework of a "greedy algorithm". It takes the roster (players which queued together, including a solo player) at the front of the queue and examines it for a match by trying to build a match (10 players) around that roster. It does so by finding the next 100 max rosters in queue whose rating falls within the examined roster's rating range. Rating range expansion is applied only on the examined roster. If not enough players are found (min 20) or if there are no desirable matches (this is where roster size and class distribution come into play), the examined roster moves to the back of the queue. The process then repeats.

If we make the assumption that for low population, the 5min rating expansion is kicking in almost always, then your opt-in roster will typically not find a match in that initial window either. The normal (not opt-in) rosters then expand their rating. That means the most likely successful match is expanded normals with some non-expanded normals. The matchmaker puts them into a game. Meanwhile, the opt-ins languish in queue. There are likely enough total people queuing to find a match for opt-in rosters, but because they're not all in queue at the same time and the expansion time is much less than the time of a game, it's unlikely that there will be sufficient matches for opt-ins before the normals are pulled out for games. And due to that assumption about low population, the opt-ins will sit in queue for 10-15min until a single active opt-ins only match finishes.

If the opt-ins become a majority in a low population scenario, normal rosters will be at their mercy for queue length, and will have even worse matches as their rating margin increases with their longer queue times. You can already see from the above scenario how queue time dramatically increases for opt-ins. Again, remember the assumption about not enough non-expanded players already. Because there are now too few normal rosters in a small expanded range, those rosters must sit longer to expand their rating margins even further than the current system. Worse yet is that for extreme outliers (e.g. your legendary players), a large number of opt-ins in high platinum would cause them to be matched with gold players after longer queue times. Those players are unable to use the opt-in if they want to play because there are rarely enough people in their base rating range queuing at the same time.

---

It would be interesting to look at the data on how many players the matchmaking is considering for low population scenarios, since there is a minimum to make a match. If there's 15 people in the range but not 20, maybe it's better to use the lower matching population for a narrow rating. The downside is that those smaller pools becomes more easily manipulated.

[bethekey.8314]

@"Exedore.6320" said:This design of optional more precise matchmaking has severe flaws. It's extra development work which at best accomplishes nothing and at worst penalizes people who don't opt-in.

Great post and I appreciate you taking the time to write this all out / think about it more. I agree that there are flaws, but I believe they can be managed and the system can have a major benefit. I'll try to address them here.

First, you have to realize that In order for the opt-in condition to be met for a single player, all 10 players must be within a certain rating difference when the matchmaker tries to make a game. If the population is high - which it is for most skill levels most times of the day - there's no difference between normal and opt-in queues. So this will only matter when the population is low - very off peak hours and at the rating extremes. Keep in mind that where you fall on rating doesn't really matter an analysis of this idea's flaws; rating only affects how often you experience the low population cases.

Yes, this system is specifically meant to address problems at rating extremes. If you fall within the bulk of the player rank distribution, you are likely getting well-matched games already.

Next is how the matchmaker works, which has the framework of a "greedy algorithm". It takes the roster (players which queued together, including a solo player) at the front of the queue and examines it for a match by trying to build a match (10 players) around that roster. It does so by finding the next 100 max rosters in queue whose rating falls within the examined roster's rating range. Rating range expansion is applied only on the examined roster. If not enough players are found (min 20) or if there are no desirable matches (this is where roster size and class distribution come into play), the examined roster moves to the back of the queue. The process then repeats.

Your argument seems to be based on the existing design of the matchmaker algorithm. While this can be changed, I don't believe major changes are feasible ("extra development time") and will try to account for it.

If we make the assumption that for low population, the 5min rating expansion is kicking in almost always, then your opt-in roster will typically not find a match in that initial window either. The normal (not opt-in) rosters then expand their rating. That means the most likely successful match is expanded normals with some non-expanded normals. The matchmaker puts them into a game. Meanwhile, the opt-ins languish in queue. There are likely enough total people queuing to find a match for opt-in rosters, but because they're not all in queue at the same time and the expansion time is much less than the time of a game, it's unlikely that there will be sufficient matches for opt-ins before the normals are pulled out for games. And due to that assumption about low population, the opt-ins will sit in queue for 10-15min until a single active opt-ins only match finishes.

If I understand correctly, the major flaw you point out is normal queued people with expanded rating ranges getting placed into matches faster, such that the 10 player threshold for an opt-in match is never met until a surge of new, in-range players queue. The simplest answer to this is that a 10-15 min wait for matches to finish isn't so bad to high level players looking for good games.

A better answer would be to point out that many people will simply restart their queues if they last too long, realizing they're about to be placed into a poorly matchmade game and negating rating expansion as a factor. Taking this idea, a co-opt system could be introduced, where if the matchmaker finds there are almost enough well-matched players for a full game, the normal player is either automatically co-opted into or given the option to opt-in and accept stricter matchmaking. It shouldn't be very hard to fill 10 players that way.

If there are barely enough well-matched players to fill a game, then that's just how it goes. Opt-ins would realize there's no chance for the games they want to happen, so they opt-out of stricter matchmaking or don't queue. For some people trying to rank up, no queue is often better than a bad match.

If the opt-ins become a majority in a low population scenario, normal rosters will be at their mercy for queue length, and will have even worse matches as their rating margin increases with their longer queue times. You can already see from the above scenario how queue time dramatically increases for opt-ins. Again, remember the assumption about not enough non-expanded players already. Because there are now too few normal rosters in a small expanded range, those rosters must sit longer to expand their rating margins even further than the current system.

I don't see this as an issue. If opting-in for stricter matchmaking is popular and becomes a majority, the system should have stricter matchmaking as a baseline then. The co-opt system I proposed above also addresses this issue.

Worse yet is that for extreme outliers (e.g. your legendary players), a large number of opt-ins in high platinum would cause them to be matched with gold players after longer queue times. Those players are unable to use the opt-in if they want to play because there are rarely enough people in their base rating range queuing at the same time.

It's unreasonable for legend or plat 3 people to expect full matches within their bracket. These players would necessarily be grouped in with high plat and the co-opt system addresses this too.

It would be interesting to look at the data on how many players the matchmaking is considering for low population scenarios, since there is a minimum to make a match. If there's 15 people in the range but not 20, maybe it's better to use the lower matching population for a narrow rating. The downside is that those smaller pools becomes more easily manipulated.

Good point. I'm curious as well.

[Dantheman.3589]

I’m gonna be honest and say no ty to original OP, for one these matches of 1 high rated player versus bad players only really occurs in off hours, so it’s not an inherent problem with match making just one that is a mix of somewhat low population and the fact that unless you have large population of players transferred from another time zone, that there will be a lot less ppl playing in off hours. The real solution at least in my experience in some games with a larger population is they provide a message that explains population is low, this could be maybe the only part that I think the op might be right on- you could have a check box where you either opt out or in to being ok with being placed with players well out side your elo. This would give the option to select the type of matches you want w/o being super exploitable- though tbh it will be exploited 100%, so I’m not super attracted to the idea

[Stand The Wall.6987]

yeah right lol you guys would be willing to wait 15 mins for a q until 3 days later the forums get flooded with "zomg q time too long anet pls fix" threads. it would be the same handful of ppl playing each other over and over. also this opt in would not be available for ppl not in prime time.

I still think it would be worth it tho.

[bethekey.8314]

@"Stand The Wall.6987" said:yeah right lol you guys would be willing to wait 15 mins for a q until 3 days later the forums get flooded with "zomg q time too long anet pls fix" threads. also it would be the same handful of ppl playing each other over and over.

You think most of the people who "zomg anet pls fix" are in high plat or legend?

[Stand The Wall.6987]

@bethekey.8314 said:

@"Stand The Wall.6987" said:yeah right lol you guys would be willing to wait 15 mins for a q until 3 days later the forums get flooded with "zomg q time too long anet pls fix" threads. also it would be the same handful of ppl playing each other over and over.

You think most of the people who "zomg anet pls fix" are in high plat or legend?

a lot would be after enough time of atrocious q's.

[bethekey.8314]

@Stand The Wall.6987 said:a lot would be after enough time of atrocious q's.

See, I consider the atrocious queues to be matching legend/high plat players with gold/silver players.

I believe there should be options for people who think differently, as we clearly do.

[Exedore.6320]

@bethekey.8314 said:See, I consider the atrocious queues to be matching legend/high plat players with gold/silver players.

I've seen no factual evidence of this large of a gap (legendary with silver) happening except during extremely dead times. Everything is anecdotal or incorrectly interpreting decayed rating as actual rating.

[bethekey.8314]

@Exedore.6320 said:

@bethekey.8314 said:See, I consider the atrocious queues to be matching legend/high plat players with gold/silver players.

I've seen no factual evidence of this large of a gap (legendary with silver) happening except during extremely dead times. Everything is anecdotal or incorrectly interpreting decayed rating as actual rating.

If it really needs to be said, that was a rough example. And to be clear, none of us (on both sides of the argument) have enough data to accurately assess this problem. The purpose of most forum posts here is to simply suggest things to Anet, which does have that capability, and not to arrive at a definitive conclusion ourselves.

[Dantheman.3589]

@bethekey.8314 said:

@Exedore.6320 said:

@bethekey.8314 said:See, I consider the atrocious queues to be matching legend/high plat players with gold/silver players.

I've seen no factual evidence of this large of a gap (legendary with silver) happening except during extremely dead times. Everything is anecdotal or incorrectly interpreting decayed rating as actual rating.

If it really needs to be said, that was a rough example. And to be clear, none of us (on both sides of the argument) have enough data to accurately assess this problem. The purpose of most forum posts here is to simply suggest things to Anet, which does have that capability, and not to arrive at a definitive conclusion ourselves.

It’s pretty simple though, watch streams of any high ranked player during prime time- almost all those matches are at least plat2+ sometimes all plat3+. While if it’s late asf matches can be mixed but still a few decent players are playing during those hours. By high ranked I mean, at least for NA, like naru not players like mukluk or noody, sometimes players who are at max like plat1 and don’t even play meta builds them selves have the audacity to claim there teams low ranked when really they aren’t high enough themselves. While actually high ranked player, even if they complain, it’s usually just becuz ppl aren’t playing well or to what their standards are for Ats or team qs

[bethekey.8314]

@Dantheman.3589 said:It’s pretty simple though, watch streams of any high ranked player during prime time- almost all those matches are at least plat2+ sometimes all plat3+. While if it’s late asf matches can be mixed but still a few decent players are playing during those hours. By high ranked I mean, at least for NA, like naru not players like mukluk or noody, sometimes players who are at max like plat1 and don’t even play meta builds them selves have the audacity to claim there teams low ranked when really they aren’t high enough themselves. While actually high ranked player, even if they complain, it’s usually just becuz ppl aren’t playing well or to what their standards are for Ats or team qs

I believe matchmaking is fine right now during prime times. That being said, these are extraordinary times with many people forced to stay home and the announcement of a new expansion, likely raising the population. Furthermore, some people aren't able to play during prime times.

This problem was more apparent in January 2019 when I created the thread. I believe the game will return to that state in time.

[Dantheman.3589]

@bethekey.8314 said:

@Dantheman.3589 said:It’s pretty simple though, watch streams of any high ranked player during prime time- almost all those matches are at least plat2+ sometimes all plat3+. While if it’s late asf matches can be mixed but still a few decent players are playing during those hours. By high ranked I mean, at least for NA, like naru not players like mukluk or noody, sometimes players who are at max like plat1 and don’t even play meta builds them selves have the audacity to claim there teams low ranked when really they aren’t high enough themselves. While actually high ranked player, even if they complain, it’s usually just becuz ppl aren’t playing well or to what their standards are for Ats or team qs

I believe matchmaking is fine right now during prime times. That being said, these are extraordinary times with many people forced to stay home and the announcement of a new expansion, likely raising the population. Furthermore, some people aren't able to play during prime times.

This problem was more apparent in January 2019 when I created the thread. I believe the game will return to that state in time.

Ok ok the thread is older and yeah some ppl are playing a little more and there was also a big balance patch, but seriously like I have a title from then which was plat3+ only at that point and literally the only time you would get lopsided matches was Qing at late hours and q dodging, otherwise you would always see high level matches, so I’m not sure if this applies for 1 highest ratings last season were low asf so maybe if you meant that season but even earlier than that? High ratings for top 10-25 were pretty decent so I don’t get the problem.But your right, sry I thought this was a current thread