Team Matchmaker Info&Discussion

[FtXCommando]

As some of you might have heard, Team Matchmaker is the new “Big Step” that’s being taken for the FAF Client now that ICE and the general client transfer to java has been undertaken. This thread is going to function as a way for me to tell you, the public, about what the devs are up to and why we’ve taken the course we’ve taken on certain aspects. Don’t worry it isn’t a dictatorial relationship and there will be things I hope to get some responses and debates about in order to create a better synthesis of the issue myself. This thread will work as a way for you to influence the devs through me (and any devs that want to read the thread) provided you’re willing to expand on your rationale.

So in order to understand why I’ve decided on a certain course with some of my decisions on the Team Matchmaker I’d like to break down an introductory question before getting into the details of the concept. Why have we chosen Team Matchmaker as the next project to undertake?

Well several reasons:

1. The need for a competitive teamgame environment.
2. The need to improve new player retention on FAF.
3. The desire to foster an environment that can properly unite the various existing community’s of FAF together.

Spoiler: show

How does TMM address competitiveness? Well it requires us to go even DEEPER and breakdown trueskill and global rating. For those that don’t know, trueskill is a rating system developed by Microsoft and utilized for games like Halo 2. It’s great not only for 1v1 situations but also for team scenarios.

So why doesn’t it work in the current environment? Why is TMM needed? Simple. Our current custom game environment allows you to control several factors that trueskill does not naturally account for thus creating an environment with a predisposition to rating manipulation/integrity issues/inaccuracy, you name it. Things like choosing your map, your opponents, and your allies all before the game launches will harm the integrity of trueskill and create these situations were you see 2000+ rated players that can’t properly manage a 1500 on traditional teamgame maps because their rating has come from defeating high deviation players on simple maps.

Now even though global rating has been corrupted to the point of meaning little about your real capacity as a player, it still is really the only tool that can be uses to balance a game ‘competitively.’ So when the 1500 player that got his endorphin hit destroying in his astro games decides to join a wonder game, he will get flamed for his inability to realize what reclaim is and likely refuse to leave his bubble again. He does not want a poor game experience (getting crushed), to be abused (flamed), or to lose rating (ego). So he manipulates the factors within his control and goes back to the map he knows.

This is a major factor for why popular maps stay popular. Dual gap and astro are always up and so getting good at them is immediately rewarded with more games compared to ‘competitive’ teamgame players that just whine in their 3/8 lobbies on a map no one has seen that month. This directly harms the competitive players on a manpower basis leading to less games and overall a less competitive atmosphere.

So how exactly would TMM deal with that? Isn’t it just going to move the whining from the ‘competitive’ lobbies into aeolus? We expect there to be a couple reasons that TMM juices up the competitive scene and it relates to a lot of the issues already found with global as well as some client readjustments.

For one, the issue of a person farming 1500 in astro will not factor into TMM. If someone decides “I’m sick of this map I want to play something different” then he will be able to match in TMM starting from a new rating where he is much less likely to be flamed for his inability. Trueskill, after 5 or so games, should put him at the level he ‘deserves’ to be at in regular games and he can continue as he pleases. The point is that once the spark to play more competitive games comes into the individual, FAF will now have a way to fan the flame rather than indirectly snuff it out. This should increase the pool of regular FAF players that eventually become ‘competitive’ players and thus further increase the pool/health of the scene.

For new players, it’s irrefutable that custom games aren’t exactly the best environment. What the hell is 1000+? Why did he kick me with a message telling me to go away? I waited in this lobby since it was 3/8 and the host kicks me at launch? What a waste of an evening I’m going to go play smm2 instead.

We will work on a way to put matchmaker front and center for the player and make it so that if they want a game, all they need to do is hit a button and wait. New players tend to prefer custom games because a teamgame is much safer and easier to swallow than a 1v1, and we hope that TMM provides them a way to get that safe experience without the toxicity of custom games. So overall, we expect both new player retention and for more regular, niche players to enter the general, competitive community.

SO NOW TO DISCUSS THE DEVIL IN THE DETAILS!

I’ll break this one down by section rather than making a thick rant.

First Dilemma: How to make TMM come out in a reasonable time?

The solution to this one was to implement it a step at a time.

Spoiler: show

We’re going to start with 2v2 matchmaker and work our way up. However, we’re also hoping to see that it won’t be wasted effort to keep going and increasing the complexity of the matchmaker so if you want to see 3v3/4v4 TMM we’d still like to see some participation! As more players are factored into the equation, you will need to implement ways for the system to create good games and while there is a general idea floating around it’s not a good idea to implement it without getting a solid foundation with 2v2. Likewise, a matchmaker with several game options will require a way to multi-queue as only being able to queue for a singular matchmaker would create silly situations of potentially a dozen players searching and not getting a game even if they all would be willing to play other gametypes.

Second Dilemma: Premade teams vs Random teams

We’ve decided on implementing premade teams rather than having full random teams.

Spoiler: show

Why would we make that decision after I just argued for TMM to increase competitiveness? Well, because sometimes making a system as competitive as possible for the individual will harm the greater whole of society. Now FAF isn’t some market economy, but I feel the comparison still holds. Premade teams reward the develop of connections between players that eventually are what bring them back to the game. I can personally attest that if I didn’t build the friends I made when I came back to FAF in 2016, I’d probably have left like I did in 2014 rather than the road that led me to Player Councillor. Random queueing, not only in FAF but in other games, greatly harms the communal aspect that must be nurtured in order to maintain a loyal playerbase. We cannot allow that communal aspect to create a wholly cliquish environment (as we now have with global games), but we cannot eliminate it without destroying the strands that bind us together.

A major reason I reached this decision was clans. I see very little point in the establishment of clans on FAF now. It’s mostly just dudes that share a sense of personality and that’s about it. I see clans as an organization to group individuals that want to create a community on FAF and it’s this precise group that needs to be encouraged to improve by playing together in things like a TMM. This will lead to, overall, a healthier sense of belonging and an increase in retention.

I also feel that this provides a healthy quantity of control for players. If they want a quick game, they have random queue that can put them with a player to get them into a game. If they specifically want a highly rated game, they can queue with a high rated friend and wait to get a game.

I have little reason to believe that a premade team of 2 will beat a random team of 2 on random maps judging by the results of the morax 2v2 tourney that was recently hosted. That tourney provided an excellent case study because the (arguably) two most well-known “premade” 2v2 teams of petric/bh and blodir/nexus that have played together for years had lost to jagged/farm. Jagged and Farm had been a team made spontaneously the day of the tournament and were able to draw the other two teams constantly until they finally won just recently. This tournament, at least to me, concluded that the biggest factor in premade teams that scares players is the fact they often have a map specific BO they whored to farm rating on a map. If at the absolute highest tier of FAF gameplay, a premade team was insufficient for dealing with a random team, there is no reason to worry about it at the lower ratings either.

Third Dilemma: Full Share vs Share Until Death

We’re going to be using full share for 2v2 matchmaker.

Spoiler: show

Honestly I didn’t expect this one to be so controversial. Many devs argued that share until death was the condition that would lead to the best user experience and gameplay for several reasons. I personally was willing to go as far as to require all matchmakers above 1v1 to require full share but I compromised on only doing so with 2v2 for now.

So why is it necessary for 2v2? Well the basic reality is that the high level of FAF has experimented with both and came to the conclusion that full share is a necessity for a good gameplay experience. If we refuse to accept the commonly reached opinion of the truly competitive and expect to reinforce a competitive atmosphere with said matchmaker, we’ve kind of destroyed a major point of the system.

A 2v2 game is not like other teamgames where a loss can be significant but not absolutely game defining. There is basically no 2v2 map where losing an ally will not lead to an automatic loss you cannot stop without killing the enemy in the next minute. However, full share allows a player to recover from their loss and potentially win the game back. This is not easy. Running on half the apm and suddenly adjust to new gameplay circumstances is not a simple skill to acquire. Those that argue full share incentivizes turtle gameplay and allowing an easy win for the side that should have lost for losing as important of a piece as an ACU are looking at one type of game. The sort of game where you invest 10k mass into gunships/some army/whatever at the cost of your eco just for the sake of killing an ACU.

Yes, the ACU is not an autolose condition anymore. In a full share game you must be able to not only kill an ACU but project control over the area that the ACU was killed in within a reasonable timeframe if you expect to win. It requires a different mentality but it is necessary to avoid 2v2 devolving into no one leaving their base in fear of a double eco snipe against the individual.

With regards to 3v3 and 4v4, my central argument rests on potential connection issues. I simply see full share leading to a generally less toxic experience. I do not have a real opinion on the matter gameplay wise as the gameplay seems the same to me regardless of whichever option you choose but if people are often having connection difficulties I will likely ask for full share to be introduced.

Fourth Dilemma: Multi-queueing

Dynamic Interval System (Go join FAF Discord Tester Group to help out!)

Spoiler: show

I’m sure most people can agree that dividing a playerbase into 1v1-4v4 and not having them interact within the matchmaking algorithm is a recipe for disaster. So how do we deal with it? Multi-queueing! Easy devs just do it! Except not so easy.

In fact it’s probably not possible with the current continuous queue system unless we got some newton-tier bigbrain not contributing somewhere on FAF. So what was the solution we came up with? A dyanmic interval system?

Word sounds sexy what does it do though? Well basically you will queue for a certain matchmaker (eventually several) and then you will get a time to wait for games to get set up based on the last queue. This means the system will have a certain quantity of players to create a game all at once and can potentially go down the list from 4v4 to 1v1 matching up high quality games based on what you queued for.

If you want to help out with implementing this, Askaholic will be working on testing it in the near future. Please go to the FAF Discord and join the “Testers” group or PM Askaholic there if you want to help out with the code.

Fifth Dilemma: Global Rating

Will be kept, but will be suppressed as much as possible.

Spoiler: show

Yeah that sounds great FtX thanks for wasting my evening with your 20th thread over 2000 words. I don’t really give a shit about the matchmaker and want to play sentons. Are you gonna be bothering me?

MAYBE!

I’m not going to be deleting global. I do plan on eliminating it as much as possible, though. Why? Well, read above. It’s inherently a bad system prone to manipulation and I want to minimize it as much as humanely possible without driving away our beautiful senton pros that memorize tree groups for fun.

I plan, once matchmaker is in full force:
- to eliminate global rating leaderboards. It isn’t a leaderboard it’s a high score chart for who can get away with the most garbage before a mod takes interest in their replays.
- to get rid of global rating as the ‘face’ of a user. When a new player hovers over a user in aeolus, they are first greeted with their global rating. They will assume these are the best players. These become the players that these new players end up looking up to and they begin copying their behavior. I wish to instead have players show their highest matchmaker rating and a drop menu of the rest of their matchmaker ratings. The reason why it will be the highest will be revealed at a later point in time when a new system that is being worked on is closer to production.
- I also plan on having your global rating replaced in lobby with your highest matchmaker rating if that matchmaker rating is higher than your current global. Global games will still only affect your global rating and the matchmaker rating will appear as a different color in the UI so you can tell whether a rating is global or not.

Any further questions, comments, disagreements, or issues? Feel free to use this thread for all TMM thoughts. I will be using this thread to post updates about feature implementations as they become relevant.

Stay tuned!

[Blodir]

I agree with everything!

[Farmsletje]

I have little reason to believe that a premade team of 2 will beat a random team of 2 on random maps judging by the results of the morax 2v2 tourney that was recently hosted. That tourney provided an excellent case study because the (arguably) two most well-known “premade” 2v2 teams of petric/bh and blodir/nexus that have played together for years had lost to jagged/farm. Jagged and Farm had been a team made spontaneously the day of the tournament and were able to draw the other two teams constantly until they finally won just recently. This tournament, at least to me, concluded that the biggest factor in premade teams that scares players is the fact they often have a map specific BO they whored to farm rating on a map. If at the absolute highest tier of FAF gameplay, a premade team was insufficient for dealing with a random team, there is no reason to worry about it at the lower ratings either.

This isn't entirely true because me and jagged already knew eachother + got in voice even though our team wasn't premade. In your average non premade team in TMM the players will most likely not know eachother and certainly not be in voice with eachother (For the higher rated players this will probably be less of an issue). Maybe you could do something like making 2 ppl only TMM voicechannels in the discord server and promote it?

[Bennis-]

Yeah that sounds great FtX thanks for wasting my evening with your 20th thread over 2000 words. I don’t really give a shit about the matchmaker and want to play sentons. Are you gonna be bothering me?

MAYBE!

I’m not going to be deleting global. I do plan on eliminating it as much as possible, though. Why? Well, read above. It’s inherently a bad system prone to manipulation and I want to minimize it as much as humanely possible without driving away our beautiful senton pros that memorize tree groups for fun.

I plan, once matchmaker is in full force:
- to eliminate global rating leaderboards. It isn’t a leaderboard it’s a high score chart for who can get away with the most garbage before a mod takes interest in their replays.
- to get rid of global rating as the ‘face’ of a user. When a new player hovers over a user in aeolus, they are first greeted with their global rating. They will assume these are the best players. These become the players that these new players end up looking up to and they begin copying their behavior. I wish to instead have players show their highest matchmaker rating and a drop menu of the rest of their matchmaker ratings. The reason why it will be the highest will be revealed at a later point in time when a new system that is being worked on is closer to production.
- I also plan on having your global rating replaced in lobby with your highest matchmaker rating if that matchmaker rating is higher than your current global. Global games will still only affect your global rating and the matchmaker rating will appear as a different color in the UI so you can tell whether a rating is global or not.

This is a highly tendentious statement that suggests that many high level setons players behave badly and dont deserve their rating. Try beating good players on setons before you denounce their skill, its hard and takes hundreds of games of practice. If these players chose to play teamgames or ladder for 1000 games they would be equally as good.

I dont agree that we need to get rid of global rating because "it’s a high score chart for who can get away with the most garbage before a mod takes interest in their replays"
You are referring to 1-2 people, for example Suzuji. Another example of extremely biased public shaming. Suzuji is a fucking god on dual gap. He played this map thousands of times and knows every single BO, timing and has great "reading" abilities. Try playing against the 2500 rated astro crater guy, similiarly he knows the map well, knows timings and BOs. Surely its not a great benchmark of overall skilllevel, but nonetheless these people "worked" hard for their respective rating.
Suzuji maybe manipulated 1-2 games for the lulz, but mostly deserves his gap rating. Since he cant play up to his rating on other maps he only plays gap, so whats the problem? He is only seen in such lobbies, and the few people on faf (yes, its only a few in general) - they know who he is.

Of course I see the general argument that global rating isnt the best measurement for overall faf skill, i.e. the ability to win on any map, and with any amount of players per game. but ladder rating isnt much better!

Blodir, thomas hiatt, very strong ladder players will lose regularily on setons, dual gap, astro crater, phantom or even isis to regulars on these maps. All these maps are popular and legitimate game modes. Why? Because they have no clue about these maps! Therefore they are inferior players, they arnt experienced, ie. "skilled" enough to beat these players, they arnt good enough to beat them on these specific maps.

Its like taking the game of motorsport. A formula one driver like michael schumacher was considered to be a star, "the best driver". But he would be hopeless against a rally driver, a drag driver, a hillclimb driver or a street racer because thats just not what he practiced alot. Is one better or more important than the other? I wouldnt say so.

You are looking for a specific type of faf skill rather than general skill, and such skill doesnt exist. You cant be good at any game mode, because the day has only 24 hours and nobody can practice all the maps and game modes.

Its funny how people look down at Nexus and call him the setons pro. He chose to play alot of setons because he likes it and without a doubt rose to the best setons player there is. At the same time it means that he couldnt play ladder at the same time and some other people practiced it more and became better. If he chose to practice ladder he would probably beat them, but he chose something else. And then they laugh when he doesnt win a 1v1 tourney because he is king of the "highscore board", but at the same time they dont want to see how he just set different priorities.

I think global rating is a good measure. You need to be good at something, in fact you need to be comparatively better than most people at something, in order to get a high rating and thats quite hard to cheat. Its in my view the best measure we currently have, so why abandon it?

Ladder isnt appealing to many people, for most people faf is appealing due to the TEAM ASPECT, and ladder rating doesnt measure in any way how someone synergises with others towards a common objective. Look at foley, he has 2400 rating because he knows how to win despite having a comparativly weaker team. Thats a skill that no ladder elitist has, and thats a skill in RTS that deserves to be reflected in his overall skill assessment.

So whatever you plan to do, my recommendation would be to you as the PC or the coucil in general to not give in to the ladder 1v1 elitists, preserve the ratings nature to reflect a persons abilitiy to make a group of people win against another group and be less tendentious in your view of what constitutes the ideal faf player.

[Geosearchef]

Showing only the highest team matchmaker rating is a bad idea as it throws 1v1/2v2 and team games into one display. I also don't know what to think of one rating per team size.

About having disclrd channels for team matchmaking, I don't know, we could also implement a VoIP service into FAF. I've heard of a protocol which was developed for establishing exactly that type of p2p voice connections called ICE, we'd just have to integrate that into FAF and add some controls to the client / game UI.

[---]

"TrueSkill is patented, and the name is trademarked, so it is limited to Microsoft projects and commercial projects that obtain a license to use the algorithm."
https://en.wikipedia.org/wiki/TrueSkill

Could there be any legal issues with this system?

[tatsu]

@--- there's a thread on this on reddit : https://www.reddit.com/r/gamedev/commen ... _us_indie/

they basically conclude that A. math isn't patentable, that B. whatever implementation of TrueSkill you come up with will be at least somewhat different than their implementation, and that C. maybe there is more or less an issue with using the name

it's a good question though, maybe we should indeed call it something else. it's probably fair as, from what I've heard, our implementation is quite different.

@FtXCommando comming from a sentonner who, I guess, should maybe feel threatened by the reduction in importance or dissapearance of general matchmaking, I have to say I agree 100% with everything that you said.

Maybe the morax tourney was read into a bit too much, if anything the sample size was way too small to draw any conclusions.

but anyways I really think lately we have been suffering as a community from bad maps hosting without issue and good maps never getting used past their release day and especially not if they don't have a news page about them.

Team matchmaking is going to revitalise all that and give all those maps made with sweat, blood and tears a chance.

Ideally the map wouldn't be a choice of the player and voting a map in or out of the rotation would be a privelege earned with rating.

The only thing I fear is map rotation for team matchmaking being 100% out of the hands of the playerbase. The FAF client ui must provide at leat one mechanism to allow players to influence it.

and another thing : something that could spur engagement is having a "map of the week" button. this would immediatley throw you into a team matchmaking of that map.

I guess in that case the number 2v2/3v3/4v4 would either be determined by the map itself, or if we're talking about a map natuarally flexible to all of those or unaturally flexible (cookie's ultimate editions), then you'd be prompted with your preference (2v2/3v3/4v4/all) beforehand and promptly plopped into the matching matchmacker of that setting and for that map.

this would be great because it would be a go-to for people feeling disgruntled because of team matchmaking ousting general : they could thanks to this option know ahead of time which map they'd be in and still be able to do a moderate ammount of BO-whoring.

also the maps for "normal" team matchmaking queue in 2v2 / 3v3 / 4v4 should be few... 4 to 6 and change every week also.

this is the kind of solution that worked great for League of Legends ARAM : consent that you won't have your favorite character but you'll get to ban your least favorites to face beforehand and you'll experience a blast changing things up.

[EcoNoob]

- I also plan on having your global rating replaced in lobby with your highest matchmaker rating if that matchmaker rating is higher than your current global. Global games will still only affect your global rating and the matchmaker rating will appear as a different color in the UI so you can tell whether a rating is global or not.

I agree with everything except for this one. I'm not sure what problem this is trying to solve. Would the game quality/optimal balance use the MM rating (which one if multiple) or will it use the 'global' rating but show the MM rating in the lobby?

[Dro]

I have little reason to believe that a premade team of 2 will beat a random team of 2 on random maps judging by the results of the morax 2v2 tourney that was recently hosted. That tourney provided an excellent case study because the (arguably) two most well-known “premade” 2v2 teams of petric/bh and blodir/nexus that have played together for years had lost to jagged/farm.

Firstly, how many games recently have Blodir and Nexus played and how many games recently have Blackheart and Petric played. I would argue Farms and Jagged have similar numbers on that front and they both have way more experience with current balance and, to add to that they were also on comms and they still nearly lost to a lot of inactive players who just knew each others playstyle. The reason team match maker will be annoying with pure premades is possibility of a vast amount of spamming games on a limited number of maps that could occur. If people spam an insane amount of games with one person, obviously that team will have an advantage against a random premade with voice, playstyle ect. and since you will know exactly how your team mate plays you can tailor your bo very specifically, e.g you know he walks his acu at x or you know he goes second air here or you know he goes intie first here. Also you didn't talk about inactivity at the highest levels. If you limit the people to only being able to play when 2 people are online you severely limit the amount of total games that can be played. Since people won't want to queue with some random when they could just easily wait and not lose any rating. This I feel is going to create a similar scenario as people who only play dual gap and don't play anything other because they want to maintain their rating which leads to fewer games overall and, I think to some degree, stifles activity. I agree with what you said also about creating uncertainty on an individual level in rating and i'm glad you're no longer maintaining tWO pEopLe is exACtly the samE as oNe Do yoU EveN uNDERSTaND TRuE Skill nonsense... Full share makes sense at least in 2v2 but anything higher i feel the games are gonna be hours long, multi queue makes a lot of sense also perhaps showing the queue times (might be an awful idea though maybe after playerbase is established it can be considered so it doesn't just get DOA), your vendetta against global rating i feel will only harm the community, however. The point of the team match maker is to show the ratings within that current system. Just using it to completely overturn global might rub some people the wrong way. If people really cared enough they could look up everyone's match making ratings anyway so using it to blanket over global is just another way to piss people off in a sort of "you play setons/dual gap astro... Well f*** you your rating doesn't matter!".

[Blodir]

a really long post

I feel like most of what you said just isn't true, but I symphatize with you in that the "it’s a high score chart for who can get away with the most garbage before a mod takes interest in their replays" really is unfair. Top sentoneers are definitely good at what they do and they should be allowed to be proud of their achievements.

"ladder rating isnt much better!" is it not? Since you brought it up let's look at thomas and myself, thomas has 2080 ladder rating currently. I think it's fair to say that on any teamgame thomas plays at least at 1800 level with only possible exception being senton. That gives us a 280 rating difference between the high and low boundary of possible ratings. My ladder rating is 2341 and I would say that I play at least 2100 on every map (although this is a bit unfair comparison because i have a history of playing all maps). That gives us a 241 difference. As confirmation you may also notice that both Thomas and I are well above the lower boundary in global rating. If you look at ladder players in general you may notice a trend that the difference between their ladder and global is rarely larger than 300. This trend doesn't exist for non-ladder players.

Now let's look at other players you mentioned: Suzuji and Foley. Suzuji is at 2943 global rating and 1155 ladder rating, although he hasn't played very many ladder games so this is unlikely to be accurate, I think it's fair to say that he is below 1500 ladder. That's a 1443 rating difference. Foley is 2399 global rating and 1556 ladder rating. That makes for 843 rating difference.

Conclusion: ladder rating is much better indeed.

Btw since you mentioned gap pros and astro pros, I have played (and watched) both and to put it bluntly they were really bad... The quality of play on these maps is just not anywhere close to ladder or sentons.