Team Matchmaker Info&Discussion

[Blodir]

About premade vs random 2v2:

I don't think any existing premade has a very large advantage over random teams by the merit of them being a premade. BH/Petry & Nexus/me are just teams where both players are at the very top of ladder and so it's not surprising that they are powerful teams. However I do have slight concerns. The possibility of premades encourages superteams like this, which will lead to either them never finding a game or that game being a total stomp. If it was purely random, that would undoubtably make for more balanced games and there would be no temptation for high rating players to team up. While I'm talking about people in general I basically just mean that this is how it might affect myself. I might end up playing only premade and find less games as a consequence.

So ye I'm on the fence about having only random teams or both randoms and premades

[armacham01]

I think we will see a significant difference between premade and random. People who take the time to team up are going to discuss strategies in advance and are more likely to be on voice chat. Two 1500s who are paired up at random (who might not even speak the same language) are probably not going to work as well together as two 1400s from the same clan, on voice chat.

Also re: Farms + Jagged, I believe that they watch each other's streams from time to time, and they have played against each other quite a bit. So each had a lot of familiarity with the other, even if they were not used to teaming up together for 2v2s.

The most important thing is to keep statistics, so we can measure it, and make changes in the future accordingly. For now, it is fine to treat premade matchups as the same as random, as long as the data is being tracked so we can answer the question definitively. Right now the problem is "we don't have a 2v2 matchmaker system live." The problem is not "we need to make sure that the 2v2 matchmaker system deploys as smoothly as possible with nobody getting upset over rating issues." It's not going to be a terrible disaster if, for the first 3 months of 2v2 ladder games, the rating system is kind of messed up. People will still be able to have fun games while the issues are worked out. There is something to be said for "just implement it and see what happens" rather than "try very very hard to predict in advance what will happen and address all of the problems before they arise"

I know there is some ratings manipulation going on with the global rating, but it is still a very useful tool for balancing games. Think of where we would be if we DIDN'T have that. In theory, FAF could track separate ratings based on certain maps (for example: a crater/gap-type map rating, a seton's-variant map rating, a "1v1 custom match" rating, and an "everything else global rating"). To avoid overwhelming the players, FAF could keep these ratings more or less secret but still use them behind the scenes when players ask FAF to balance games for them. By not showing people the four sub-ratings it would make it a little harder to manipulate them. And we could try to scientifically measure whether someone really is better at a certain type of map. That would be my solution IF we needed to fix the global rating. At this point, I think global rating is "good enough" for most of what we want to do, so it probably isn't worth it to try to fix. There are better ways to invest resources, including setting up the 2v2 matchmaker (and then 3v3, etc.). Eventually it may come to pass that fixing the global rating becomes worth investing in.

I would even be interested in a 3v2 matchmaker or 4v2 matchmaker. If you have 4 1200s, and 2 1500s, maybe on some maps it is "fair enough" to set up a 2v4. But that is something to look into in the future. Obviously we get 2v2, 3v3, and 4v4 up before we look at asymmetrical matchups (and even: asymmetrical maps, eventually--with the neroxis generator that might be possible to have balanced-enough asymmetrical maps that people would play 1 time only, or generate 1 map every 8 hours, and keep only the 3 newest generated maps in the ladder pool, so there is enough time for some sort of meta to develop but if the map isn't super-balanced, it won't ruin FAF forever). It is only fitting that the absolute best RTS of all time, with its over-the-top "track every projectile" approach, would have an over-the-top awesome map generator for random maps to keep things fresh and over-the-top flexible matchmaking system that could set up balanced games with odd numbers of players. With enough testing and data collection over time, I think it will be very possible to have a robust, fair asymmetrical matchmaker.

[FtXCommando]

Premade Comments:
Okay so I do agree that the farms/jagged tourney situation isn’t exactly similar to the situation that a random 600/800 would have against a 700/700. The big reason why would be language issues and an inability to be in voice together. As Geo mentioned already, there has been some attempts at integrating a voice chat service within the java client in order to alleviate the latter issue somewhat. The language dilemma can’t exactly be solved, however.

Yet I do not think it would lead to a productive environment for TMM if you force people that cannot properly communicate with one another to play with one another. While a premade team that can both speak English/German/Russian will have an advantage over a team that does not have that ability, I do not think dragging everyone down to the level where they could have a teammate they cannot communicate with is a good idea. Again, providing the user with a degree of control over the matchmaker experience is necessary to create a better environment overall here.

I would not be averse to utilizing the FAF Discord similarly to the way that say the PUBG discord works where you have people in a #looking-4-game channel where they post what matchmaker they want to play, their language/country, and perhaps a rating they would want their ally to be. I think that would lead to a great way for people to form bonds in the community and it’s how I often made friends in other communities that utilize a matchmaking service.

I really do disagree with a lot of the analysis about the jagged/farm team. It seems like monday-morning-quarterbacking (read: analysis in hindsight) about the tournament results. Jagged/farm was definitely not a favorite to win this tourney compared to bh/petric or blodir/nexus. But again, I do concede there are some aspects that do make the team ‘kinda premade kinda not’ here. Also petric/bh have been extremely active recently as they basically play as often as farm so I disagree with calling them inactive. And even when they actually WERE inactive they were going like 18-0 in 2v2s last year.

Global Rating:
I worded it the way I did because it’s impossible to quantify the precise degree of manipulation within the system. Taking the most noteworthy examples is only looking at one aspect of the problem. There are people that farm twin rivers to 1900, there are people that can only play wonder open (on one slot!) at 1700 level, there are people that can basically only play isis at 1800 level, etc. Am I upset at these people playing FAF and doing what they enjoy? Of course not. If I was, I would want to delete global rating entirely.

However, I do not consider these games to be worth being noted as competitive outside of their subcommunity. If those within the subcommunity wish to continue playing in their own ecosystem they are certainly welcome to do so by balancing with the global rating that would be used with the lobby. If they dislike using TMM ratings to balance games then they can easily see whether someone’s rating in lobby is a TMM rating or not. They could then kick those with a TMM rating if they desire to. I would always see TMM ratings as more accurate than global rating regardless barring the noteworthy example of highlevel sentons where the same 12 dudes are playing one another.

If you think global is the best system we have, well, that’s simply not based in objective reality. Maybe that’s true for individuals that will never, ever, in 100 years on FAF play a game other than what they have decided to play. It is an absolutely toxic system for players that wish to branch out and it is those players that desire to branch out that eventually become the foundation of the high level competitive scene. I define the high level competitive scene as the group of players that are willing to regularly play in tournaments because that’s the best current metric now.

In regards to EcoNoob’s question, my goal was to prevent as many barriers of entry as possible within the community. I see TMM as more accurate than a global rating gained in twin rivers/wonder that is then transferred over to sentons and so I figured it would not harm the general environment of custom games more than it already is. I do understand the issue of people not losing rating for actually losing games if their TMM rating is higher but I do not really know how I would prevent the problem. If people can play 100 games in TMM and then get kicked from custom games because they have zero rating, well, we’re back in this subcommunity divide that confuses new players and harms player retention. Basically, it’s an area where I decided to go with assisting the new player experience over the high level experience.

Geo:
I’d rather discuss this one privately as it relates to another concept being worked on. I expect to treat this system of replacing global with other trueskill values to merely be a placeholder in the longterm. I would be willing to only use it for 2v2+ ratings, however as custom games do not see a lot of 1v1 action.

Choosing Maps:
This is mostly in response to tatsu. The system is sound and I would love to have a choice pool concept similar to what I’ve elaborated on in the ladder thread. However, until then, I’m likely going to run with a system where I choose maps similar to the current ladder system. I haven’t put an extreme degree of thought into it just yet but rest assured I would love to at least give players a veto capability.

This is a dilemma I will go more into depth as we get closer to pushing the concept, unfortunately not a lot of details that I’ve ironed out but thanks for the ideas.

Asymmetrical Teams:
Uh, not for a long time at least. The maps that allow for asymmetry to be balanced need to be built significantly different to traditional FAF maps as an extra ACU is a ridiculous advantage in the early game. It practically immediately restricts you to only 20x20 maps by default and the map becomes more turtle favored as you increase the player quantity imbalance. If it would ever be entertained, it would be very far down the line and honestly it truly seems more of a casual game experience rather than a competitive one in my eyes.

[FtXCommando]

Thanks for all the posts, I wasn’t expecting so much to read after a day.

The major two things I’m currently wondering about is the feasibility of implementing a sort of rating restriction for your ally so that a 2000 rated player doesn’t match with an 800 for 3/4 of their random games. However, as arma said above, I really can’t tell how much of a problem this would be until the matchmaker system is actually implemented. It is a tweak to keep in mind, though. I’m still unsure if random functioning as a way “to get a game, any game” and premade as a way to control your type of game is necessarily bad for FAF.

The other thing I’m thinking about is a regional divide (as FAF already uses a service to locate you by your IP) but again, not sure how damaging this could be to the matchmaking process.

[Bennis-]

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.

Because you dont look beyond apm, unit formation and the like. It takes more to actually win on dual gap with 8 mexes against 24 other mexes, and suzuji can pull it off, when his other 20 mexes are controlled by total baboons. Its about understanding whats important to do: air defense, nuke defense, spotting their anti nukes, having a feeling for their anti nuke % etc.

I mean I do see that these people suck at 1v1, that these people have not the slighest clue about inti micro, how to use an inti air advantage (gunships, drops, bombers), T2 unit mix and timings, I know. But I think that within their little gap ecosystem their rating is actually a useful benchmark to form a good balance for matches.

heck what are we even argueing here. Im just against the elitism. when I got 2k rating I got bored with setons and tried some Gap of Rohan, some dual gap, some astro crater and teamgames and realised: oh wow, these maps actually have a specific meta that I dont know and therefore even though I am, lets say, 1600 ladder, I cant really "carry" easily here. Sorry I dont have much to contribute to the 2v2 topic, I think you guys are unfortunately debating a subject that will never find enough people that play it. when we dont have an active ladder scene, I think its utterly dilusional to think that we can get 4 ladder players to chose to play AT 2v2 at the same time. faf has other construction sites that need to be tackled before, like recruiting new people and not banning the active ones.

[tatsu]

One point hasn't been argued all throughout this thread : lag.

lag is a non-issue in 1v1, heck in almost any 5x5 map or 10x10 map situation.

but with team matchmaking since you no longer have control over who is in the game it should be a factor.

one idea would be to use the player's calculated cpu score and under a certain score the matchmaker doesn't include 4v4 matches and 20x20 and above maps

above that score and anything goes. also the player clocked under that certain score does get to matchmake on 4v4 matches and 20x20 and above maps but only with players who are under that certain score as well.

under the hood the matchmaker will be doing this silently, no one's none the wiser and everyone's overall playing experience is improved.

how does that sound?

[biass]

cpu score

Probably a good idea, but might be an easy way to abuse.
Bring your friend playing on a toaster into a game and intentionally attempt to make the enemy team leave from frustration - especially if you cannot see cpu score before the game starts.

asymmetry

Probably too far an example of project creep.

I don't think that people are expecting the system to really function at an acceptable standard for a number of months after it's released - as you said yourself. But if in a blue moon everything is of an alright quality, FAF should be focused on encouraging people to play it. Divisions and the like. It's all well and good that you created something, but once the initial wonder and mystery of a new system wears off (or is sullied with a poor release) faf needs to give people a reason to stay or like bennis says, it will be dead.

The map generator might be too much a stretch for asymmetry. While it just kinda places down stuff whereever, as is the point, it's justified in the sense that the other player has the exact same random stuff as you. Balanced asymmetrical maps require a precision blend of terrain and resource layout that due to the amount of testing and planning, isn't something a map generator can provide. It even isn't worth the time for mappers despite the current demand because of the high skill (of the craft) required. I think FAF should leave the random nature to something like galactic war, which should seek to encourage emergent gameplay over fair competition.

However about maps: FAF is gonna need a lot of them clearly, and with "getting into ladder" being the only incentive for mappers, a lot of people neglect teamgame maps and focus on 1v1, leaving FAF at a serious deficit.

FAF should quickly implement some simple strategies to get content ready for the matchmaker post haste. It's way too late for me to cash in on this statement: but it's time to start rewarding people for their contribution, no?

bennis doesn't last 2 posts without mentioning his bans

ngmi

[Blodir]

Because you dont look beyond apm, unit formation and the like. It takes more to actually win on dual gap with 8 mexes against 24 other mexes, and suzuji can pull it off, when his other 20 mexes are controlled by total baboons. Its about understanding whats important to do: air defense, nuke defense, spotting their anti nukes, having a feeling for their anti nuke % etc.

Please stop putting words in my mouth / assuming my thoughts, it's really annoying.
you do realise when i started faf i played teamgames exclusively until 2300 rating right so what you are saying makes precisely zero sense

[Apofenas]

Says

...Our current custom game environment allows you to control several factors...

allows premade teams against random players who may not even speak same language... hmmmmmm

And exciting full share condition with premade teams! You just get yourself a low rated guy, teach him to push with t1 arty, do com drop or some other suicidal play. Enemy team would take some damage and you would take double eco. And if we extend TMM to 4v4s, the team may just take a low rated balance slave to simply leave the game letting better player take the eco.

Share untill death is default for a reason. It is a measure to prevent suicidal game plays (along with armor types for ACU explosion) when enemy team kills you but suffers damage, while your teammate gets extra base. It's a significant advantage if 2300 player throws 700 rated and gets double eco against a couple 1500s.

If Share until death encourages players to keep ACUs in their bases, the Full share encourages players to suicide. You just wait to see it. I can clearly see some ways to abuse the f*** out of this system and turn TMM rating into Global 2.0


Watching specific map rating discussion reminds me of this
When the flame from playing with crater player is so strong, that this energy transforms into an art!


Also i didn't see the Sixth dilema: Maps. You get more people in TMM and the chance someone hates some specific map grows accordinly. Any inside info on how this will work?

[FtXCommando]

I really don’t get why you’d bring that up when I already talked about how premade teams bring inaccuracy into the system. Making a purely random matchmaker will require increasingly complex code to curate teams and remove as many potential terrible scenarios as possible without creating situations where there are many people searching yet incapable of playing a game. We could wait 3 years for this code to not only be written and ironed out, but then tested on a large scale where we can be confident that it properly handles faf population sizes. OR we could just use premade teams and work the complexity in as we need it.

Share Until Death does absolutely zero to prevent the situation you describe. 0 rated dude makes a factory, donates it, then goes and acts as a glorified pd for his 2k ally against 1ks. Same exact suicidal gameplay you’re worried about. Beyond the first 2 minutes, the 0 rated dude is going to more than likely already inefficient compared to his ally so the longer you wait to donate (in share until death) or suicide (full share) the less payout this strategy has. Keep in mind you have no absolute control over your map so you can’t beat a BO into the head of your 0 rated balance slave. Now if you manage to incidentally train this dude to be able to last 10 minutes on every map, chances are he’s just been made into a 1k player. Now thanks for your desire to cheese, we’ve potentially got a new player that will stay with the community. Sadly your farming of him will only work for perhaps (10?) games and then you’ll need to go train another poor soul or get stomped by high rated players ready to farm your points.

Also, good luck finding a dude that wants to spend his evening just not playing on FAF. Good luck winning those games as you’re underestimating the difficulty of this situation. Anyway, you could already do the exact thing you’re describing in global as it is and yet no one does it.

The real issue here would be people abusing the initial deviation of 0 rated players and other new people. I’m unsure how widespread it would be and whether in the longterm trueskill would end up solving the issue through other games in the matchmaker.

Maps will likely work with me picking them until some sort of choice system can be made or neroxis map generator can be trusted to develop maps for the matchmaker.