How exactly do we expect low rated players to play the game?
-
Yes I think my implicit point was more options for smaller games would be better.
-
@nex said in How exactly do we expect low rated players to play the game?:
@clyf yeah, the rating system isn't designed to handle large teams
On the contrary
https://www.microsoft.com/en-us/research/wp-content/uploads/2006/01/TR-2006-80.pdf -
@blodir said in How exactly do we expect low rated players to play the game?:
On a related note, I've been playing ladder in a different game recently where the matchmaking range is far too broad. I have 73 wins and 15 losses. It is not fun for me that the majority of my games are essentially meaningless stomps. I don't think it's fun for the opponents either.
It's been my assumption that, when a player queues for a game in FaF, the system begins by trying to find a player near their rating,
but as they queue for longer and longer, it 'expands' the range of ratings it's looking for.Is this not the case?
It seems like the most elegant solution, where people will get games against equally-skilled players, but if they're really really desperate for a game, they can wait for longer and fight someone outside their skill level.When I started out (not long ago!) I WANTED to fight stronger players, and would queue for ages in the hope that a strong player had also been queueing for a long time. I also understand people not wanting to play people far outside their rating, at which point cancelling then re-joining the queue seems like a fine solution, giving control to the users.
I guess another option is allowing a player to set a range for which players they are happy to queue against - although it could allow for rating manipulation if the player got too much control over it, so maybe increments of 100 would be more suitable?
@nex said in How exactly do we expect low rated players to play the game?:
so it takes a lot more games to accurately reflect a players skill when they play 8v8 than if they played 1v1.
This will always be the case. There's no impersonal (/automatic) system that can make a 4v4 'zero in' on a players real skill any less than 4 times slower than 1v1s.
That's just the nature of team games, in that players can 'lean' on allies in 3/4 of the games they play.
I played tons of a game called 'battlezone 2', whose community was miles smaller than FaF. In that game, everybody basically knew one another (you'd have maybe 20 total players online at peak times), and could balance teams based on knowledge of the players. When new players arrived, their skill would be very quickly assessed by players - but this doesn't really work with any 'automatic' (maths-based) systems that I'm aware of.
So basically, if your community is large enough to necessitate ratings as a 'skill currency', you're kinda forced to have skill take ages to sort out when playing large team games.
(A similar issue occurs with the very popular 'MOBA' games - players can play for months or years without every really getting an accurate skill, because of the randomness introduced by only controlling 10% of the players in a match, rather than 50% in a 1v1). This is a mostly the reason I prefer playing 1v1 games! -
^The issue is that team performance in FAF is less like a sum and more like a "who is the weakest or strongest link in a critical position".
*Edit: at lower levels
-
@blodir I have missphrased my point, which is it is less efficient at evaluating players that play in teams. Whether it does this well compared to other rating systems, I can't say.
So while they tried to make it work for teams, it still suffers compared to 1v1 games.@clyf said in How exactly do we expect low rated players to play the game?:
The issue is that team performance in FAF is less like a sum and more like a "who is the weakest or strongest link in a critical position".
that is just a superficial issue, as that is the case for both teams and will average out over multiple games, so it doesn't impact the rating system a lot.
-
@clyf said in How exactly do we expect low rated players to play the game?:
The issue is that team performance in FAF is less like a sum and more like a "who is the weakest or strongest link in a critical position
Usually strongest, yes?
The fact that when a weak player gets knocked out, all their stuff get's given to stronger players, seems like it would make the team games lean more towards 'which team has the strongest player?' than 'which team has the weakest player?'
I don't play many team games, but I was led to believe that some games 'turned around' when the 'bad' player got knocked out, HELPING the team that lost an ACU!
-
@redx said in How exactly do we expect low rated players to play the game?:
How are people going to get better when they're lucky to get 1 or 2 decent games in a night?
Your mistake is assuming that you get better by playing matches, or by playing matches with higher-rated players.
The truth is: you get better either with coaching from a better player, or by being your own coach. Which means understanding why you lost a match (it's usually not "unit composition" or "strategy," usually it is ECONOMY, raiding and being raided, expanding, gathering and spending resources). It means accepting responsibility for every loss as your own fault and your own lack of skill. It means watching replays and being critical of bad decisions you made and opportunities you missed.
Unfortunately, just hanging out with higher rated players doesn't transfer skill to you by osmosis. Usually you just end up being carried by them, they scout for you, they overflow energy to you, they beat the other team for you, you're just going along for the ride.
If you really want to get better, you could start by reading my guide. (Not the whole thing--just read until you've absorbed a few new ideas, practice them, then you can go back and read more of it.) If you really want to get better, it means playing lots of 1v1 matches and getting advice from higher-rated players (ask them to watch a replay and tell you what you should do better). You have to git gud at 1v1 before you can be good in team games. Even just trying to get in to team games means you're not understanding how to get better.
By the way, no one is obligated to get better. Some people might be different and learn in different ways. But I think this is the only viable path to improving rating for 95% of FAF players.
-
If the model for team performance doesn't reflect individual performance, you're just averaging out noise. Consider team slayer: your individual performance is far more impactful (with your kills and deaths being reflected directly in the game deciding score) than two ~400 ranked players sitting in the back of an otherwise 1000-1500 ranked match.
it is less efficient at evaluating players that play in teams
What's the basis for you believing this if you don't agree with the above?
all their stuff get's given to stronger players
Heavily dependent on share mode, yes. And yes, usually it's the stronger player that has the ability to finish the game out that plays a critical role.
-
Wow lots of discussion about score and rating here.
An important thing that makes FAF different from a shooter game (on which much of this score research has been done) is that the map - and the finite resources on the map - matter more in FAF. Everybody knows that if you kill the 900 player in a 900/1500/1900 team fighting a 1400/1400/1500 team, that you are doing that team a favor, particularly in full share, as the 1900 player will make far-greater use of the limited resources. I've seen at least one game in which the lower rated player gifts the higher rater player his base and hides his ACU in water for this exact reason. This is why "Share Until Death" became meta in the first place. (In unmodded SupCom, there were no restrictions on share... either you gifted your base manually when you died, and your team got it, or you didn't do this and they didn't.)
-
@funkoff As a somewhat new player 'looking in', the current 'share until death' mechanics seem really, really problematic!
I can see how high-rated players like it, but honestly it seems like a disaster for matchmaking.(I have similar 'noobie' misgivings about being able to gift engineers to allies... I guess at least that makes faction balancing easy, since they're all basically the same! :D)
-
If you think it’s problematic or a disaster for matchmaking you haven’t realized
- how much of an issue d/c’s can be in faf
- how little map variation is allowed to make share until death remotely playable
- How one note the tactics are
T2 air is considered OP in full share where killing any ACU doesn’t carry an innate 20k of mass killed in infrastructure at min 10.
-
@clyf said in How exactly do we expect low rated players to play the game?:
with your kills and deaths being reflected directly in the game deciding score
Without knowing much about professional Halo(that's what team slayer is reffering to?) I can tell you that is wrong. The only relevant metric is you won, rating players based on their kills turns the team game into a FFA where there are some people you can't shoot.
@clyf said in How exactly do we expect low rated players to play the game?:
performance is far more impactful [...] than two ~400 ranked players sitting in the back of an otherwise 1000-1500 ranked match.
Impact on the game ofc differs by rating and position, but over several games there will be bad players in impactful positions in both teams.
@clyf said in How exactly do we expect low rated players to play the game?:
If the model for team performance doesn't reflect individual performance
ofc the outcome (win or loose) depends 100% of the performance of each player, so the metric measures to a certain degree the performance of each player no matter how much they contributed, but as the number of players go up the noise (as you also mention) goes up, but over several games that noise gets evened out by putting that player in different teams.
So if team A wins, you know that team is (probably) better, but you have no idea if player 1 was better than 2 or 3 was better than 4, to find that out you need more games with different permutations of the players -
@blackyps said in How exactly do we expect low rated players to play the game?:
@redx said in How exactly do we expect low rated players to play the game?:
How are people going to get better when they're lucky to get 1 or 2 decent games in a night? I understand not wanting to play with terrible players, but we also need to be realistic about how many players are actually on at a given time and what's realistic to expect if want a health playerbase.
Low rated people are be the biggest rating bracket. I don't know why we don't see loads and loads of noob lobbies. Maybe these people don't feel confident to host a lobby. Maybe they all play in the matchmaker instead, after all that was a big reason to create it. But you already said that the matchmaker is dead in your timezone, so honestly I have no idea where the low rated players in your timezone are hiding. But they must be somewhere.
I'm not sure where all the low rated players go either. Even in all welcome lobbies it's usually a bunch of 1k-1600 players and a couple 800s or so. Maybe they all hide in gap, I tend to just ignore gap lobbies altogether.
Side note, can we keep the full share debate to another thread and keep from sidetracking this one too far?
-
"professional" Halo
team slayer
I can tell you that is wrongWell, good that you A. don't have the faintest idea that I'm talking about B. couldn't take 30 seconds to google it and C. were still confident enough to tell me that I'm wrong!
-
I'd imagine this "lost" group of low rated players are just fighting AI. Stress free, toxicity free, etc. Venturing into playing against other people is a big step.
-
@clyf I googled
that's how I got that "team slayer" is some halo gamemode where 2 teams compete to get 50 kills first and it fit into the rest of your statement, so I assumed that's what you meant.
But it doesn't matter which game this is about. 4v4 FAF is about the first team to reach 4 ACU kills, but rating players based on how many ACUs they killed would be stupid. -
but rating players based on how many ACUs they killed would be stupid
Exactly! Why does rating by number of kills make perfect sense for team slayer, but would be stupid for FAF?
-
@clyf said in How exactly do we expect low rated players to play the game?:
Why does rating by number of kills make perfect sense for team slayer
It doesn't because it disregards players actually playing in a team.
what if you have some supportive role that allows your team to get more kills?
What if you secure important map positions that hold good weapons?
Just by rating players by this metric you change the game from a team game into a single player target shooting game, where your "teammates" are your opponents and your "opponents" are just targets you are competing to shoot. That's where a lot of toxicity in online teamgames comes from (see league of legends and the term "killsteal").
Players never play what you believe your game is, they play what the rating system tells them what the game is. So if your rating system tells them the more kills you get the better, then they will do everything to get more kills themselves and won't do anything else that could help their team. some would even start shooting teammates if that didn't have penalties, just to get more kills to themselves.
Consider hurdling: The runners don't avoid the obstacles because the games designer envisioned it so, but because there's a penalty on their rating when they don't.The rating system is the actual game in which you compete. The "game" is just the means to do so.
-
what if you have some supportive role that allows your team to get more kills?
What if you secure important map positions that hold good weapons?My brother, the name of the gamemode, the gamemode that TrueSkill was invented to judge, is called slayer. It is judged by the number of kills you get, and how few kills you give up to the enemy team.
The good weapons allow you to get more kills. Not a ton of support interactions in team slayer (CTF, oddball, another story). Definitely different in other shooters, but in Halo you support your team by helping to kill your opponents before they can kill your teammates.
TrueSkill judges team performance as a sum of individual performance because, in a game where you get points by killing the other team and the other team gets points by killing you, team performance literally just is a sum of individual kills and deaths in a way that doesn't relate to an RTS where map texture is much more important.
But we're getting way off track here. Why do you think TrueSkill is not as efficient at evaluating players on teams if you don't agree with any of the above?
-
@clyf like I said above since you only have data on the performance of the team not the individuals, since there is no 100% correct algorithmic way to tell how much someone contributed to the win. So to rate a single player accurately you need them to play with different teams.
So to find the correct rating for a player in team games, that player will need to play more games than they would need to if they only played 1v1.
