I would argue that the ideal outcome is to have fun. Phrased from a system design perspective: that the largest number of players possible experience games as often as possible which convey a sense of fun.
Winning is generally fun, but the interesting bit about using Fun as a guiding light in matchmaking design is that its a non-zero-sum game. Losing teams can also have fun. Maybe not as often, depending on the game and such, but its definitely a number larger than 50%.
I know this seems reductive and whatever, but I think its a really critical discussion that the gaming industry isn't having. Skill-based matchmaking and ELO are pretty tuned-in to, well, "skill" (whatever that means in modern gaming), such that you're pinned at that 50% rate to the best of the system's ability. Some would argue that this pinning actually results in less fun games, for two extremely valid reasons:
1) If the system is constantly adjusting to keep your win rate at 50%, you never have the opportunity to see yourself improve through the metric of your win rate. This effect is especially pronounced in games like Call of Duty, where the metric we're talking about is really something more like your KDR; if the system is engineered to constantly keep your KDR at 1.0, there's startlingly few metrics to judge your own performance by. Many games have tried to address this through the award of opaque ranks and badges; weirdly, Call of Duty doesn't, but some games do. In other words, your ability to feel like you're getting better is replaced by being told that you're getting better; many would agree that while it can feel good in the moment to get that badge, its less satisfying in the long-run.
2) A perfectly balanced skill-based match making system would reasonably want to keep every player's win-rate at 50%, over a long period of time (or, KDR at 1.0, or in Apex a win rate of 5%, or whatever is balanced for that game). But that isn't necessarily what happens over short periods of time; and players experience games in the moment, players aren't statistical databases of their entire gaming history with some title. Thus its normal to ask not just the win rate, but the standard deviation of that win rate. People play the lottery, and find it fun, for a very real psychological reason: fun can have an outsized impact relative to the odds of some game if the game has an abnormally high/unfair standard deviation of expected success. If you paid $1 for a lottery ticket, and you were guaranteed to the best of the system's ability to win $2 every 2 lottery tickets, no one would have fun playing the lottery (and the system wouldn't make any money, but that's beside the point). Yet; this is how SBMM is quite literally designed; to reduce the standard deviation of your performance as much as possible. Variety is the spice of life, as they say; but the better and more consistent SBMM systems get, the less spice the matches you play tend to have. Put another way; "popping off" for a game is supposed to be a fun, rewarding experience; but players are now universally trained to know that the system will actually punish you for being abnormally good in a match, by placing you against better players next game.
My overall point being: there's a gulf of difference between "players have to have an average aggregate win rate of 50%" and "all players should have an individual win rate of 50%", and systems which try to guarantee the latter can actually be less fun.
I also find it interesting to think about an ELO system which tries to operationalize the concept of Fun, to try and guarantee a Fun game for as many players as possible. What would that look like? Most simplistically, and probably ineffectually: ask players after the match "did you have fun?" A player's performance in the game influences their own skill rating; but a player's response to this question would instead influence every other player's "fun rating"; asserting whether or not the competitors (and teammates) were fun to play with; as well as signaling internal metrics to the development team for tuning of things like map, weapon, and game design.
That fun rating is a great idea, and something I was thinking of as well. The data science folks could have some fun figuring out what to do with it.
But for my own psychological health, just telling the matchmaker how I felt about the match would go a long way to taking the bite off a bad experience.
HoTs used to do that when I played. You would sometimes get a survey asking you to rank the cooperation of teammates, communication etc. Looking back I sure felt a lot less salty after a loss when I got one of those surveys. Unfortunately Blizzard did absolutely nothing with it.
Another lost opportunity due to a corporation being more concerned about the next quarterly reports than anything else..
HoTs failed miserably at both the fun part and the skill based match making.
Sometimes I had more fun losing but those were exceedingly rare cases that generally involved a friend or two in the group.
I still played the game chasing those rare highs where I actually enjoyed the experience. After losing 30something games in row spread out over a week of nightly playing I just rage quit. Even today years later the game's MM still pisses me off. Mostly because I found the game play mechanics to be fun and I miss the experience of the good games.
Winning is generally fun, but the interesting bit about using Fun as a guiding light in matchmaking design is that its a non-zero-sum game. Losing teams can also have fun. Maybe not as often, depending on the game and such, but its definitely a number larger than 50%.
I know this seems reductive and whatever, but I think its a really critical discussion that the gaming industry isn't having. Skill-based matchmaking and ELO are pretty tuned-in to, well, "skill" (whatever that means in modern gaming), such that you're pinned at that 50% rate to the best of the system's ability. Some would argue that this pinning actually results in less fun games, for two extremely valid reasons:
1) If the system is constantly adjusting to keep your win rate at 50%, you never have the opportunity to see yourself improve through the metric of your win rate. This effect is especially pronounced in games like Call of Duty, where the metric we're talking about is really something more like your KDR; if the system is engineered to constantly keep your KDR at 1.0, there's startlingly few metrics to judge your own performance by. Many games have tried to address this through the award of opaque ranks and badges; weirdly, Call of Duty doesn't, but some games do. In other words, your ability to feel like you're getting better is replaced by being told that you're getting better; many would agree that while it can feel good in the moment to get that badge, its less satisfying in the long-run.
2) A perfectly balanced skill-based match making system would reasonably want to keep every player's win-rate at 50%, over a long period of time (or, KDR at 1.0, or in Apex a win rate of 5%, or whatever is balanced for that game). But that isn't necessarily what happens over short periods of time; and players experience games in the moment, players aren't statistical databases of their entire gaming history with some title. Thus its normal to ask not just the win rate, but the standard deviation of that win rate. People play the lottery, and find it fun, for a very real psychological reason: fun can have an outsized impact relative to the odds of some game if the game has an abnormally high/unfair standard deviation of expected success. If you paid $1 for a lottery ticket, and you were guaranteed to the best of the system's ability to win $2 every 2 lottery tickets, no one would have fun playing the lottery (and the system wouldn't make any money, but that's beside the point). Yet; this is how SBMM is quite literally designed; to reduce the standard deviation of your performance as much as possible. Variety is the spice of life, as they say; but the better and more consistent SBMM systems get, the less spice the matches you play tend to have. Put another way; "popping off" for a game is supposed to be a fun, rewarding experience; but players are now universally trained to know that the system will actually punish you for being abnormally good in a match, by placing you against better players next game.
My overall point being: there's a gulf of difference between "players have to have an average aggregate win rate of 50%" and "all players should have an individual win rate of 50%", and systems which try to guarantee the latter can actually be less fun.
I also find it interesting to think about an ELO system which tries to operationalize the concept of Fun, to try and guarantee a Fun game for as many players as possible. What would that look like? Most simplistically, and probably ineffectually: ask players after the match "did you have fun?" A player's performance in the game influences their own skill rating; but a player's response to this question would instead influence every other player's "fun rating"; asserting whether or not the competitors (and teammates) were fun to play with; as well as signaling internal metrics to the development team for tuning of things like map, weapon, and game design.