Why is MMR system based only on win/loss?

[Stop! Stop! He's Already Dead!]

Is it really so hard to rank MMR by the performance of the individual player regardless of whether the team wins or loses, or at least factor it in somehow?

I don't understand why you can get all these stats after a match telling you how you ranked against the entire spectrum of dota players, and even more stats with Dota+ but all of this quantifiable data doesn't factor in at all into your MMR.

Playing with randoms is a crap-shoot much of the time. You can outperform everyone on your team and still lose for all kinds of reasons.

Why is MMR based solely on win/loss?

Here's an example, please Valve tell me how this fair and sensible:

I just finished a role-ranked game while in calibration. The two supports did not choose heroes, losing gold until they randomed. Then, the mid player disconnected for 5 minutes and got an abandon. I played safelane axe and outperformed everyone on my team, but we lost as 4 against 5.

I wasn't even allowed to quit the game after the mid officially dropped without taking it as a loss that affects my MMR calibration. I played it through anyway, but how ridiculous is it, that it doesn't even matter how well I did? Because all that matters is whether you win or lose.

[Pepega or Gigachad?]

It's not possible to measure performance. Performance is objective.

[Marcel]

Originally posted by ----:

It's not possible to measure performance. Performance is objective.

you mean subjective?

[Zockolade]

The Dota 2 matchmaking system is just the cheapest that is possible.

[Stop! Stop! He's Already Dead!]

Originally posted by ----:

It's not possible to measure performance. Performance is objective.

You mean subjective, but I respectfully disagree, there is a ton of quantifiable data available that could be factored in, how many last hits, denies, kills, deaths, assists, damage, and on and on. You could be last hitting at the same level of ancients in a herald game, kill the enemy 20 times and never die and still lose because one of your teammates is afk, but still the system doesn't care, all that matters is whether you win or lose. It's unacceptable that with all the data they have, and in the age of AI, they can't come up with a better and more fair system. The people who put this together are practically running a casino.

[Pepega or Gigachad?]

Originally posted by 4 practice:

Originally posted by ----:

It's not possible to measure performance. Performance is objective.

you mean subjective?

Thanks. Yes. Otherwise, all mobas will just use it instead of win-lose. There are many reasons like people will pursue stats over game objectives.

[HardRock]

Originally posted by Purrlock Holmes:

You mean subjective, but I respectfully disagree, there is a ton of quantifiable data available that could be factored in, how many last hits, denies, kills, deaths, assists, damage, and on and on. You could be last hitting at the same level of ancients in a herald game, kill the enemy 20 times and never die and still lose because one of your teammates is afk, but still the system doesn't care, all that matters is whether you win or lose. It's unacceptable that with all the data they have, and in the age of AI, they can't come up with a better and more fair system. The people who put this together are practically running a casino.

The goal of the game is to destroy enemy base, not to kill enemy heroes.

You can win a game dealing a total 0 damage to enemy heroes, if all you do is hitting towers. But if you only push lanes - you can still win the game, while dealing 0 damage to enemy duildings yourself, creeps will do all the work for you.

All those statistics like KDA or networth doesnt mean anything. The only measurable progress is building health. But since you can damage them indirectly, even this statistic will be inaccurate.

Ofcourse, its cool to have detection when someone is intentionally ruining the game. Like breaking items, staying afk or feeding, placing all wards at the same spot.

But other than that its hard to tell who made the most impact in game: carry can have high KDA, but he was stealing farm from all of his teammates.

[Stop! Stop! He's Already Dead!]

Originally posted by HardRock:

Originally posted by Purrlock Holmes:

You mean subjective, but I respectfully disagree, there is a ton of quantifiable data available that could be factored in, how many last hits, denies, kills, deaths, assists, damage, and on and on. You could be last hitting at the same level of ancients in a herald game, kill the enemy 20 times and never die and still lose because one of your teammates is afk, but still the system doesn't care, all that matters is whether you win or lose. It's unacceptable that with all the data they have, and in the age of AI, they can't come up with a better and more fair system. The people who put this together are practically running a casino.

The goal of the game is to destroy enemy base, not to kill enemy heroes.

You can win a game dealing a total 0 damage to enemy heroes, if all you do is hitting towers. But if you only push lanes - you can still win the game, while dealing 0 damage to enemy duildings yourself, creeps will do all the work for you.

All those statistics like KDA or networth doesnt mean anything. The only measurable progress is building health. But since you can damage them indirectly, even this statistic will be inaccurate.

Ofcourse, its cool to have detection when someone is intentionally ruining the game. Like breaking items, staying afk or feeding, placing all wards at the same spot.

But other than that its hard to tell who made the most impact in game: carry can have high KDA, but he was stealing farm from all of his teammates.

Those are some good points, I appreciate your input. I don't really have all the answers here on how to develop a better system, but I do believe it is possible to have something better than just basing it on win/lose alone with no other factors in play. Aside from that, I think it's obvious there are a number of stupid things happening like the example I gave in the original post, that should/could be addressed.

What you've described too, is emphasis on pushing and taking towers, yes that is really the most important part of it, I won't speculate is it possible to incorporate elements of that into consideration outside of win/loss.

[HardRock]

One possible suggestion:
- The faster the game ends, the more MMR you gain\loose as a team.

It doesnt count individual impact of players, but at least its better than having fixed MMR gain for any matchup.

[ʍolɟ ǝɥʇ ǝsɹǝʌǝɹ]

If it was small like 20% of what you gain or lose it might be ok. But it can't be substantial enough people go for this instead of trying to win

[ʍolɟ ǝɥʇ ǝsɹǝʌǝɹ]

Originally posted by HardRock:

One possible suggestion:
- The faster the game ends, the more MMR you gain\loose as a team.

It doesnt count individual impact of players, but at least its better than having fixed MMR gain for any matchup.

This is an interesting idea since longer gamed show they are more evenly matched. And base rush/zoo is all butdead

[Schopenhauer]

Originally posted by ----:

It's not possible to measure performance. Performance is objective.

And why is there a mvp after every match? 2 of winner team and one of loserteam? If its not possible?
Why does the one mvp of team who lost, lose mmr?
Thats just stupid.

[Absche]

I think besides the overall challenge to judge individual performance in a team game with a mix of full stack, small parties and solo gaming, everybody in solo queue would more play for the individual performance instead of for the win. The system would support selfish gaming even more than the game design supports it anyway right now.

[Zockolade]

Originally posted by HardRock:

One possible suggestion:
- The faster the game ends, the more MMR you gain\loose as a team.

So if the matchmaker make it more winable for one team they also get more mmr.
Sounds like freeloot for a matchmaking system that already sucks.

[ʍolɟ ǝɥʇ ǝsɹǝʌǝɹ]

Originally posted by Zockolade:

Originally posted by HardRock:

One possible suggestion:
- The faster the game ends, the more MMR you gain\loose as a team.

So if the matchmaker make it more winable for one team they also get more mmr.
Sounds like freeloot for a matchmaking system that already sucks.

But on the other side, a reason to drag out unwinnable games

[Pepega or Gigachad?]

Originally posted by Schopenhauer:

Originally posted by ----:

It's not possible to measure performance. Performance is objective.

And why is there a mvp after every match? 2 of winner team and one of loserteam? If its not possible?
Why does the one mvp of team who lost, lose mmr?
Thats just stupid.

MPV is not accurate. Its just based on a few stats like kills, stuns and heal, etc. They don't reflect game performance.