General DiscussionShould matchmaking reward performance?

How do you calculate if someone does expectionally well compared to team?

If so, this could only be like +3 points or something.

There is no way to measure performance.

I imagine that's the ideal system where you are assesed individually if say you are in a solo Queue, but it's not something that is easy to implement.

it's hard to judge performance objectively

imo Valve should implement a vote at the end of the match, where both teams vote for 1 MVP who would get extra 25 points (can be on both winning and losing team)

I think that the result of such a system wont be any different to the actual one. Current MM system measures how you perform on average (let me skip easy maths that proves it), while the discussed system would measure your skill taking into account error term. Sum of error terms in longrun is 0, so the obtained result gonna be exactly the same.
Well, there are some little details also like additional motivation (that imho wont influence that much), and difficulties we face when trying to elaborate such a system. Overall, I think valve wont ever introduce personal performance evaluation, and I agree with them.

There is an initial phase on accounts where your hidden MMR is very sensitive to your performance. This dies off pretty quickly, but it moves you roughly where you should be. Over the course of the remaining 100 or so games + further calibration phase, unless you are outside the boundaries, your MMR should be somewhat accurate.

Its fine for any further MMR changes to be based purely on win/loss.

Now if everyone started at X mmr, and got +y/-y every game, that would be awful because it would take a considerable number of games to separate players.

That's what i'm talking about, the game HAS the mechanics to assess how well a player did. There are stuff like how many creeps near you die without you last hitting, hero damage, gold per minute, gold fed to the enemy, hero damage. Of course, stuff like how much space did you create for your team can't be measured and as a result, it's possible that offlaners would end up with lower MMR.
They still would get points for low gold fed to enemy, etc.

These stats are a good way of assessing how good a player is, and should determine the end score. If you performed well, you shouldn't take the same punishment as someone that under-performed just because you lost the game.

Also, a voting option would be terrible because pubbers are ignorant.

PS: The current system promotes more player hostility, since winning is all that matters, players blame each other etc.

Well whatever you ^wrote there, is calculated when you get your initial mmr, right?

There's no reason to give people additional points. Just because you overperformed one game doesn't mean you're better than the game you're playing.

I don't understand why people aren't satisfied with the system, and want changes.
The only problem there is, is people with horrible attitude and lack of very good players.

Valve can't make the community play better, unless they turn the game into league of legends.

Easy observation - if you think u gonna benefit from such a system, then you consider that you are playing better than your teammates most of the time. If it is really like that, in longrun your winrate gonna be positive (reasons are obvious), and u will rise mmr with the existing system, untill u get to the point where on average everyone plays same good as you do. The suggested system won't contribute any differences, it will just be more difficult to develop. Senseless work for programmers.

Well what i'm proposing is more of a every month, 10 games will be sensitive to these settings.
Kind of like calibration matches, but of course with less magnitude.

This would really be great, because even if you belong to a higher bracket than you currently are (inb4 you never belong to a higher bracket), and you win like 60% of the games, it takes too long to get to where you want with the current system. And some people just don't have the time for it, and end up giving up on ranked, and keep on playing in the same old skill level and never improve.

Time is what improves your skill in dota my ass, playing against good players and with good players is what improves a player without attitude issues. Being around them, being able to soak up the experience, seeing what they do, but most importantly, trying harder because you know the enemy is worth your time.

It's monkey see monkey do, if you see your team trying hard to win, playing it right, you automatically do the same. You know you should buy a smoke if you want to gank a warded spot, but do you really if you know the match is full of baddies? If you're playing against smart opponents, you would.

What u propose looks like spherical caw in a vacuum case. And furthermore such a system gonna be easy to abuse, and therefore the results will be negative.
The bracket u r playing in usually reflects ur skill rather accurately, and I cant think of any way you could progress in dota without actually improving your rank, either in mmr games or hidden one in normal matchmaking.

How is it a spherical cow?
It's a modification of an already existing system. I ditched the performance idea.

edit: also are you a physicist?

@waow

We are playing a game called DotA. Not who can get the highest GPM/XPM or CS. MMR is supposed to reflect how good you are at winning games, and you can do that indirectly too. I remember watching the EG knight strat, omni would feed, get no xp and lose his tower, but it created so much space for the rest of the team. Now should that player be punished while his team gets the benefit?

Take a look at my stats. They are decent yeah? Yet my win rate is barely above 50% and I was hovering around 40% for ages, because while I am pretty good at farming, I just don't have knowledge or mindset for playing aggressively (without a significant advantage). You will see the better players make the most of the opportunities they have, while I play a lot more conservatively.

Individual performance matters when you are or could potentially be significantly better than others on your team, it makes sense when they have no idea what your skill level is. Once they have a rough idea, and you are playing well consistently, that should be reflected in wins and thus, an increase in MMR.

nah, i suck in physics although trying to improve. just used this old meme in the meaning that the situation in which such a system gonna be useful is highly hypothetical.

@Howdoyougetsliced
Thank you for the clarification, for some reason i thought i was playing league of legends

people shouldn't be making huge jumps in ability though. And if they are, that should influence their ability to win games at their current MMR, they should be the difference. At the moment I'm probably a lot better than my current rating, but it would be silly for the system to push me up without any real reason given that my previous performances have put me where I currently sit.

summarizing all said above (not only my comments), the one-and-only measure of how much u really contribute to the win is the share of wins/losses wheen playing in certain mmr level and then putting you to the bracket where your winrate is exactly 50%. All the other types of measurement can be easily abused (kda, hd, lh, etc.). So in conclusion the system we have is optimal one.