I think it's time to start considering Performance Based MMR

General Discussion
Let me start with this. Carrying is possible in this game, to a degree. Playing a high impact hero skillfully well, can, and for the most part will, lead to a win for your team.

However, with that said, it is not possible to carry every game.

When there are 9 other equally skilled players, and you are the clear outlier, you very likely will win the game for your team given that your teammates are at least halfway competent.

Sometimes, you will never win that game. I'm not talking about team composition losses. Those happen and I realize that sometimes the matchmaker will just be completely unfair to you. I'm talking about those times you get this person that has some really warped sense of how to play the game (Almost always a Smurf) that thinks that as Samuro their entire time should be focused on taking camps, even if those camps have zero impact on the game and can apply absolutely zero map pressure. Those Varians and Artanis players that think they can 1v5 the enemy team and end up dying. The Probius, Sylvanas, or Sgt Hammer that respond to absolutely no objectives whatsoever, stay alone pushing a lane, then get ganked and die.

The simple fact of the matter is, for all of your personal skill in the world, you cannot carry these players. Their negative impact on the game is much greater than your positive impact, and try as hard as you may, that game is a decided loss.

Sometimes if you have 2-3 people who are on point for most of the game, they might have a slight chance of offsetting that player, but most of the time that one person decides the loss.

Now, I don't think you should go up in MMR for losing. That would be stupid and almost completely trivialize the system. Winning the game should be the end all be all deciding factor in whether or not you gain or lose MMR. However, I don't think the team's star performer should lose as much as the rest of the team.

I think it's time to start considering using performance metrics to help people gain or lose more or less MMR based on how they performed during a game.

Blizzard's current algorithm isn't perfect, but it would help mitigate some of these issues. A player who was present for 90% or more of the team's skirmishes, has the least amount of deaths, most hero damage, etc, should lose less MMR than that Sylvans who never showed up for a single Tribute and spent all game working on taking down a single fort after dying 9 times within the first 15 minutes of the game.

I'm reposting this because it does not violate any terms in the Code of Conduct. Blizzard, fix your moderation.
04/10/2017 10:10 AMPosted by Drothvader

Blizzard's current algorithm isn't perfect, but it would help mitigate some of these issues. A player who was present for 90% or more of the team's skirmishes, has the least amount of deaths, most hero damage, etc, should lose less MMR than that Sylvans who never showed up for a single Tribute and spent all game working on taking down a single fort after dying 9 times within the first 15 minutes of the game.

[/quote]

Who wants to go solo lane???? No one??? why...because the system you described put the solo laner at disadvantage. Who wants to play Abby? Everyone??? why, you can be at every team fight on the map.................and you can even sit at starting point so you never die........let the team carry you and you will get MVP if team wins, and still don't lose much points if your team loss.

Sounds like great way to abuse your system.
04/10/2017 10:15 AMPosted by DarkAngel
Who wants to go solo lane???? No one??? why...because the system you described put the solo laner at disadvantage.

That's the point. You shouldn't be spending the entire game solo laning. If you're one of the people I'm describing you don't respond to map threats or don't show up for any objectives just so you can push a single lane all game, you're a part of the problem and a large part of why people like me are so frustrated.

Also, Abathur acts through his allies. His physical body may be elsewhere, but he has a very strong team presence and this is accounted for in Blizzard's algorithm.

04/10/2017 10:15 AMPosted by DarkAngel
why, you can be at every team fight on the map.................and you can even sit at starting point so you never die........let the team carry you and you will get MVP if team wins, and still don't lose much points if your team loss.

Not sure what universe you reside in but more than just "not dying" goes into making MVP. Sometimes MVP is given to the absolute worst person on the team, but this isn't common.

I'm pretty sure you actually have no idea what you're talking about.
Good players win games.
Another tl;dr thread brought to you by Drothvader.
04/10/2017 10:22 AMPosted by Akrazo
Good players win games.

I'm pretty sure I opened with that...

However, there are games that you will still lose even if you're a Grand Master.
I think the algorithms and possibilities for "good plays" would be really hard for them to base that off of. There are a lot of intangibles that come into play during any given match.
Your answer to Darkangel's post is completely off, read his post again.

There are many ways to abuse your system and that is why Blizzard never implemented a system like that. We want players to try to actually win the game and not to maximise their stats and any system that is based on game stats is going to be abused one way or an other, and even if there is no great way to abuse it, it will always distract players from the one and only goal: trying to win the game.

You complain about players who do not participate enough or whatever, but your system is only going to encourage toxic behavior because then players will have a good reason to do it : maximise their stats and their MMR.
04/10/2017 10:26 AMPosted by Akrazo
I think the algorithms and possibilities for "good plays" would be really hard for them to base that off of. There are a lot of intangibles that come into play during any given match.

This is stuff Blizzard is already iterating on and trying to improve. It's what the MVP and card system is based on.

All they need is some small tweaks to how they value certain metrics. For instance, I believe that "Headhunter" is valued too highly. They could tone it down a bit and it could fix some of the issues.

Either way, the game already knows how to do this. It's what the "On Fire," MVP, and End Game Card system are based on. Very rarely do I see bad players on fire so I'm somewhat confident Blizzard has the right idea about how to measure performance metrics. Very rarely are the best players in the game not on fire.
You said yourself that if you play better you are more likely to win games, so MMR is already doing what it's intended to do. It doesn't matter if it goes down from one bad game, what matters is that over time it accurately represents your skill level. If you keep consistently playing better than everyone else at your MMR then you should win more games on average and your MMR will rise higher than theirs. So what if you have a bad game and lose MMR with them? You'll go onto to other games where you'll be more likely to win and gain that MMR back, while the others will go onto games they might lose.

Also, changing MMR to take into account anything but win/lose is just asking for people to game the system. Whatever stat or stats you want to choose to adjust MMR, there are scenarios where a player didn't do well in those stats but was important to their teams performance, or vice versa.
04/10/2017 10:27 AMPosted by Flammy
There are many ways to abuse your system and that is why Blizzard never implemented a system like that. We want players to try to actually win the game and not to maximise their stats and any system that is based on game stats is going to be abused one way or an other, and even if there is no great way to abuse it, it will always distract players from the one and only goal: trying to win the game.

What part of you only gain MMR if you win a game is being lost in translation?

Sure, you could pad stats all you want, but unless you're winning games it won't matter.

Very rarely will someone who is spending all game solo laning ever become "On Fire." I'm fairly confident the devs already have a good sense of how to measure performance based metrics and the game more or less knows who the star performers are.

This rampant "abuse" you guys think is going to suddenly appear, won't.
MMR for rank
MMR + Performance metric for matching.

I don't think any performance-based thing is a good idea for affecting your MMR directly, as people will exploit the system - and then continue to complain when their grand master games are full of other noobs who only got there because they exploited the system, too. Meanwhile, the people who sacrificed in-game stats for a victory will be relegated to platinum or gold, which defeats the purpose altogether.

However, as an additional number to quantify players when it comes time to match them together in a game, I think it's a great idea. Imagine a 2D grid. On one axis is your MMR, and on the other is your PR (performance rating). Matches are made by finding a pairwise cluster of players, but your MMR/Rank adjustment after each match is determined only by the MMRs of the players involved.

Keep in mind your PR doesn't need to be a single number, either. There is a lot to this game - time spent dead, soak, draft awareness, map awareness, merc captures, healing done, T/D... these are all possible metrics for a "PR vector". While some sort of PCA could be used to create a single number, it's not needed as k-means clustering can be done in any dimensional space.

This would essentially quantify playstyle alongside MMR. The idea being, if everyone in a game, for example, thinks that fighting in the jungle is the best way to win, then it will still be a good game for them.
04/10/2017 11:16 AMPosted by normalice
MMR for rank
MMR + Performance metric for matching.

I don't think any performance-based thing is a good idea for affecting your MMR directly. However, as an additional number to quantify players when it comes time to match them together in a game, I think it's a great idea. Imagine a 2D grid. On one axis is your MMR, and on the other is your PR (performance rating). Matches are made by finding players who cluster together, but your MMR/Rank adjustment after each match is determined only by your MMR.

If you want to make them separate, I'm fine with that as well.

If someone is doing really well for almost all of their games, try throwing them in an environment where they have to prove themselves worthy. If they can hang there competently, adjust their MMR to compensate.

With what you're suggesting though, there's a possibility for a huge disparity between player MMR and this Performance Rating. There needs to be at least some semblance of parity between them.
"I'ma bail when I don't need to to preserve my KDA."
04/10/2017 11:16 AMPosted by normalice
Meanwhile, the people who sacrificed in-game stats for a victory will be relegated to platinum or gold, which defeats the purpose altogether.

Like that's any different from now... which is why I do strongly support some kind of change like this.

04/10/2017 11:47 AMPosted by Amaru
"I'ma bail when I don't need to to preserve my KDA."

And that should be taken into account. Not kills. Not damage. Not merc captures. Everything. There's a thing for "Clutch Heals" and "Escape Artist", they're not only monitoring how much you heal or tank, but when you do it as well. So if you're leaving a fight at full health and mana, that should be considered. If you leave and your team dies, maybe that was a bad decision. If you leave and your team wins, maybe you had a reason and it was a good decision. If you ping retreat and 3 allies leave but 1 stays behind, maybe it was THEIR bad decision, and not yours to retreat. I think they can figure it out at least a little bit with what they already calculate for the score screen.
I play KT/Luna/Naz/Zag/Sylv/Tass - all of which stat bloat.

Performance = no
04/10/2017 11:57 AMPosted by Fawxkitteh
And that should be taken into account. Not kills. Not damage. Not merc captures. Everything. There's a thing for "Clutch Heals" and "Escape Artist", they're not only monitoring how much you heal or tank, but when you do it as well. So if you're leaving a fight at full health and mana, that should be considered. If you leave and your team dies, maybe that was a bad decision. If you leave and your team wins, maybe you had a reason and it was a good decision. If you ping retreat and 3 allies leave but 1 stays behind, maybe it was THEIR bad decision, and not yours to retreat. I think they can figure it out at least a little bit with what they already calculate for the score screen.

That's why I'm saying these people have no clue what they're talking about.

The game already recognizes good plays through end game cards. The performance metrics aren't as simple as KDA and Hero Damage. I don't think people are actually giving Blizzard enough credit for the system they're building here.

The next logical step would be to apply those metrics toward your MMR.

As I said earlier, rare do I ever see someone who is just stat padding end up On Fire. You have to actually be making an impact to do well. Having high damage doesn't cut it.

The only place where I see some questionable callouts is in the MVP system.
04/10/2017 10:22 AMPosted by Akrazo
Good players win games.


If no incompetent smurf around, who single handed will cost you the game. The MMR system has flaws, just read threads about Ranked in general. Rank is not equal with skills, and that is a problem in every tier.
Maybe a skill / performance based Rank make more sense, and cannot be abused that easy like smurf accounts do all around in the present system.
04/10/2017 11:19 AMPosted by Drothvader
With what you're suggesting though, there's a possibility for a huge disparity between player MMR and this Performance Rating. There needs to be at least some semblance of parity between them.
it's a possibility, sure, but I don't think it would be that big of an issue. For one thing, what I'm suggesting is a (non-binary) continuous metric of performance. The vast majority of people would still fall somewhere between "good" and "bad."

Keep in mind the point of the matching system is not to rank people up because they are good, but to give them fun/competitive games. I would be more than happy to remain in Bronze 5 forever, if it meant that I was facing equally skilled opponents, with equally skilled allies. But that's not what happens. My only interest in ranking up is to diminish the frequency with which I get bad players on my team. But, I've been in diamond and I've been silver, and I've yet to see any difference in teammate quality. In fact, the most intense games I've played were unranked games..

Anyway, the point of this is, if you have a 'bad' player on your team, this sort of "pair-wise clustering" system would help ensure your opponents also have a player who is bad for similar reasons. Once that is done, rank should take care of itself. And even if it doesn't, it wouldn't matter because everyone would still be (theoretically) teamed up with and against like-minded players, resulting in better games.

Plus, keep in mind that this PR would be in conjunction with MMR, meaning if someone in silver is performing exceptionally well, they could very easily end up in a platinum game, if everyone in platinum has a similar PR vector. And if they show that's where they belong by winning, the MMR difference would accelerate their ascent through the ranks. I think that, after some time, the disparity you mentioned would directly correlate to rank - meaning those with a high PR would also have a high MMR, and vice versa.

Join the Conversation

Return to Forum