|
Please be advised:
We will be closing this General thread in 24 hours. It will remain searchable.
After that we will require new threads to discuss topics.
Questions should go in the stickied Q&A thread, screenshots and PotG will go in the PotG sticky, QQ/Rage/Complaints should go in the QQ/Rage thread. If you want to talk about maps or strategies open a new thread.
Any comments or concerns will be logged please forward them to ZeromuS. This new forum is still fluid so we will try this out. General TL rules will still apply to new threads. |
On October 30 2015 02:09 Gorsameth wrote:
if not via MMR what system would you recommend is used for matchmaking?
Because you want something that attempts to construct reasonable equal teams and seeks to limit one sided stomps.
MMR is far from perfect. It has a ton of flaws. It just happens to be better then the alternatives.
I am not saying it is bad at its job. I am saying its not something that needs to be surfaced for the end user to see improvement. A league or season system is more effective, since they are completing against people who also want to compete. It also resets, allowing people to see how much better they do in the next season or try a new role/hero.
In short: there are better ways to show improvement than just surfacing the match making number that is designed to plateau.
|
It is a shit system to reflect skill and always has been. Many active dota players mock the MMR system because of it. Just like the SC2 ladder, where people got to Grand master by six poolin
You are looking at extreme outliers. With that type of logic you can defy all types of metricsthat are very valuable in normal circumstances. In fact, there are almost no metrics that reflects everything perfectly evertime. But as long as a metric gives us some extra type of information and we understand its potential flaws so we can take them into account in the overall assesment, it's a useful metric.
And if we look at Starcraft, maybe there is one or 2 guys who actually are slightly inferior than a high master league players in a real macro game. But as someone who actually played against lots of GMs, I can tell you that there almost always is a difference. And you don't get into GM if you are worse than a midmaster.
With regards to MOBA's, if one massed one hero while climbing, then the MMR would ofc just reflects how good you are with that specific champion. Noone will be acquired by a pro team only due to MMR, but it's one of the major criterias that pro teams asses before they decide whether you are worthy of consideration.
Another criteria is ofc that you have a proper hero pool/can play those that are viable in the meta.
But as someone who has no willingness to go pro, but want to have fun and improve at the same time, I am gonna play the heroes that I have fun with and not gonna play the heros that are boring. At the same time I want to be able to measure whether I am getting better playing those heroes. Thus some type of MMR system is of importance for me (I don't need the actual number though - the LOL model is good enough for me).
|
On October 30 2015 02:14 Plansix wrote:Show nested quote +On October 30 2015 02:09 Gorsameth wrote:
if not via MMR what system would you recommend is used for matchmaking?
Because you want something that attempts to construct reasonable equal teams and seeks to limit one sided stomps.
MMR is far from perfect. It has a ton of flaws. It just happens to be better then the alternatives.
I am not saying it is bad at its job. I am saying its not something that needs to be surfaced for the end user to see improvement. A league or season system is more effective, since they are completing against people who also want to compete. It also resets, allowing people to see how much better they do in the next season or try a new role/hero. In short: there are better ways to show improvement than just surfacing the match making number that is designed to plateau.
Why does it matter if it surfaces or not? It literally means nothing other than the level that you're matched at. If you want to track your improvement, you can do so by watching your own games and seeing your decision making and plays. No formula will track that for you. And if you don't care about MMR or rankings or whatever, than you can just not look at the number and play with the focus of improving. Simple as that.
MMR is not designed to plateau. If you improve, you win more games, and it increases. There's no one out there with a vendetta against you going "well let's just match him against shitty/pro players to make sure he never changes".
|
On October 30 2015 02:14 Hider wrote:Show nested quote + It is a shit system to reflect skill and always has been. Many active dota players mock the MMR system because of it. Just like the SC2 ladder, where people got to Grand master by six poolin What the hell. You are looking at extreme outliers. In almost all cases, a grandmaster is someone that's noticeably more skilled than a midmaster player. And if you massed one hero in a moba, then the MMR just reflects how good you are with that specific champion. Noone is going to go on a pro team only due to MMR, but it's one of the major crtierias that pro teams asses before they decide whether you are worthy of consideration.
You are correct. Mason in dota had 4K MMR and played with EG at one TI. MMR does not denote skill or reflect if the player is good. It only reflects their ability to win within the match making system against opponents it deems to be similarly skilled.
The argument is: Surfacing MMR is an effective way to show improvement for the end user. I say it isn't because the system is easily gamed and naturally plateaus. It is good at creating levels of skills in the user base, but not good at showing people making the slow climb to becoming better, especially in team games when the player makes of 20% of his/her chance to win.
Or more importantly, that argument does not take into account designing a separate system for competitive players to compare their accomplishments.
On October 30 2015 02:18 Requizen wrote:Show nested quote +On October 30 2015 02:14 Plansix wrote:On October 30 2015 02:09 Gorsameth wrote:
if not via MMR what system would you recommend is used for matchmaking?
Because you want something that attempts to construct reasonable equal teams and seeks to limit one sided stomps.
MMR is far from perfect. It has a ton of flaws. It just happens to be better then the alternatives.
I am not saying it is bad at its job. I am saying its not something that needs to be surfaced for the end user to see improvement. A league or season system is more effective, since they are completing against people who also want to compete. It also resets, allowing people to see how much better they do in the next season or try a new role/hero. In short: there are better ways to show improvement than just surfacing the match making number that is designed to plateau. Why does it matter if it surfaces or not?
At the end of the day, there is no proof it helps or hurts the game. There are people in the Dota community who liked the game before MMR was a thing and everyone focused on ranked games. In team games, it only shows your ability to win in solo que games, not with a team.
|
How's Mason a good example of anything? He fools around half the time and other half the time he quits games. Big surprise, if you're not gonna try you're not gonna rise in MMR.
Also, I'm not sure how you think playing 2-3 heroes is gaming the system...by that logic playing mid is gaming the system as well - you're applying more control/impact over a particular game.
MMR= ur ability to win pub dota matches, how it doesnt matter
|
On October 30 2015 02:21 Andre wrote:
How's Mason a good example of anything? He fools around half the time and other half the time he quits games. Big surprise, if you're not gonna try you're not gonna rise in MMR.
Also, I'm not sure how you think playing 2-3 heroes is gaming the system...by that logic playing mid is gaming the system as well - you're applying more control/impact over a particular game.
MMR= ur ability to win pub dota matches, how it doesnt matter
So is MMR the best metric for someone to judge their improvement in Dota over all?
|
You are correct. Mason in dota had 4K MMR and played with EG at one TI. MMR does not denote skill or reflect if the player is good. It only reflects their ability to win within the match making system against opponents it deems to be similarly skilled.
The argument is: Surfacing MMR is an effective way to show improvement for the end user. I say it isn't because the system is easily gamed and naturally plateaus. It is good at creating levels of skills in the user base, but not good at showing people making the slow climb to becoming better, especially in team games when the player makes of 20% of his/her chance to win.
Are you talking about variance?
Surely if you have some lucky teammates that carry you, your short-term ELO can be boosted. But long-term you need to get better to actual climb in MMR/ELO.
It is good at creating levels of skills in the user base, but not good at showing people making the slow climb to becoming better,
Sure it does. If I main Riven in League of Legends and go from gold league to diamond, that's a pretty strong indicator I got better at the champion.
Now obviously if I only played Riven, you could still argue that I am not as good of an overall player as other diamond players that have a wider champion pool, but it definitely measures my own progress.
|
On October 30 2015 02:25 Plansix wrote:Show nested quote +On October 30 2015 02:21 Andre wrote:
How's Mason a good example of anything? He fools around half the time and other half the time he quits games. Big surprise, if you're not gonna try you're not gonna rise in MMR.
Also, I'm not sure how you think playing 2-3 heroes is gaming the system...by that logic playing mid is gaming the system as well - you're applying more control/impact over a particular game.
MMR= ur ability to win pub dota matches, how it doesnt matter So is MMR the best metric for someone to judge their improvement in Dota over all?
I'm not sure, it's the only reliable metric we have though. From anecdotal evidence, I've had people on my flist stay at a certain MMR range for ages then climb very quickly to a new MMR and stay there. Short term MMR changes don't really matter though.
|
On October 30 2015 02:18 Plansix wrote:
The argument is: Surfacing MMR is an effective way to show improvement for the end user. I say it isn't because the system is easily gamed and naturally plateaus.
It plateaus because the player plateaus. That is not the fault of Elo, it shows what it is supposed to.
On October 30 2015 02:18 Plansix wrote:
In team games, it only shows your ability to win in solo que games, not with a team.
Yeah and I only play soloQ games. So showing my ability to win soloQ games is exactly what I want from it. Why? Because it makes me care if I win said soloQ game or not. If the system makes me win 50% of my games no matter what I do and does not reflect me winning n any way, I have no reason to care about winning a particular game. Then I will rather try to get kills or whatever instead of going for objectives which worsens the game experience overall for everyone involved.
Overall Elo gets mostly critized by people that overestimate their own skills cant cant deal with Elo showing a lower number than they feel it should. Basically it is people that care too much about their e-peen that dont want it.
|
On October 30 2015 02:29 Andre wrote:Show nested quote +On October 30 2015 02:25 Plansix wrote:On October 30 2015 02:21 Andre wrote:
How's Mason a good example of anything? He fools around half the time and other half the time he quits games. Big surprise, if you're not gonna try you're not gonna rise in MMR.
Also, I'm not sure how you think playing 2-3 heroes is gaming the system...by that logic playing mid is gaming the system as well - you're applying more control/impact over a particular game.
MMR= ur ability to win pub dota matches, how it doesnt matter So is MMR the best metric for someone to judge their improvement in Dota over all? I'm not sure, it's the only reliable metric we have though. From anecdotal evidence, I've had people on my flist stay at a certain MMR range for ages then climb very quickly to a new MMR and stay there. Short term MMR changes don't really matter though.
I think this is the argument against surfacing it. It fluctuates a ton early on and doesn't really tell us much in the long run.
It's a good enough system for matching making, but once you surface it it starts to take on meanings it didn't/shouldn't have. Like that awful Gearscore mod in WoW and it's in-game successor... character iLvl (and I suppose now it's Proving Grounds, although I didn't play much after those were introduced).
|
Yeah and I only play soloQ games. So showing my ability to win soloQ games is exactly what I want from it. Why? Because it makes me care if I win said soloQ game or not. If the system makes me win 50% of my games no matter what I do and does not reflect me winning n any way, I have no reason to care about winning a particular. Then I will rather try to get kills or whatever instead of going for objectives which worsens the game experience overall for everyone involved.
Yes that's my stand to, and I think that Plansix looks at "improvement" from an esports-perspective and not a casual perspective. It seems that the only thing that matters is if youimprove your skillset in a way that's relevant for competitive play.
But what's relevant for most of us is that we improve our skillset so we can climb in soloQ. As long as we have fun in the proces, the gamedesigner has done his job.
|
On October 30 2015 02:37 Hider wrote:Show nested quote +Yeah and I only play soloQ games. So showing my ability to win soloQ games is exactly what I want from it. Why? Because it makes me care if I win said soloQ game or not. If the system makes me win 50% of my games no matter what I do and does not reflect me winning n any way, I have no reason to care about winning a particular. Then I will rather try to get kills or whatever instead of going for objectives which worsens the game experience overall for everyone involved. Yes that's my stand to, and I think that Plansix looks at "improvement" from an esports-perspective and not a casual perspective. It seems that the only thing that matters is if youimprove your skillset in a way that's relevant for competitive play. But what's relevant for most of us is that we improve our skillset so we can climb in soloQ. As long as we have fun in the proces, the gamedesigner has done his job.
Its also improvement over time and long term. Peoples MMR isn’t going to be reset often because it ruins the player experience for the entire player base. So if someone tries to play a new hero, their MMR does down. Get into a new mode or play style, MMR goes down. So the system punishes players for doing the specific things that will make them better long term.
In your case, a surfaced MMR system would encourage you to play Riven times and never change it up until you are as skilled with another hero. And if that is the case, is there a way to design a better system that doesn’t discourage players from practicing new heroes within the competitive system. Because there are people in dota that say hat the system of surfaced MMR isn't fun, but unranked games have become to clowny for them to enjoy.
Note: And before people say “you just ignore that stuff” remember all the threads about ladder anxiety. That is not an argument that surfaced MMR is the best system to show improvement.
|
On October 30 2015 02:46 Plansix wrote:Show nested quote +On October 30 2015 02:37 Hider wrote:Yeah and I only play soloQ games. So showing my ability to win soloQ games is exactly what I want from it. Why? Because it makes me care if I win said soloQ game or not. If the system makes me win 50% of my games no matter what I do and does not reflect me winning n any way, I have no reason to care about winning a particular. Then I will rather try to get kills or whatever instead of going for objectives which worsens the game experience overall for everyone involved. Yes that's my stand to, and I think that Plansix looks at "improvement" from an esports-perspective and not a casual perspective. It seems that the only thing that matters is if youimprove your skillset in a way that's relevant for competitive play. But what's relevant for most of us is that we improve our skillset so we can climb in soloQ. As long as we have fun in the proces, the gamedesigner has done his job. Its also improvement over time and long term. Peoples MMR isn’t going to be reset often because it ruins the player experience for the entire player base. So if someone tries to play a new hero, their MMR does down. Get into a new mode or play style, MMR goes down. So the system punishes players for doing the specific things that will make them better long term.
That heavily depends on how you define "punishes the player".
If you're talking about short term MMR drops when trying new things, then yes. However, if the purpose of trying new things is to improve your play holistically, then that short term drop shouldn't matter because, ideally, it would result in long term gains.
MMR benefits from a large number of games. The more games you play, the more accurate your relative MMR will be. If you get better, your number trends upwards.
As for ladder anxiety - I don't disagree with a Normal queue that has hidden MMR and a Ranked queue that makes MMR visible. If you're anxious about rank, play normals - it shouldn't matter to you anyway. If you want to play in a more structured, competitive setting, play ranked.
|
On October 30 2015 02:25 Plansix wrote:Show nested quote +On October 30 2015 02:21 Andre wrote:
How's Mason a good example of anything? He fools around half the time and other half the time he quits games. Big surprise, if you're not gonna try you're not gonna rise in MMR.
Also, I'm not sure how you think playing 2-3 heroes is gaming the system...by that logic playing mid is gaming the system as well - you're applying more control/impact over a particular game.
MMR= ur ability to win pub dota matches, how it doesnt matter So is MMR the best metric for someone to judge their improvement in Dota over all?
The answer is still yes, obviously. On the balance of probability your MMR is the best single metric indication of your skill level. Trotting out Mason as a counter-example is meaningless when taking a macro view of this topic. That's like saying height doesn't correlate to success in basketball by talking about Nate Robinson.
|
On October 30 2015 02:53 jcarlsoniv wrote:Show nested quote +On October 30 2015 02:46 Plansix wrote:On October 30 2015 02:37 Hider wrote:Yeah and I only play soloQ games. So showing my ability to win soloQ games is exactly what I want from it. Why? Because it makes me care if I win said soloQ game or not. If the system makes me win 50% of my games no matter what I do and does not reflect me winning n any way, I have no reason to care about winning a particular. Then I will rather try to get kills or whatever instead of going for objectives which worsens the game experience overall for everyone involved. Yes that's my stand to, and I think that Plansix looks at "improvement" from an esports-perspective and not a casual perspective. It seems that the only thing that matters is if youimprove your skillset in a way that's relevant for competitive play. But what's relevant for most of us is that we improve our skillset so we can climb in soloQ. As long as we have fun in the proces, the gamedesigner has done his job. Its also improvement over time and long term. Peoples MMR isn’t going to be reset often because it ruins the player experience for the entire player base. So if someone tries to play a new hero, their MMR does down. Get into a new mode or play style, MMR goes down. So the system punishes players for doing the specific things that will make them better long term. That heavily depends on how you define "punishes the player". If you're talking about short term MMR drops when trying new things, then yes. However, if the purpose of trying new things is to improve your play holistically, then that short term drop shouldn't matter because, ideally, it would result in long term gains. MMR benefits from a large number of games. The more games you play, the more accurate your relative MMR will be. If you get better, your number trends upwards. As for ladder anxiety - I don't disagree with a Normal queue that has hidden MMR and a Ranked queue that makes MMR visible. If you're anxious about rank, play normals - it shouldn't matter to you anyway. If you want to play in a more structured, competitive setting, play ranked.
So why is it better than a ladder or league system?
|
On October 30 2015 03:02 Plansix wrote:Show nested quote +On October 30 2015 02:53 jcarlsoniv wrote:On October 30 2015 02:46 Plansix wrote:On October 30 2015 02:37 Hider wrote:Yeah and I only play soloQ games. So showing my ability to win soloQ games is exactly what I want from it. Why? Because it makes me care if I win said soloQ game or not. If the system makes me win 50% of my games no matter what I do and does not reflect me winning n any way, I have no reason to care about winning a particular. Then I will rather try to get kills or whatever instead of going for objectives which worsens the game experience overall for everyone involved. Yes that's my stand to, and I think that Plansix looks at "improvement" from an esports-perspective and not a casual perspective. It seems that the only thing that matters is if youimprove your skillset in a way that's relevant for competitive play. But what's relevant for most of us is that we improve our skillset so we can climb in soloQ. As long as we have fun in the proces, the gamedesigner has done his job. Its also improvement over time and long term. Peoples MMR isn’t going to be reset often because it ruins the player experience for the entire player base. So if someone tries to play a new hero, their MMR does down. Get into a new mode or play style, MMR goes down. So the system punishes players for doing the specific things that will make them better long term. That heavily depends on how you define "punishes the player". If you're talking about short term MMR drops when trying new things, then yes. However, if the purpose of trying new things is to improve your play holistically, then that short term drop shouldn't matter because, ideally, it would result in long term gains. MMR benefits from a large number of games. The more games you play, the more accurate your relative MMR will be. If you get better, your number trends upwards. As for ladder anxiety - I don't disagree with a Normal queue that has hidden MMR and a Ranked queue that makes MMR visible. If you're anxious about rank, play normals - it shouldn't matter to you anyway. If you want to play in a more structured, competitive setting, play ranked. So why is it better than a ladder or league system?
Ladder and League systems still use MMR under the hood. They just hide the exact number from the player and fudge it by using broader definitions when displaying it.
I also fully expect a ranked mode in overwatch to use leagues rather then the actual MMR number as per Hearthstone/HotS/SC2
|
MMR and ELO dont just give you a metric to measure yourself... they also give you a competitive goal you can set for yourself.
And I dont agree in ANY way whatsoever that resetting a players MMR/ELO is a negative to ANY player.
Everyone LOVES the chance to start over and a clean slate.
Its probably the most popular time of League right after a reset, people are playing more ranked than before a season end, 100%.
|
On October 30 2015 03:17 MaestroSC wrote:
MMR and ELO dont just give you a metric to measure yourself... they also give you a competitive goal you can set for yourself.
And I dont agree in ANY way whatsoever that resetting a players MMR/ELO is a negative to ANY player.
Everyone LOVES the chance to start over and a clean slate.
Its probably the most popular time of League right after a reset, people are playing more ranked than before a season end, 100%.
League is not a complete reset. it is a sort reset based on brackets.
SIlver gets reset to 1200, Diamond to 1600 ect (numbers fictional)
A full reset is negative because it means for the next few days the Pro's are going to be stomping the Casual's face into the dirt and that makes the Casual want to not play.
|
Got beta and played some games. Looks fun, but i will wait until they put a competive mode.
|
On October 30 2015 03:02 Plansix wrote:Show nested quote +On October 30 2015 02:53 jcarlsoniv wrote:On October 30 2015 02:46 Plansix wrote:On October 30 2015 02:37 Hider wrote:Yeah and I only play soloQ games. So showing my ability to win soloQ games is exactly what I want from it. Why? Because it makes me care if I win said soloQ game or not. If the system makes me win 50% of my games no matter what I do and does not reflect me winning n any way, I have no reason to care about winning a particular. Then I will rather try to get kills or whatever instead of going for objectives which worsens the game experience overall for everyone involved. Yes that's my stand to, and I think that Plansix looks at "improvement" from an esports-perspective and not a casual perspective. It seems that the only thing that matters is if youimprove your skillset in a way that's relevant for competitive play. But what's relevant for most of us is that we improve our skillset so we can climb in soloQ. As long as we have fun in the proces, the gamedesigner has done his job. Its also improvement over time and long term. Peoples MMR isn’t going to be reset often because it ruins the player experience for the entire player base. So if someone tries to play a new hero, their MMR does down. Get into a new mode or play style, MMR goes down. So the system punishes players for doing the specific things that will make them better long term. That heavily depends on how you define "punishes the player". If you're talking about short term MMR drops when trying new things, then yes. However, if the purpose of trying new things is to improve your play holistically, then that short term drop shouldn't matter because, ideally, it would result in long term gains. MMR benefits from a large number of games. The more games you play, the more accurate your relative MMR will be. If you get better, your number trends upwards. As for ladder anxiety - I don't disagree with a Normal queue that has hidden MMR and a Ranked queue that makes MMR visible. If you're anxious about rank, play normals - it shouldn't matter to you anyway. If you want to play in a more structured, competitive setting, play ranked. So why is it better than a ladder or league system?
What makes it worse?
MMR is used for ladder systems' matchmaking, it's just obscured through simpler groupings. Personally, I have no real preference one way or the other, but have absolutely no issues with a more exact visible MMR instead (this is how it was in the early days of League).
Citing ladder anxiety as part of an argument against MMR is meaningless because it would be more or less the same in a ladder system.
|
|
|
|
|
|
EPT
