Issues currently with MMR based tier

Issues currently with MMR based tier

in PvP

Posted by: Xter.6271

Xter.6271

As much as I’d love to rant about the matchmaking, I’ll do something more constructive to vent my frustrations.

What is the issue with current matchmaking? Plenty. Are some people receiving the quality match ups they want? Yes. Are some people receiving match ups that are blow outs/no fun/frustrating as hell? Yes.

The current MMR system is based on team performance. It’s no secret and no surprise. You gain MMR if your team wins and lose MMR if your team loses.

Now the matchmaking has a large range it seems to draw from for MMR due to population, that way Queue times are to a minimum. You can find evidence for this on the forum and by conducting your own research. The problem is the matchmaking seems to draw from a too wide of range at times (noticeably off hours) to make fast queues.

Now this shouldn’t be an issue right? Faster queues, more games, more balanced MMRs across the board. Well the issue arrives in that the matchmaking seems to at times to stack teams (again, forum/reddit research, and you can conduct research yourself). As long as the average MMR of both teams is relatively the same, the matchmaking will pair all 10 players.

Now the huge problem in this, is that the one team can have 2 great players, 3 almost average players, matched against 5 average players.

Let’s say Team Red has; 1400, 1350, 1150, 1100, and 1100
Team Blue has; 1250, 1250, 1200, 1200, and an 1100

Red has an average of 1220 and Blue has an average of 1200. Seems like it would be a quality match up right? Well Red also has 2 players that are a great deal better. while having 1 player roughly equal to Blue’s best player and 2 players equal to Blue’s worst.

Lets say those two great players carry Red to a win, maybe even a blow out because they rotate and manage better while the other 3 guys just zerg, In theory we should be able to say the least that Blue will learn from Red’s win and pick up a few things and Red’s 3 other guys should have first hand experience of proper tactics for a win in conquest. Except that communication is often minimum before, during and after a match. That’s no secret, many people QQ in game, on the forums, and on reddit about team members not listening or responding.

You could say “Well, look. Red had a slight advantage, but blue was equally as good but just didn’t do what they should have. They should of rotated more, disengaged a losing fight, and know when to +1 and when to try another point.”

Fair enough, but let me point you to this for the lower divisions.

Team Red; 1050, 900, 900, 850, 700
Team Blue; 850, 850, 850, 750, 750

Red’s Average is 880
Blue’s Average is 810

These numbers are well within the matchmaking MMR’s limits. (Forum research, reddit, and do your own if you want again).

But this match up is destined to be a blow out for Blue. Why? Well Red has 4 players that are equal or greater then Blue’s best player(s). Red will only have to carry 1 player and Blue will have 2 to carry, plus Blue will be at a disadvantage due to having 1 player on Red being equal and 3 players better then Blue’s best 3 players.

Here lies the huge problem. In Bronze it’s easier for a team comp to be within the average limits yet still having one team stacked. And when you sum up your MMR gains or loses due to Team performance, some people literally can be stuck where they don’t belong. The system does not factor in what you as the individual has done during the match. It just looks at the fact that your team ether won or lost. Clean cut and dry. This issue can arise in other divisions because again, the MMR averages can be within the threshold but still stack teams.

I’m not going to lie, I started Silver T2 and suddenly just dropped to Bronze T3. I seem to start climbing back to my original ranking and then plummeted to Bronze T2 and I’m becoming extremely frustrated with almost every one of my teammates I’ve had. They’re not rotating, not calling targets, not +1 at critical fights, being meat fodder to the other team, double capping home at start, trying to 2v4 at Mid, not regrouping, no communication on what’s happening where, trying to fight away from points, camping at points when they could be used better else where, and some just not being able to play their class at all.

Issues currently with MMR based tier

in PvP

Posted by: Xter.6271

Xter.6271

I’m going to keep trying to move back up but at this point I’m frustrated my MMR is plummeting because of the teams I keep getting paired with. Yes I’m not perfect, I’m by no means a pro, and yes I know I need to L2P and get “gud”. But all I want is match ups where I feel we have a legit chance, players with the same competence as me at least, and to actually have a quality match.

When I enter a match and see 4 guys decked out in starter stuff, no guilds, no titles, and don’t seem to know how to type into the chat, all I think is “Great, here we go again. Do I switch classes and try to carry them the best I can? Do I just trust they can avoid being fodder? Am I just going to be typing into an empty team chat all match again trying to coordinate with them? Why doesn’t Anet allow tonics in sPvP, I could at least give the other team a laugh by charging as a Moa into mid and have some memorable experience of this match.”

Anet really needs to develop an MMR system that accounts for personal performance and team performance. Like this guy captured the most points, killed the most players, and his team won, +47 MMR. or, this guy killed the most players, decapped the most points, but his team lost, +7 MMR.

A system that rates you most actively for participation to winning conditions. If your team won, but you performed poorly (AKA; Died the most, didn’t kill anyone or help kill anyone, captured no points and decapped no points) you should gain little MMR or none, maybe even lose, to place you with players of equal skill. Let’s say you, performed average compared to your team mates, the match was close, but you lost. You should lose very little MMR because it means you’re around your skill level. Or you were exceptional compared to your team mates, you did way more, and your team won, you should gain a lot of MMR to propel you towards players of equal rating.

A system set up like this would be less frustrating for everyone because even in a losing condition team wise, you can still move up towards players around your skill level if you played well on a personal level. Players that perform poorly will move up very slowly and should fall back to their tier fast if they don’t self improve.

The biggest issue with the current system is “your” personal MMR is impacted by your team. Whether they are worse then you, better then you, or equal skill wise. In a system that is random, you control no variables too outside of duo queuing with someone you trust, the worst thing is to affect everyone’s MMR on the losing team due to 1 player ruining it, or 4.

My best suggestion is to move to a system that rates personal performance, and team performance to help move everyone where they should be respectively and attempt to provide everyone with better match up qualities. We don’t need everyone to have constant win streaks, but we need to make the losing streaks to a fairer match so we can at least walk away from defeat and say “If I wouldn’t of done this, and done that instead, maybe we could have won the match” and therefore hopefully provide everyone with a better experience and better insight on what to do next time.

Sincerely,
X

Issues currently with MMR based tier

in PvP

Posted by: BlaqueFyre.5678

BlaqueFyre.5678

Sounds like they need to scrap Solo queue because it’s incompatible with the organized team gamemode known as sPvP, and implement a pure 5v5 team Queue, and a 2/3 mans team Queue as well. Boom all problems solved with matchmaker especially once they remove class stacking, queue times should never be a reason for compromised Matchmaking.

Issues currently with MMR based tier

in PvP

Posted by: Faux Play.6104

Faux Play.6104

Wins are the best metric. It is simple, and the people that are the most consistent and helping their teams win will end up with the highest ratings.

Example you are on a 2v2 at home and it is neutral. You down the 1st, and both you and your teammate are in good shape health wise and the other player is 30%. It is a favorable match up 1v1 for your teammate even if the other player was at 100% health. The best play for the team would be to leave and help another fight, but you are going to miss out on a kill and a capture. What do you do if it is individual score based?

Issues currently with MMR based tier

in PvP

Posted by: Cynz.9437

Cynz.9437

Lol it is even worse at higher ratings since half of 250 are alt accounts/season afkers so MM can’t find equal teams so you have 4 top 10 players on one team and full team of 1700-1800 on other.

What needs to happen:
1. players shouldn’t be placed higher than 1700
2. you need to play at least 100 matches to qualify for top 250
3. rating gains/loses should depend on how close the match was. It will give motivation for players to actually try. Losing 20 ranks for 499-500 match is extremely discouraging.

All is Vain~
[Teef] guild :>

Issues currently with MMR based tier

in PvP

Posted by: Lordrosicky.5813

Lordrosicky.5813

Lol it is even worse at higher ratings since half of 250 are alt accounts/season afkers so MM can’t find equal teams so you have 4 top 10 players on one team and full team of 1700-1800 on other.

What needs to happen:
1. players shouldn’t be placed higher than 1700
2. you need to play at least 100 matches to qualify for top 250
3. rating gains/loses should depend on how close the match was. It will give motivation for players to actually try. Losing 20 ranks for 499-500 match is extremely discouraging.

I agree. Not sure on 3 though. Dont think i like it. People will still afk and it will suck even more when you lose 30 pts cos the game is unbalanced and you lose 500-0

As for the op. I placed at 1040. It is impossible to get stuck at bronze unless you belong there. People are so deluded as to their own skill level.

Character: Henry rank million/Duke Henry
Necromancer/Casual Warrior
[Team] Best WvW guild of all time. EASILY.

Issues currently with MMR based tier

in PvP

Posted by: Pakkazull.6894

Pakkazull.6894

Anet really needs to develop an MMR system that accounts for personal performance and team performance.

I doubt ArenaNet has the expertise, time or money to develop such a system, nor do I think it’s feasible at all.

Like this guy captured the most points, killed the most players, and his team won, +47 MMR. or, this guy killed the most players, decapped the most points, but his team lost, +7 MMR.

Bad idea. None of those metrics can accurately capture the impact of a player on a game, and furthermore can be gamed which could encourage undesirable behaviour (such as several people staying on a point just to get the cap).

“If I wouldn’t of done this, and done that instead, maybe we could have won the match”

Wouldn’t have done this*. Sorry, pet peeve of mine.

Sounds like they need to scrap Solo queue because it’s incompatible with the organized team gamemode known as sPvP, and implement a pure 5v5 team Queue, and a 2/3 mans team Queue as well. Boom all problems solved with matchmaker especially once they remove class stacking, queue times should never be a reason for compromised Matchmaking.

Terrible idea, you’d completely kill sPvP in one swoop. I agree that teams need a competitive venue though, but that can be handled separately from the league system.

Issues currently with MMR based tier

in PvP

Posted by: Wolfric.9380

Wolfric.9380

It is more of the problem that one player drags down the team. It might be an AFK/disconect but more often it is a bad player that just feeds and doesnt listen. For me it often comes down to which side gets the worst one.
Examples. We got a DH on our side (forest). Seeing oponent has two thiefs writing DH close thief will come. Well basically expecting DH can hold the point… thief comes killing DH in 10 seconds, stealing beast going back killing respawend DH again.
Game over you know this DH is nearly worthless. Since oposing team is not dump no way to carry or win. The situation snowballs by someone trying to help out close, oposing respawn wins 3:4 mid. Chaos spreads and its a blowout.
Other way around. Team two necros, me ele, 1 war 1 scrapper. Oponent is good mix i feared we loose with 2 necro. We send scrapper close…. It showed he was a good one. Oposing mesmer attacks but scrapper held point stalling him endless and even could manage a +1 for a shot while till i could arrive. This stable spot sowballed the match in the oposite direction. I could keep the necros with ele mid and a blowout for us was the result. Warrior was even free for a moment to decap far ….
its a 10:10 teamgame but one can make it uncarryable affacting all MMR´s.

100 MMR is not realy much skill diffrence. I can move within 300 MMR range from low silver up and don´t feel a diffrence. OK match quality increases a bit mid silver but results are often uncarriable and blowouts.
I encounter about 2/3 of all matches are 200+ diffrence.

Such a match is usually a disaster for blue:

Team Red; 1200, 1100, 1100, 1100, 1000
Team Blue; 1300, 1200, 1200, 1100, 700

Red’s Average is 1100
Blue’s Average is 1100

MMR loss is significant and the 1300 is unable to carry the 700 that does feed all 30 seconds and the result that might happen if someone tries to compensate.

In adition never place two duo on one side and five solo the other! The solo average can even be better and will still have the odds against them.

Issues currently with MMR based tier

in PvP

Posted by: Xter.6271

Xter.6271

Wins are the best metric. It is simple, and the people that are the most consistent and helping their teams win will end up with the highest ratings.

Example you are on a 2v2 at home and it is neutral. You down the 1st, and both you and your teammate are in good shape health wise and the other player is 30%. It is a favorable match up 1v1 for your teammate even if the other player was at 100% health. The best play for the team would be to leave and help another fight, but you are going to miss out on a kill and a capture. What do you do if it is individual score based?

Have the system account for kill assisting and divide the solo kill performance metric by the number of teammates helping with the kill if they contributed enough damage or support via boons, auras, etc to your teammates.

Onto the capture, have the system factor in enemy range, teammates on point, and whether one player is wasting time or if there is a probable cause like an enemy in range or combat is currently active via damage being dealt. If none is of these are happening then the system will give nothing to both players. This doesn’t mean they’ll lose MMR but nothing is factored into the personal performance for a lack of anything. If someone +1 in a fight to get that point and that player leaves while one player caps that point then both will receive the merit for the capture being it was a team based captured.

Anet really needs to develop an MMR system that accounts for personal performance and team performance.

I doubt ArenaNet has the expertise, time or money to develop such a system, nor do I think it’s feasible at all.

Like this guy captured the most points, killed the most players, and his team won, +47 MMR. or, this guy killed the most players, decapped the most points, but his team lost, +7 MMR.

Bad idea. None of those metrics can accurately capture the impact of a player on a game, and furthermore can be gamed which could encourage undesirable behaviour (such as several people staying on a point just to get the cap).

“If I wouldn’t of done this, and done that instead, maybe we could have won the match”

Wouldn’t have done this*. Sorry, pet peeve of mine.

Sounds like they need to scrap Solo queue because it’s incompatible with the organized team gamemode known as sPvP, and implement a pure 5v5 team Queue, and a 2/3 mans team Queue as well. Boom all problems solved with matchmaker especially once they remove class stacking, queue times should never be a reason for compromised Matchmaking.

Terrible idea, you’d completely kill sPvP in one swoop. I agree that teams need a competitive venue though, but that can be handled separately from the league system.

I think some of my ideas would help. I don’t have the knowledge or expertise to balance a system like this. My +47 MMR gain was just an example. The numbers and metrics would have to be balanced.

It would be up to Anet to decide if such a system was possible and if they’d want to take the plunge. I’m just picthing an idea and hopefully or maybe they’ll get some inspiration or at the minimum decided this idea was a good try but not worth it to pursue.

I tried to elaborate some on how such a system may work in certain circumstances to reward players for doing well, while not giving them anything favorable for trying to abuse it.

Maybe my idea wouldn’t factor hugely if your team won, but factors into it if you lose so if you play well the MMR lost is to a minimum and you can’t tumble down the ratings as fast for playing well on your end, being you rely on 4 other random people.

In a team queue like 5 man premade, if we got such a thing, should be entirely based on a win/lose condition being you have plenty of team to coordinate, resources like voice comm. to help, and everyone is geared towards building a functional team at the start and coordinating themselves to adapt at the match start if need be/they choose.

Issues currently with MMR based tier

in PvP

Posted by: Fay.2357

Fay.2357

Xfer, you’re making the faulty assumption that matchmaking is out to get you and is always placing you in the losing team of those matchups. The reality is quite different. You’re right that MMR matching will not produce perfect matches every time. However, you have an equal chance of being on the lucky team as you do of being on the unlucky team. Over time these average out and end up having no long term effect on your rating.

Issues currently with MMR based tier

in PvP

Posted by: Xter.6271

Xter.6271

Xfer, you’re making the faulty assumption that matchmaking is out to get you and is always placing you in the losing team of those matchups. The reality is quite different. You’re right that MMR matching will not produce perfect matches every time. However, you have an equal chance of being on the lucky team as you do of being on the unlucky team. Over time these average out and end up having no long term effect on your rating.

No I’m not. I have had games with great team mates and it was fun at the start till it just turned into a blow out and we literally had 3 point caps and the other team showed no effort into doing something. They’d ether respawn and be fodder, or gave up.

Now the match went from being enjoyable at the start because I had players that wanted to try and win, communicated, rotated, and did their absolute best, but the other team was awful and blowing out the opposing team isn’t really fun when you’re just waiting on 500 and killing the same guys over and over.

I’m just trying to propose something that will help the current matchmaking place people with people of similar skill quickly and hopefully create better matches. I personally don’t care too much about losing, but I get irritated when I lose and drop down farther and get even worse team mates that seem to have just started the game 5 hours ago. Yes, eventually I’ll climb out and hopefully be in a better division. But it’s going to be painful and frustrating doing so knowing it’s probably going to take 30 matches of the same thing over and over.

Issues currently with MMR based tier

in PvP

Posted by: Fay.2357

Fay.2357

Xfer, you’re making the faulty assumption that matchmaking is out to get you and is always placing you in the losing team of those matchups. The reality is quite different. You’re right that MMR matching will not produce perfect matches every time. However, you have an equal chance of being on the lucky team as you do of being on the unlucky team. Over time these average out and end up having no long term effect on your rating.

No I’m not. I have had games with great team mates and it was fun at the start till it just turned into a blow out and we literally had 3 point caps and the other team showed no effort into doing something. They’d ether respawn and be fodder, or gave up.

Now the match went from being enjoyable at the start because I had players that wanted to try and win, communicated, rotated, and did their absolute best, but the other team was awful and blowing out the opposing team isn’t really fun when you’re just waiting on 500 and killing the same guys over and over.

I’m just trying to propose something that will help the current matchmaking place people with people of similar skill quickly and hopefully create better matches. I personally don’t care too much about losing, but I get irritated when I lose and drop down farther and get even worse team mates that seem to have just started the game 5 hours ago. Yes, eventually I’ll climb out and hopefully be in a better division. But it’s going to be painful and frustrating doing so knowing it’s probably going to take 30 matches of the same thing over and over.

Well, correct me if I’m reading this wrong, but your whole argument seems totally pointless. You start by using MMR examples as a demonstration as to how unbalanced matches can be formed. This inherently also means that the MMR used for these matches is accurate, but the system is using it wrong. You then go on to talk about how the system ends up evaluating your MMR totally wrong and you introduce some suggestions to evaluate it better.

Pick one, because you can’t have both. Either the MMR is accurate and it’s the matching that is bad or the MMR is bad itself. But as is, your arguments are logically inconsistent with each other. Let me give an example:

You claim that the system has artificially lowered your MMR more than it should have. If that’s true, you’ll end up in a match like your second example… except now team blue has an ace up their sleeve. You’re that guy with the 750 MMR, but your actual rating is closer to 900. In this case, team blue should dominate the match because you are much stronger than the matchmaker thinks you are. This will keep happening until your MMR has risen to the appropriate level.

And so now we see how matchmaking works fine. Even if your MMR somehow gets lowered more than it should, the system is self-correcting. Even though the system doesn’t make perfect matches every time, it still automatically irons out improper MMR ratings until they’re correct. Your arguments are inconsistent with each other AND inconsistent with reality. I recommend finding some new ones.

(edited by Fay.2357)

Issues currently with MMR based tier

in PvP

Posted by: hackks.3687

hackks.3687

I started in Bronze 2 after abysmal placement matches, climbed to Gold 2 quickly, then within the last weekend the match making went to complete kitten and now I just can’t buy a win. I call all the targets, all the rezzes, all the rotations, enemy numbers and movements, I win 1v2’s even. But what I can’t do is win the other fight for my teams second point, too. As soon as I rotate, the fight is lost and my team respawns and scatters to the wind even though I ask them to regroup. Matches have turned into blowouts regardless of my effort. And even when they remain competitive, it’s inevitable that someone will make a wild mistake and throw the match.

I don’t have this problem to this degree in any other game. Any of them.

What’s wrong? As OP states, the player draw is from too wide a range, which skews the actual skill rating for each team. Worse than that though, I’m sharing a rating with the people who are making the horrible mistakes ONLY because I can’t gain enough rating separation from because there’s no criteria for individual contribution.

Now if you think ANet can’t apply some criteria to include individual contribution then I kindly invite you to take your ignorant kitten elsewhere, because these systems already exist. Go play any other competitive game and you’ll find out pretty quick just how ANet’s system is deficient.

Also, there is a way to make sure the progression is positive rather than negative and to keep people from tending to their ratings by only playing the minimum number of games to prevent rating decay. It’s simple: Everyone places into bronze. Period. Now everyone has to play to climb. Not just those of us who didn’t get the fortune of great placement teams.

Hackkz/Riggamaroll
I’ve stayed at this party entirely too long

Issues currently with MMR based tier

in PvP

Posted by: Fay.2357

Fay.2357

Now if you think ANet can’t apply some criteria to include individual contribution then I kindly invite you to take your ignorant kitten elsewhere, because these systems already exist. Go play any other competitive game and you’ll find out pretty quick just how ANet’s system is deficient.

Yeah! Just look at League of Legends, the most successful eSport of all time! Their matching system….Oh wait, their matching system actually works exactly the same.

Issues currently with MMR based tier

in PvP

Posted by: megilandil.7506

megilandil.7506

Lol it is even worse at higher ratings since half of 250 are alt accounts/season afkers so MM can’t find equal teams so you have 4 top 10 players on one team and full team of 1700-1800 on other.

What needs to happen:
1. players shouldn’t be placed higher than 1700
2. you need to play at least 100 matches to qualify for top 250
3. rating gains/loses should depend on how close the match was. It will give motivation for players to actually try. Losing 20 ranks for 499-500 match is extremely discouraging.

3 its a big system problem, a close match is a fair match and the one that works for calculating properly the “skill” of players a blowout is a system failure and counting it for calculations is unfair, something go “wrong” with the calculations of mm and losing team is in an unwinnable match, the only form to adress this is separating the rank from mmr and this is another version of the old pip system, blow ot lose small amount of mmr lose big rank, close encounter lose correct mmr lose small rank, or directly change the complete system, to a true league ofseason are clasificatorys and seasons players fighting to be the best of their (being randomly matched whith any player of your tier r puted in the lowest tier)league whithout league climbings (but that have the problem of trolling alts of “pros” being champions on all tiers)

Issues currently with MMR based tier

in PvP

Posted by: hackks.3687

hackks.3687

Now if you think ANet can’t apply some criteria to include individual contribution then I kindly invite you to take your ignorant kitten elsewhere, because these systems already exist. Go play any other competitive game and you’ll find out pretty quick just how ANet’s system is deficient.

Yeah! Just look at League of Legends, the most successful eSport of all time! Their matching system….Oh wait, their matching system actually works exactly the same.

The player pool in LoL is MASSIVE compared to GW2 and can draw from much narrower range of skill ratings as a result without sacrificing queue times – which is why it works, and also why it doesn’t in GW2…

CS:GO has MVP system.

Get informed. You just sound stupid.

Hackkz/Riggamaroll
I’ve stayed at this party entirely too long

Issues currently with MMR based tier

in PvP

Posted by: Pakkazull.6894

Pakkazull.6894

Now if you think ANet can’t apply some criteria to include individual contribution then I kindly invite you to take your ignorant kitten elsewhere, because these systems already exist. Go play any other competitive game and you’ll find out pretty quick just how ANet’s system is deficient.

Please, enlighten me, because I barely know of any game that uses any other criteria than wins and losses. And yes, CS:GO has an MVP system, but it’s also easier to implement in CS:GO since the only real metrics it uses are kills/assists and plants/defuses (and even that system isn’t entirely accurate. You’ll get MVP for planting the bomb even if someone else carried hard and killed 4 guys solo to get you to the bomb site).

It’s simple: Everyone places into bronze. Period. Now everyone has to play to climb. Not just those of us who didn’t get the fortune of great placement teams.

Well, that’s the silliest suggestion I’ve seen in this thread so far. You’d basically condemn everyone to two months of awful matches out of spite, because you feel you’ve been wronged by “the system”.

(edited by Pakkazull.6894)

Issues currently with MMR based tier

in PvP

Posted by: megilandil.7506

megilandil.7506

other implementation that can work for preventing to much blowouts in the first days/weeks, when players are misplaced, is making supositions and little or not so little adjustemments in the mmr number used in mm calculations, a player on a winstreak supose that has a higher mmr than the actual one, a player in a lose streak supose lower and a player stabilized in a near 50% w/l the actual that can prevent the actual high volatility in first weeks where you get misplaced players that are falling ranks or climbing fast and have a big impact in the outcome of your match even it is fair in numbers

Issues currently with MMR based tier

in PvP

Posted by: hackks.3687

hackks.3687

Now if you think ANet can’t apply some criteria to include individual contribution then I kindly invite you to take your ignorant kitten elsewhere, because these systems already exist. Go play any other competitive game and you’ll find out pretty quick just how ANet’s system is deficient.

Please, enlighten me, because I barely know of any game that uses any other criteria than wins and losses. And yes, CS:GO has an MVP system, but it’s also easier to implement in CS:GO since the only real metrics it uses are kills/assists and plants/defuses (and even that system isn’t entirely accurate. You’ll get MVP for planting the bomb even if someone else carried hard and killed 4 guys solo to get you to the bomb site).

It’s simple: Everyone places into bronze. Period. Now everyone has to play to climb. Not just those of us who didn’t get the fortune of great placement teams.

Well, that’s the silliest suggestion I’ve seen in this thread so far. You’d basically condemn everyone to two months of awful matches out of spite, because you feel you’ve been wronged by “the system”.

What exactly are you advocating for? You want a system that protects you from having to play kittenty matches, but not everyone else?

And you’re reason for not doing anything apparently is because it’s hard and not as easy as copy/paste? We don’t have metrics for our game? Reliable stats that point to winning behavior? Kills/Deaths/Revives/Caps/Decaps/Defense? Just because ANet isn’t using them correctly doesn’t mean that can’t change.

What is this point of this post? Have you ever made anything????

Hackkz/Riggamaroll
I’ve stayed at this party entirely too long

Issues currently with MMR based tier

in PvP

Posted by: megilandil.7506

megilandil.7506

Now if you think ANet can’t apply some criteria to include individual contribution then I kindly invite you to take your ignorant kitten elsewhere, because these systems already exist. Go play any other competitive game and you’ll find out pretty quick just how ANet’s system is deficient.

Please, enlighten me, because I barely know of any game that uses any other criteria than wins and losses. And yes, CS:GO has an MVP system, but it’s also easier to implement in CS:GO since the only real metrics it uses are kills/assists and plants/defuses (and even that system isn’t entirely accurate. You’ll get MVP for planting the bomb even if someone else carried hard and killed 4 guys solo to get you to the bomb site).

It’s simple: Everyone places into bronze. Period. Now everyone has to play to climb. Not just those of us who didn’t get the fortune of great placement teams.

Well, that’s the silliest suggestion I’ve seen in this thread so far. You’d basically condemn everyone to two months of awful matches out of spite, because you feel you’ve been wronged by “the system”.

What exactly are you advocating for? You want a system that protects you from having to play kittenty matches, but not everyone else?

And you’re reason for not doing anything apparently is because it’s hard and not as easy as copy/paste? We don’t have metrics for our game? Reliable stats that point to winning behavior? Kills/Deaths/Revives/Caps/Decaps/Defense? Just because ANet isn’t using them correctly doesn’t mean that can’t change.

What is this point of this post? Have you ever made anything????

caps——> doublecaping
kills—→ teamfigters get more deads than roamers
defenses——>the same
deaths—→a holder dead more often when outnumbered but let the other players to be numerical superior on map
revives——> teamfighters get more and not all the classes revive whith the same easines
caps/decaps——→roamers get more than teamfighters
these stats are highly dependant on role on match not on playing good

Issues currently with MMR based tier

in PvP

Posted by: Pakkazull.6894

Pakkazull.6894

What exactly are you advocating for?

I think the system is fine as it is, provided they add more aggressive decay to the top divisions and a x number of games played requirement to place on the leaderboard. So I guess that’s what I’m “advocating for”.

You want a system that protects you from having to play kittenty matches, but not everyone else?

No, I don’t know where you’d get that idea, but strawmen are always nice I guess. I was merely remarking on how your suggestion seemed driven purely by spite.

And you’re reason for not doing anything apparently is because it’s hard and not as easy as copy/paste? We don’t have metrics for our game? Reliable stats that point to winning behavior? Kills/Deaths/Revives/Caps/Decaps/Defense? Just because ANet isn’t using them correctly doesn’t mean that can’t change.

We have metrics, yes, but that’s exactly the point; we have a lot more metrics than CS:GO, which is why making a system that accurately reflects the impact a player has on a game is much harder. Furthermore, you can’t game CS:GO’s metrics (except for hogging the bomb I guess?), whereas you can definitely game the GW2 metrics you listed.

Ultimately it’s up to ArenaNet if they want to try or not, but in my opinion it’s prohibitively difficult to get right and needless when wins/losses are a good enough metric already.

What is this point of this post? Have you ever made anything????

I’m not sure what you mean. Have I made anything? Like what, woodwork?

(edited by Pakkazull.6894)

Issues currently with MMR based tier

in PvP

Posted by: Rhiannon.1726

Rhiannon.1726

How do you measure individual skill? “Skill” in pvp isn’t determined by top stats.

I’ll try to make examples:

Besides you there is another necro in your team. You are more skilled.
The enemy prefers to focus you the whole time (because they know you are more dangerous, you are a Charr, or you have pink hair…).
You have to kite a lot and maybe you die several times and the other necro has it way more easier to attack the others and gets credit for dps and kills, because he is mostly ignored.
So you are skilled but because of bad luck the other necro gets rewarded?

There is a tempest and a druid in one team. The druid is more skilled. The druid is nearly always defending home against a thief and the tempest is in mid where the most fights are. The druid doesn’t heal anyone else, doesn’t kill anyone, doesn’t do the most damage and doesn’t capture any points (only close once).
The tempest is constantly healing others (by randomly spamming auras…), captures mid more than once and takes part in several kills/rezzes.
So how does the system know who is better?

Issues currently with MMR based tier

in PvP

Posted by: Nightshade.2570

Nightshade.2570

I have noticed the quickening of pace with which matches are made. In effect the average queue time has been sucessfully reduced compared to prior seasons. Now, I think this a great thing except, I fear that in lower divisions of match making, gold and below this may result in more, poorly matched teams. IE, quickly matched but not Well matched. Now, this would not cause a problem during, high peak times in PvP, this would be of no issue when the player pool is vast and wide. IE, more players good matches are wrought.

At the beginning of the season the matches were very good, it seemed the achievement of a balanced match regularly was not so far out of grasp. The strides of doing placement matches were great and I not only did well in my placements but held my general rating for a month at least. Sadly I noticed that as time went on and the newness of the PvP season wore off. The player pool lowers and suddenly there are not as many people for the match making system to select from.

Yet, I still see fast matchmaking queue times and in fact suspect that queue times take precedence over the quality now of the match, more then they did before.

Has this been adjusted? I can not say, I understand the want for fast queue times, I understand the complaints in past seasons over it getting longer as the seasons ended.

I simply must ask, has this been changed this season? As it has become impossible for me to hold my spot where I was in the past week or so, and the change I suspect has to do with the waning population pools in PvP. The broadening of matchmaking searches, happening sooner.

Issues currently with MMR based tier

in PvP

Posted by: SlippyCheeze.5483

SlippyCheeze.5483

Well, that’s the silliest suggestion I’ve seen in this thread so far. You’d basically condemn everyone to two months of awful matches out of spite, because you feel you’ve been wronged by “the system”.

…not even: it would only take ten to twenty games to put them roughly back at the same MMR they have now, because (shocking news!) the gliko algorithm converges fairly quickly on the target MMR.

The biggest thing that distinguishes it from ELO or other algorithms, in fact, is the volatility estimate, which roughly speaking translates to “how much does a surprise win or loss move your MMR estimate”, and reduces as you stabilize.

Issues currently with MMR based tier

in PvP

Posted by: hackks.3687

hackks.3687

What exactly are you advocating for?

I think the system is fine as it is, provided they add more aggressive decay to the top divisions and a x number of games played requirement to place on the leaderboard. So I guess that’s what I’m “advocating for”.

You want a system that protects you from having to play kittenty matches, but not everyone else?

No, I don’t know where you’d get that idea, but strawmen are always nice I guess. I was merely remarking on how your suggestion seemed driven purely by spite.

And you’re reason for not doing anything apparently is because it’s hard and not as easy as copy/paste? We don’t have metrics for our game? Reliable stats that point to winning behavior? Kills/Deaths/Revives/Caps/Decaps/Defense? Just because ANet isn’t using them correctly doesn’t mean that can’t change.

We have metrics, yes, but that’s exactly the point; we have a lot more metrics than CS:GO, which is why making a system that accurately reflects the impact a player has on a game is much harder. Furthermore, you can’t game CS:GO’s metrics (except for hogging the bomb I guess?), whereas you can definitely game the GW2 metrics you listed.

Ultimately it’s up to ArenaNet if they want to try or not, but in my opinion it’s prohibitively difficult to get right and needless when wins/losses are a good enough metric already.

What is this point of this post? Have you ever made anything????

I’m not sure what you mean. Have I made anything? Like what, woodwork?

What I’m asking is if you’ve ever contributed to producing a product for mass consumption. You’ve already made it clear you don’t as you’re advocating for a status quo that diminishes the quality of the end product and there by wastes any investment for reaching this point in PvP’s life cycle. It makes no sense to stop the product changes here, which is your suggestion essentially – do nothing.

On top of which none of what you’re saying is a reason not to do these things. Just because there are problematic variables doesn’t mean they can’t be accommodated and addressed. You simply want to perpetuate an unhealthy system that benefits you even though it’s obviously faulty and the continued decline of the player population is your proof – not that you’re interested in facts or reality – and will likely drive you from the game eventually, too. Yes, I’m saying your suggestion and viewpoint are self-destructive and not even for your own benefit.

The irony is that you, as well as large portions of the player base, would likely benefit from these changes as well. If you were a truly skilled player, you wouldn’t care about the weighting of the system, you’d just want good match ups. I sense that’s not the case.

Thanks for your self-centered and myopic viewpoint, though…

Hackkz/Riggamaroll
I’ve stayed at this party entirely too long

Issues currently with MMR based tier

in PvP

Posted by: Pakkazull.6894

Pakkazull.6894

…not even: it would only take ten to twenty games to put them roughly back at the same MMR they have now, because (shocking news!) the gliko algorithm converges fairly quickly on the target MMR.

I could see that being the case if a high MMR player was artificially lowered to bronze and inserted into an active system, but if everyone was reduced to bronze level MMR? I dunno, you might be right, I’m not pretending to be an expert.

What I’m asking is if you’ve ever contributed to producing a product for mass consumption. You’ve already made it clear you don’t

I don’t see how it matters in the slightest, but I sense that you just want to put me down and claim some sort of superiority, so… go ahead I guess.

It makes no sense to stop the product changes here, which is your suggestion essentially – do nothing.

I see reading comprehension isn’t one of your strong suits, because I suggested two refinements.

On top of which none of what you’re saying is a reason not to do these things. Just because there are problematic variables doesn’t mean they can’t be accommodated and addressed.

No, I honestly don’t think they can be adequately addressed — I’d gladly be proven wrong though.

You simply want to perpetuate an unhealthy system that benefits you even though it’s obviously faulty and the continued decline of the player population is your proof – not that you’re interested in facts or reality

Hilarious, considering you have no proof of “decline in population”. Only ArenaNet knows for sure; the rest of us have only anecdotal evidence.

(edited by Pakkazull.6894)

Issues currently with MMR based tier

in PvP

Posted by: hackks.3687

hackks.3687

No Proof of low player population?

Why else do you end up with matches taking minutes to pop and yet end up with the same players as the previous match?

And when the match finally pops, why is it do you think that the players in the match that took so long to pop are all wildly different ratings? Perhaps because ANet had to make it possible in order to ensure matches popped more frequently than every 20 minutes?

Let’s try this another way:

2 + 2 = ?

Hackkz/Riggamaroll
I’ve stayed at this party entirely too long

(edited by hackks.3687)

Issues currently with MMR based tier

in PvP

Posted by: Pakkazull.6894

Pakkazull.6894

No Proof of low player population?

Why else do you end up with matches taking minutes to pop and yet end up with the same players as the previous match?

And when the match finally pops, why is it do you think that the players in the match that took so long to pop are all wildly different ratings? Perhaps because ANet had to make it possible in order to ensure matches popped more frequently than every 20 minutes?

Yeah, like I said, anecdotal evidence. What you’re saying is entirely possible, but we don’t know if that holds true across all timezones on both NA and EU servers and across all rating ranges. Only ArenaNet has the entire picture.

Let’s try this another way:

2 + 2 = ?

Yes, your condescension is funny.

Issues currently with MMR based tier

in PvP

Posted by: hackks.3687

hackks.3687

No Proof of low player population?

Why else do you end up with matches taking minutes to pop and yet end up with the same players as the previous match?

And when the match finally pops, why is it do you think that the players in the match that took so long to pop are all wildly different ratings? Perhaps because ANet had to make it possible in order to ensure matches popped more frequently than every 20 minutes?

Yeah, like I said, anecdotal evidence. What you’re saying is entirely possible, but we don’t know if that holds true across all timezones on both NA and EU servers and across all rating ranges. Only ArenaNet has the entire picture.

Let’s try this another way:

2 + 2 = ?

Yes, your condescension is funny.

Anecdotal? They’ve said as much on numerous occasions that the rating deviation on match ups is so large in order to improve the long queue times due to low player population.

You can go look at the algorithm and formula yourself. It’s right there on the wiki.

I’m sure its fun to be contrarian but at some point it helps to know what you’re talking about and what’s implied within the specific context.

This is about the furthest thing from a reasonable discussion, so I’ll bow out here.

Hackkz/Riggamaroll
I’ve stayed at this party entirely too long

Issues currently with MMR based tier

in PvP

Posted by: Pakkazull.6894

Pakkazull.6894

Anecdotal?

Yes, the “proof” you presented is anecdotal. Do you know what it means?

They’ve said as much on numerous occasions that the rating deviation on match ups is so large in order to improve the long queue times due to low player population.

It’s funny how you originally said “continued decline in population” and then changed it to low population. There’s a difference.

This is about the furthest thing from a reasonable discussion, so I’ll bow out here.

That’s the first bright idea you’ve had today, I think.

Issues currently with MMR based tier

in PvP

Posted by: hackks.3687

hackks.3687

Anecdotal?

Yes, the “proof” you presented is anecdotal. Do you know what it means?

They’ve said as much on numerous occasions that the rating deviation on match ups is so large in order to improve the long queue times due to low player population.

It’s funny how you originally said “continued decline in population” and then changed it to low population. There’s a difference.

This is about the furthest thing from a reasonable discussion, so I’ll bow out here.

That’s the first bright idea you’ve had today, I think.

How do you get to a “low population” from a large population? You have to have ’decline’…

This must be what it feels like trying to teach chimps to read…

Hackkz/Riggamaroll
I’ve stayed at this party entirely too long

Issues currently with MMR based tier

in PvP

Posted by: Pakkazull.6894

Pakkazull.6894

How do you get to a “low population” from a large population? You have to have decline…

This must be what it feels like trying to teach chimps to read…

sPvP has never had a large population, except for the influx of reward-farming PvE players at the start of each season. So yes, in that sense, there is an ebb and flow to the population… but you said “continued decline”, so that’s clearly not what you meant.

Also, you’re still here? I thought you “bowed out”.

Issues currently with MMR based tier

in PvP

Posted by: morrolan.9608

morrolan.9608

Now the huge problem in this, is that the one team can have 2 great players, 3 almost average players, matched against 5 average players.

Let’s say Team Red has; 1400, 1350, 1150, 1100, and 1100
Team Blue has; 1250, 1250, 1200, 1200, and an 1100

Red has an average of 1220 and Blue has an average of 1200. Seems like it would be a quality match up right? Well Red also has 2 players that are a great deal better. while having 1 player roughly equal to Blue’s best player and 2 players equal to Blue’s worst.

Analysis by some players suggest teams aren’t being broken up like this, that the system is matching up 5 players for a team then finding another team with the closest mmr geometric average (as in S2 and s3), hence the teams are more likely to be:

Red: 1400, 1350, 1250, 1250, 1200
Blue: 1200, 1150, 1100, 1100, 1100

Jade Quarry [SoX]
Miranda Zero – Ele / Twitch Zero – Mes / Chargrin Soulboom – Engi
Aliera Zero – Guardian / Reaver Zero – Necro

Issues currently with MMR based tier

in PvP

Posted by: hackks.3687

hackks.3687

How do you get to a “low population” from a large population? You have to have decline…

This must be what it feels like trying to teach chimps to read…

sPvP has never had a large population, except for the influx of reward-farming PvE players at the start of each season. So yes, in that sense, there is an ebb and flow to the population… but you said “continued decline”, so that’s clearly not what you meant.

Also, you’re still here? I thought you “bowed out”.

Prove it.

Present your facts that PvP has always had a low population going all the way back to launch.

Hackkz/Riggamaroll
I’ve stayed at this party entirely too long

Issues currently with MMR based tier

in PvP

Posted by: Pakkazull.6894

Pakkazull.6894

Prove it.

Present your facts that PvP has always had a low population going all the way back to launch.

You’re right, I can’t. My evidence is just as anecdotal as yours.

Issues currently with MMR based tier

in PvP

Posted by: Xter.6271

Xter.6271

How do you measure individual skill? “Skill” in pvp isn’t determined by top stats.

I’ll try to make examples:

Besides you there is another necro in your team. You are more skilled.
The enemy prefers to focus you the whole time (because they know you are more dangerous, you are a Charr, or you have pink hair…).
You have to kite a lot and maybe you die several times and the other necro has it way more easier to attack the others and gets credit for dps and kills, because he is mostly ignored.
So you are skilled but because of bad luck the other necro gets rewarded?

There is a tempest and a druid in one team. The druid is more skilled. The druid is nearly always defending home against a thief and the tempest is in mid where the most fights are. The druid doesn’t heal anyone else, doesn’t kill anyone, doesn’t do the most damage and doesn’t capture any points (only close once).
The tempest is constantly healing others (by randomly spamming auras…), captures mid more than once and takes part in several kills/rezzes.
So how does the system know who is better?

On the first example;

Maybe a kill and kill assist type system. If you helped deal enough damage (determined through balancing and what Anet concludes is good) you also get credit for a kill. But if you die at any point you’re cleared of any kill assists that may have happened between your death and respawn, that way players are forced to stay alive and actively participating. That way someone can’t load up on AoEs and try to tag as many players recklessly as possible. They are challenged to stay alive even if they have to disengage a fight temporarily rather then being revive fodder to the other team.

On the second example;

Perhaps a system like WvW currently where if someone is still providing a support role like that, the team can vote after the match ends and certify that this person was doing well. We would need a set of system checks to prompt this so it can’t be abused. Like at the beginning before a match, if one person decides to patrol home and +1 when home is safe from a roamer, they can flag themselves and the team takes a party vote to agree or disagree with this person’s choice of role (It is team based so the team gets the final say). If the team agrees to this, after the match ends, before anyone can leave the instance they are prompted to vote whether this person did well or not. If you duo queue with someone, neither you or your friend may flag themselves for this role to prevent abuse. Or they may but whoever they duo queued with cannot vote for their buddy. Not a great solution but I don’t think a computer could ever really calculate or understand this situation with today’s technology. So we would have to rely on the human aspect to reward them fairly. But if they sit all match doing nothing but camping then they will not be allowed to receive any benifits of this role. If they actively show signs (determined through real world data like watching matches and setting test matches without telling anyone what the true purpose for testing so not to skew the results) then the computer will check a set of requirements like combat at least twice during a match, enemy’s average time within range of home during match, and etc (just examples) or whatever seems to accurately gauge there was a reason for this person being there. Then the vote is considered valid if a set of standards is met. If the vote is considered invalid then normal factoring will take place. Or forgo the previous vote system and factor into this players average range of home cap, opposing teams average range of your home during the match, and a few others. This way a player is not at a disadvantage for protecting a cap or home. So really it could be used for all the cap points.

Issues currently with MMR based tier

in PvP

Posted by: Xter.6271

Xter.6271

Now onto this thread in general, I think if this system were to be implemented it cannot allow anyone to gain MMR in a lose condition, but minimalize their MMR lost to a very small number. That way if you performed well on an individual level, but your team lost, you’ll lose very little MMR from your current score to help keep players from dropping too fast. Whereas if you performed poorly as an individual and a lost condition happened, the system will factor that and you’ll lose whatever MMR that will help bring you to a skill level you belong the fastest.

Where as in a win condition and you did exceptional as an individual you’ll gain more then normal MMR to help boost you to where you belong. And if you did average you’ll gain normal MMR.

Now first thing is this system would be quite tricky to balance. But I think what Anet could do to help balance this system in the testing stages is take a wide range of players from ESL to just fresh sPvPers and group them by skill levels without a computer or algorithm. Just simple hand picking them and have them play a lot of matches within their skill range. Then take all the metrics from every match for each skill level and average them for a match.

So let’s say on average for a low skill tier, every player scored at 4 kills per the average of all the matches. Every player made at least 1 capture for the average of all matches. They healed XXX HP to their allies for the average of matches. They dealt XXX damage for the average. They revived XXX allies across all matches (Lord and any other NPC would not be counted to prevent abuse). And etc.

Now we roughly have an average of what a player at this skill level should be doing. If they go over the average we would have a formula to calculate what they would receive but it would have diminishing returns so doing 2X the average damage will not provide 2X the benefits. Like wise for all other metrics.

Deaths will not be factored into this unless their deaths exceed the average by a large amount. For example like 3X the average. That way players being focused will not suffer a penalty for being focused fired. This would need to be balanced to determine the number but that way players being fodder to an outnumbered situation (2v4, 1v4, and etc.) will have a larger MMR drop to attempt to place them where matches are better suited to them.

EDIT; And to add to the damage averages and healing and etc, you’ve set up a threshold to qualify for having average metrics in these so builds that deal less damage or offer less healing and etc are not left out and placed poorly for an individual performance factor. Maybe this could swing for kills, caps, decaps, and revives if the averages can accommodate a range below the average. You don’t want this range to extend to low but enough to make it so people are not penalized for their builds

(edited by Xter.6271)

Issues currently with MMR based tier

in PvP

Posted by: hackks.3687

hackks.3687

Simple fix to prevent gaming the point capture contribution: it only counts for the first person to step onto the point – which more likely than not is the person intended to capture it rather than the person who ran back to perhaps cleave a corpse then stay for the cap.

Xter’s ideas are right on the money IMO

Hackkz/Riggamaroll
I’ve stayed at this party entirely too long

Issues currently with MMR based tier

in PvP

Posted by: Fay.2357

Fay.2357

Simple fix to prevent gaming the point capture contribution: it only counts for the first person to step onto the point – which more likely than not is the person intended to capture it rather than the person who ran back to perhaps cleave a corpse then stay for the cap.

Xter’s ideas are right on the money IMO

Well, good thing the devs don’t take poor ideas from random forum-goers.

Issues currently with MMR based tier

in PvP

Posted by: Saiyan.1704

Saiyan.1704

On top of which none of what you’re saying is a reason not to do these things. Just because there are problematic variables doesn’t mean they can’t be accommodated and addressed.

No, I honestly don’t think they can be adequately addressed — I’d gladly be proven wrong though.

It’s nothing but code, of course it can be addressed… Remember when Anet said “Oh well, we can’t implement outside PvP Queueing because of the [eye candy] we implemented. We don’t have the tech for it.”

The bold is what they said word for word in that regards. Yet here we are, able to queue without staying in HotM to do it.

As for making Matchmaking better, they can implement criteria that literally make MM think more when allocating players on a team… ya, it can be done. Matchmaking is so concentrated on the numbers on paper that it doesn’t even look at the individual class comp. It doesn’t notice that Foefire’s Middle is 3x larger than other maps. It doesn’t notice that there’s a Lord that gives 250 points when winning.

Other than player population… all these other items are very, easily, addressed. It just takes man hours.

aka FalseLights
Rank: Top 250 since Season 2
#5 best gerdien in wurld

(edited by Saiyan.1704)

Issues currently with MMR based tier

in PvP

Posted by: hackks.3687

hackks.3687

On top of which none of what you’re saying is a reason not to do these things. Just because there are problematic variables doesn’t mean they can’t be accommodated and addressed.

No, I honestly don’t think they can be adequately addressed — I’d gladly be proven wrong though.

It’s nothing but code, of course it can be addressed… Remember when Anet said “Oh well, we can’t implement outside PvP Queueing because of the [eye candy] we implemented. We don’t have the tech for it.”

The bold is what they said word for word in that regards. Yet here we are, able to queue without staying in HotM to do it.

As for making Matchmaking better, they can implement criteria that literally make MM think more when allocating players on a team… ya, it can be done. Matchmaking is so concentrated on the numbers on paper that it doesn’t even look at the individual class comp. It doesn’t notice that Foefire’s Middle is 3x larger than other maps. It doesn’t notice that there’s a Lord that gives 250 points when winning.

Other than player population… all these other items are very, easily, addressed. It just takes man hours.

Exactly

Hackkz/Riggamaroll
I’ve stayed at this party entirely too long

Issues currently with MMR based tier

in PvP

Posted by: Olterin Fire.5960

Olterin Fire.5960

To those of you saying all the statistical evaluation being easy to address … You do realize that we’re still talking about a computer evaluation, yes? If you are so sure these issues are easy to address, please, educate us (and Arenanet) by providing details of how to do it. Because all I have seen so far is “it should do this and this” but not the details of how it can be accomplished – nor the reasoning why those particular things would be the ones used.

What I’m talking about is – how do you weigh individual contribution based on match statistics? Which ones should be used? Should they be universal or class-dependent? If they are class-dependent, shouldn’t they instead be role-dependent? If they should, how do you automatically determine the role someone is playing without human interference?

Most importantly, can you also outline how such a proposed change would impact the system on a global scale? I mean, sure, I have seen examples of individual changes, but did anyone think of the larger implications? Until all of this has been adequately addressed, I have no reason to believe the current system is inherently bad.

Sure, I would like to not get dragged down during individual matches by having “bad luck” with the team, but as has been said time and time again… Bad luck averages out over a large sample size of games. So does good luck. The matchmaker has no preference for individual accounts to be put on “good teams”, thus, even though you only control roughly 10% of what goes into a match, assuming the other 90% average out to a 50/50 win/loss over time, it’s your 10% that will decide whether you rise or fall in rating. Of course, it might take more than 50 or even 200 games, but if you’re serious about PvP you’re likely to play more than that. In which case there is no issue here.

The one and only change I would like to see to the system as it stands right now, would be for decay to occur towards the average rating instead of towards 0. I mean, the soft MMR reset puts everyone’s rating closer to the average, not towards zero, so it only makes sense that by being inactive the skill is assumed to drift towards that same point.

WIthout light, there can be no darkness. Without darkness, there can be no light.

Sword Of Justice – Gunnar’s Hold

Issues currently with MMR based tier

in PvP

Posted by: megilandil.7506

megilandil.7506

To those of you saying all the statistical evaluation being easy to address … You do realize that we’re still talking about a computer evaluation, yes? If you are so sure these issues are easy to address, please, educate us (and Arenanet) by providing details of how to do it. Because all I have seen so far is “it should do this and this” but not the details of how it can be accomplished – nor the reasoning why those particular things would be the ones used.

What I’m talking about is – how do you weigh individual contribution based on match statistics? Which ones should be used? Should they be universal or class-dependent? If they are class-dependent, shouldn’t they instead be role-dependent? If they should, how do you automatically determine the role someone is playing without human interference?

Most importantly, can you also outline how such a proposed change would impact the system on a global scale? I mean, sure, I have seen examples of individual changes, but did anyone think of the larger implications? Until all of this has been adequately addressed, I have no reason to believe the current system is inherently bad.

Sure, I would like to not get dragged down during individual matches by having “bad luck” with the team, but as has been said time and time again… Bad luck averages out over a large sample size of games. So does good luck. The matchmaker has no preference for individual accounts to be put on “good teams”, thus, even though you only control roughly 10% of what goes into a match, assuming the other 90% average out to a 50/50 win/loss over time, it’s your 10% that will decide whether you rise or fall in rating. Of course, it might take more than 50 or even 200 games, but if you’re serious about PvP you’re likely to play more than that. In which case there is no issue here.

The one and only change I would like to see to the system as it stands right now, would be for decay to occur towards the average rating instead of towards 0. I mean, the soft MMR reset puts everyone’s rating closer to the average, not towards zero, so it only makes sense that by being inactive the skill is assumed to drift towards that same point.

decay to average… great, bronze decay to silver great idea

for the other things on post, yes, evaluate player contribution on teams is pure mess, class dependant , role dependantm, situation dependant so much variables to cool machine analisis and player voting pure joke , better firts adress mm algorithm to a proper one and the other will be only short lucky or bad luck strikes and pure great numbers law

Issues currently with MMR based tier

in PvP

Posted by: pelle ossa.9705

pelle ossa.9705

i think 1 big problem is not low population (3 years ago population in pvp was even less then now , but mm was not perfect , but good cos ppl who played care about pvp) but motivations.. it seesm like there ppl that play ranked cos they want to compete , they like spvp and stuff , there are some ppl that play ranked only cos they can, they even don’t care about winning or lose (with some absurd build even in gold) , they play only cos there are achievement/cheap way for ascended and legendary stuff.

Issues currently with MMR based tier

in PvP

Posted by: Olterin Fire.5960

Olterin Fire.5960

Perhaps I should expand on the decay bit. What I would expect decay to be, would not be a flat -100/+100 per day (depending on whether the player in question is below or above average), but rather something like exponential decay. This can potentially be made far less aggressive for rankings below the average but I firmly believe it should still be present after a period of inactivity. One could also – while keeping the basic decay formula the same – modify the decay time (in the exponential decay scenario) multiplicatively with a weight factor that would be dependent on the starting rating in question. So for example, around the average rating this factor would be nearly 1 thus resulting in a normalized decay time, it would go down the higher the rating is (thus making the decay kick in and be faster) and it would go up the lower the rating is (thus making the decay kick in and be slower).

One could make the baseline rate of decay be such that over the course of half the season, a given player (with a decay time factor of nearly 1) would drift back to the average rating. Obviously, this is almost never going to be the case, and people lower on the ranking list would have slower decay, while people higher up would have faster decay. From here on out, with a model of exponential decay, it would be merely a question of playing with some sample numbers until a satisfactory model is found.

WIthout light, there can be no darkness. Without darkness, there can be no light.

Sword Of Justice – Gunnar’s Hold

Issues currently with MMR based tier

in PvP

Posted by: SlippyCheeze.5483

SlippyCheeze.5483

…not even: it would only take ten to twenty games to put them roughly back at the same MMR they have now, because (shocking news!) the gliko algorithm converges fairly quickly on the target MMR.

I could see that being the case if a high MMR player was artificially lowered to bronze and inserted into an active system, but if everyone was reduced to bronze level MMR? I dunno, you might be right, I’m not pretending to be an expert.

I wouldn’t say I’m an expert, but I have worked with these matchmaking algorithms, and have some background to understand them when I have not.

Anyway, a common gliko implementation is to forgo “placement” matches or whatever, and insert new players with a rating of 350, high volatility. They then rapidly converge on real rating as they play.

So, yeah, it’d quickly sort people out into the “right” range as you played. The only people who wouldn’t get pushed to the right place are the non-players, which doesn’t matter in the slightest unless you also tie some sort of leaderboard to it.

(Which is why the default low MMR is good, because it means that you won’t have a top N where someone randomly fit in, after the first few matches.)