Allow MMR to decrease for every loss?

StaticWave

Hello, I was just thinking about how other team-based PvP games increase or decrease your MMR depending on certain factors such as win/loss or personal performance. I'm not sure how it could be implemented in ESO, but at least it could correctly place players in their appropriate ranking, where the win rate is usually 50/50.

jcm2606

Actual performance ranking metrics often pull data from many sources to get a more complete picture of a player's performance level. Not only would they pull from win/loss, but maybe also a score that indicates how much the player contributes to the match, on top of some general data such as damage done, received, healing done, received, etc.

There really isn't a reason why ESO couldn't do the same. Keep the current medal score tally to represent how much the player contributed to the match, but also add in win/loss, some general data about the player in the match, as well as maybe some data specific to each game mode, such as kills/deaths/assists for deathmatch, amount of time spent holding the ball or amount of healing done to allies holding the ball or amount of damage done to enemies holding the ball for chaosball, etc.

Weight each of these so they each contribute to the final MMR value a different amount, and maybe have the weights change depending on the game mode, if we want MMR to be per game mode (which it should be, IMO, since players typically perform very differently between game modes).

Roll out a preliminary version of this new MMR system that isn't actually hooked into the match maker, just to gather information about how the average player performs, and use that information to determine a baseline that players must meet to have a "neutral" MMR value for that match (ie if they're below the baseline, MMR value goes negative which means MMR drops; if they're within a few percent of the baseline, MMR value approaches zero which means MMR stays the same; if they're above the baseline, MMR value goes positive which means MMR rises).

Under a new MMR system like this, a player's MMR will start fairly low, and will gradually rise, causing the player to match against harder and harder opponents, until eventually their MMR starts to plateau and even drop, until they meet the baseline, at which point it can be assumed that the player is matching against other players with roughly equal performance levels to their own.

At the baseline, a player's MMR will be constantly fluctuating, but will always reach a balancing point, as it's constantly adjusting the matching preferences to match the player's performance level.

So long as all the weights are in decent ranges, this shouldn't cause any problems regarding queue times, as the match maker is still just pulling a single number that it can consider the player's performance rank. Just, the rank is being determined using a much more sophisticated function, that is overall more indicative of how a player actually performs.

ChunkyCat

7 paragraphs? really?

SirAndy

ChunkyCat wrote: »

7 paragraphs? really?

Better than a wall of text with no paragraphs at all ...

jcm2606

ChunkyCat wrote: »

7 paragraphs? really?

What can I say, I really want to see ZoS un-*** MMR.

ChunkyCat

You need at least 10 paragraphs for that.

jcm2606

ChunkyCat wrote: »

You need at least 10 paragraphs for that.

Damn, guess that's why they haven't fixed MMR in the year or so I've been asking them to do so.

ChunkyCat

start typing

NyassaV

Or just reset MMR with every new update. 3-4 time a year.

StaticWave

NyassaV wrote: »

Or just reset MMR with every new update. 3-4 time a year.

That's not the point. It's to categorize players based on their ranking. Currently people can just move up the mmr by playing alot. It's not how it works in other pvp games where you need to become better to climb the ranks. In those games you can play for 10k hours in bgs but if you lose more than you win or have the same amount of wins as your losses then you stay in the same rank.

BNOC

Because your own and enemy players MMR is not visible, it's pretty pointless to even do anything with MMR at all.

We'd have to be able to see that and there be some kind of rewards per season otherwise there's no incentive to climb and the lowering MMR will just be used as a way to face up against weaker players and stomp them.

jcm2606

BNOC wrote: »

Because your own and enemy players MMR is not visible, it's pretty pointless to even do anything with MMR at all.

We'd have to be able to see that and there be some kind of rewards per season otherwise there's no incentive to climb and the lowering MMR will just be used as a way to face up against weaker players and stomp them.

It's not pointless when MMR directly contributes to the *** experience of regular players. High MMR directly results in longer queue times, lower pops at the higher MMR brackets leading to no match diversity, premades matching against solo queues thanks to premades having their MMR inflated, etc.

Removing MMR and replacing it with an alternative that is properly designed would go a long way towards fixing these problems, on top of more correctly matching players of similar performance levels, due to MMR having a more complete picture of a player's performance level.

Zacuel

Well in bgs I get roflstomped. So I'd love a lower mmr.

StaticWave

Zacuel wrote: »

Well in bgs I get roflstomped. So I'd love a lower mmr.

Exactly. It’s not fun to get steamrolled by premades or players that are better. It’s usually not fun for those players to steamroll others either. Lowering the mmr allows players to play at a level that they’re comfortable at.

Rahar

jcm2606 wrote: »

Actual performance ranking metrics often pull data from many sources to get a more complete picture of a player's performance level. Not only would they pull from win/loss, but maybe also a score that indicates how much the player contributes to the match, on top of some general data such as damage done, received, healing done, received, etc.

There really isn't a reason why ESO couldn't do the same. Keep the current medal score tally to represent how much the player contributed to the match, but also add in win/loss, some general data about the player in the match, as well as maybe some data specific to each game mode, such as kills/deaths/assists for deathmatch, amount of time spent holding the ball or amount of healing done to allies holding the ball or amount of damage done to enemies holding the ball for chaosball, etc.

Weight each of these so they each contribute to the final MMR value a different amount, and maybe have the weights change depending on the game mode, if we want MMR to be per game mode (which it should be, IMO, since players typically perform very differently between game modes).

Roll out a preliminary version of this new MMR system that isn't actually hooked into the match maker, just to gather information about how the average player performs, and use that information to determine a baseline that players must meet to have a "neutral" MMR value for that match (ie if they're below the baseline, MMR value goes negative which means MMR drops; if they're within a few percent of the baseline, MMR value approaches zero which means MMR stays the same; if they're above the baseline, MMR value goes positive which means MMR rises).

Under a new MMR system like this, a player's MMR will start fairly low, and will gradually rise, causing the player to match against harder and harder opponents, until eventually their MMR starts to plateau and even drop, until they meet the baseline, at which point it can be assumed that the player is matching against other players with roughly equal performance levels to their own.

At the baseline, a player's MMR will be constantly fluctuating, but will always reach a balancing point, as it's constantly adjusting the matching preferences to match the player's performance level.

So long as all the weights are in decent ranges, this shouldn't cause any problems regarding queue times, as the match maker is still just pulling a single number that it can consider the player's performance rank. Just, the rank is being determined using a much more sophisticated function, that is overall more indicative of how a player actually performs.

Excellent post. I really hope that they do something like this when all is said and done.

The only thing I'd add is MMR decaying a little bit over time. Every week, regardless of how much the player participates or wins, a static, small percentage of MMR is taken right off the top so that only the most active and skilled players stay at the top, and queue times for more skilled players will be reduced. Other MMOs/games with good MMR systems or leaderboards do this to avoid stagnation, e.g., situations where players at the extreme ends of the MMR curve find no matches because there's no one else close to the same ranking they are.

MentalxHammer

Yes pls

jcm2606

Rahar wrote: »

jcm2606 wrote: »

Actual performance ranking metrics often pull data from many sources to get a more complete picture of a player's performance level. Not only would they pull from win/loss, but maybe also a score that indicates how much the player contributes to the match, on top of some general data such as damage done, received, healing done, received, etc.

There really isn't a reason why ESO couldn't do the same. Keep the current medal score tally to represent how much the player contributed to the match, but also add in win/loss, some general data about the player in the match, as well as maybe some data specific to each game mode, such as kills/deaths/assists for deathmatch, amount of time spent holding the ball or amount of healing done to allies holding the ball or amount of damage done to enemies holding the ball for chaosball, etc.

Weight each of these so they each contribute to the final MMR value a different amount, and maybe have the weights change depending on the game mode, if we want MMR to be per game mode (which it should be, IMO, since players typically perform very differently between game modes).

Roll out a preliminary version of this new MMR system that isn't actually hooked into the match maker, just to gather information about how the average player performs, and use that information to determine a baseline that players must meet to have a "neutral" MMR value for that match (ie if they're below the baseline, MMR value goes negative which means MMR drops; if they're within a few percent of the baseline, MMR value approaches zero which means MMR stays the same; if they're above the baseline, MMR value goes positive which means MMR rises).

Under a new MMR system like this, a player's MMR will start fairly low, and will gradually rise, causing the player to match against harder and harder opponents, until eventually their MMR starts to plateau and even drop, until they meet the baseline, at which point it can be assumed that the player is matching against other players with roughly equal performance levels to their own.

At the baseline, a player's MMR will be constantly fluctuating, but will always reach a balancing point, as it's constantly adjusting the matching preferences to match the player's performance level.

So long as all the weights are in decent ranges, this shouldn't cause any problems regarding queue times, as the match maker is still just pulling a single number that it can consider the player's performance rank. Just, the rank is being determined using a much more sophisticated function, that is overall more indicative of how a player actually performs.

Excellent post. I really hope that they do something like this when all is said and done.

The only thing I'd add is MMR decaying a little bit over time. Every week, regardless of how much the player participates or wins, a static, small percentage of MMR is taken right off the top so that only the most active and skilled players stay at the top, and queue times for more skilled players will be reduced. Other MMOs/games with good MMR systems or leaderboards do this to avoid stagnation, e.g., situations where players at the extreme ends of the MMR curve find no matches because there's no one else close to the same ranking they are.

Ideally, a situation like that would be addressed by basically looking at how many people in the queue are within a certain range of MMR values centered on the player, and adjusting the MMR leniency accordingly, possibly even taking into account the difference between the player's MMR and the nearest potential match (to get a rough idea of how far away it is, and adjust matching accordingly).

The problem with MMR decay is it kinda muddies the water a little, and it's doing something (making sure everyone sits where they are) that already should happen. An actual MMR system basically tries to constantly adjust itself based on how the player does from match to match. If you happen to get lucky a few matches in a row, absolutely slay, and your MMR goes way up, above what it should be, then it'll match you against people that are outside of your performance bracket, causing your MMR adjustments to go negative, pulling your MMR down, until you reach a point where you go neutral or positive again.

It's basically always trying to reach an equilibrium, depending on how you perform. Do good, it goes up. Do real good, it skyrockets. Do bad, it goes down. Do real bad, it plummets. Over time, assuming your performance level remains constant, it'll sort of approach a value that perfectly defines how you perform, and match against other players in a similar range.

MMR decay is good for quickly getting the player in lower performance brackets when they take a break and may potentially suck as a result of that break, avoiding the player going through a few horrible matches until their MMR finds that equilibrium again, but that's really about only what it should be used for, if MMR is implemented properly.