Let's talk about the MMR...

Thogard

When you do queue for a BG on your Stam sorc, are you getting in matches with BG regulars or are you getting put with people you don’t recognize who are just queueing for their daily? Or maybe it alternates between both?

Mojomonkeyman

Yeah, good post.

I personally am pretty sure that theres win/loss taken into account. It's the only difference between Cinbri's waiting times (40min+ solo) and mine (20min+ solo). We have played similar amount of matches on our chars, similar medal scores per game (since we both can heal/deal dmg/tank on our chars) and similar k/d/a per match. We usually play in the same premade, so I know our performance very well.

Yet his waiting times are double of mine, and the only difference we can relate to is winrate (mine at 65%, his at 93%). Really looking forward to learn what exactly the formula is based on.

Aliyavana

Thogard wrote: »

Aliyavana wrote: »

Thogard wrote: »

Does anyone on PC NA know Jon Stryfe? I really think he’s the key to this. We need to see what kind of matches he’s getting. He’s played a ton of matches but he seems to always queue solo. @Aliyavana you’re another really frequent solo queuer who isn’t running away from fights just to cap flags, so you’ve probably got a relatively low win rate but tons of matches played.. could you post a PvPMeter report if you have it? I remember you used to get a lot of matches against me / my group at summerset launch but i haven’t seen you in a while and I’m not sure if that’s because the system actually lowered your MMR or if you stopped queueing on your main.

I really want to take Jonstryfe or zebendel and have them do a sync solo queue up at the exact same time as someone with around 20 games played but a really really high win rate due to aggressive focus on objectives ... and I watch and see what happens. @damarky do you have any toons like that? If they both get the same queue pop we’ll have them decline, and we’ll keep going until only one of them gets a pop, and we’ll see which one gets thrown against my premade. It’ll be a good test and should let us know if playtime or winrate is more important... and if they keep getting in the same match we might have to start exploring Other vsriables like average KDR or alliance rank..

you havnt seen me because ive been busy with pve stuff in summerset tho I still hop in for the daily rewards

I was afraid you’d say that

Do you track your win rate? I always assume that because you don’t run away from fights that you’ve got a relatively low one like the rest of us that just like to PvP rather than PvF, but you seem to almost always be solo so your stats would be more pure than mine (easy to get a W even if not aggressively playing objective when you’re a premade fighting two pugs)

For objective games I'm usually always second or first place and I prioritize the objective over kills so I always score high in chaosball and other gamemodes. Deathmatch is highly reliant on if the other teams are premades as pugs will get completely crushed by premades and in objective is where pugs have a higher chance to win as premades are more likely to just be there to have a fun fight.

Thogard

Mojomonkeyman wrote: »

Yeah, good post.

I personally am pretty sure that theres win/loss taken into account. It's the only difference between Cinbri's waiting times (40min+ solo) and mine (20min+ solo). We have played similar amount of matches on our chars, similar medal scores per game (since we both can heal/deal dmg/tank on our chars) and similar k/d/a per match. We usually play in the same premade, so I know our performance very well.

Yet his waiting times are double of mine, and the only difference we can relate to is winrate (mine at 65%, his at 93%). Really looking forward to learn what exactly the formula is based on.

If everything is the same and you play together then why is his win rate so much higher than yours?

Mojomonkeyman

Thogard wrote: »

Mojomonkeyman wrote: »

Yeah, good post.

I personally am pretty sure that theres win/loss taken into account. It's the only difference between Cinbri's waiting times (40min+ solo) and mine (20min+ solo). We have played similar amount of matches on our chars, similar medal scores per game (since we both can heal/deal dmg/tank on our chars) and similar k/d/a per match. We usually play in the same premade, so I know our performance very well.

Yet his waiting times are double of mine, and the only difference we can relate to is winrate (mine at 65%, his at 93%). Really looking forward to learn what exactly the formula is based on.

If everything is the same and you play together then why is his win rate so much higher than yours?

One solo queues a lot, the other one plays only in group (with few exceptions). So Cinbri's W/L is basically our premades W/L.

Thogard

Mojomonkeyman wrote: »

Thogard wrote: »

Mojomonkeyman wrote: »

Yeah, good post.

I personally am pretty sure that theres win/loss taken into account. It's the only difference between Cinbri's waiting times (40min+ solo) and mine (20min+ solo). We have played similar amount of matches on our chars, similar medal scores per game (since we both can heal/deal dmg/tank on our chars) and similar k/d/a per match. We usually play in the same premade, so I know our performance very well.

Yet his waiting times are double of mine, and the only difference we can relate to is winrate (mine at 65%, his at 93%). Really looking forward to learn what exactly the formula is based on.

If everything is the same and you play together then why is his win rate so much higher than yours?

One solo queues a lot, the other one plays only in group (with few exceptions). So Cinbri's W/L is basically our premades W/L.

But wouldn’t that mean that you have a much higher Number of games played?

TBH I ruled out win rate as a factor pretty early on... a lot of BG regulars on PC NA don’t play objectives, but they’re still good and have high MMR. So then I thought “what about K/D?” But then healers and tanks Would have low MMRs.

Really the only thing the high MMR players have in common is that they’ve all played a ton of games. I recognize al of them and they recognize me... we all know each other.

A 40 min queue time seems a bit long for a solo queue. We don’t get a lot of those on PC NA. 25 minutes seems to be when the queue finder gives up and lowers your MMR enough to practically guarantee a match. I don’t know of any character on PC NA that is regularly getting 40 min queues during primetime. Any premade either... and the premade queue should be even longer.

If you look at Wheem’s PvPMeter he is showing a 35-40 win rate but he’s still got a pretty high MMR even when he solo queues. I think that stat is underestimating the # of BGs he’s played because it only started counting when he installed it.

I need to find someone with the same # of BGs as me but a higher win rate and lower k/d and do sync solo queues with them to test.

wheem_ESO

Thogard wrote: »

wheem_ESO wrote: »

I'm not convinced that MMR is based on cumulative lifetime score that never resets. All 3 of the characters I've played in BGs seem to have essentially the exact same queues; pretty fast during prime time, and matched up against premade teams decently often.

Magicka Nightblade:

Magicka Sorcerer:

Magicka Warden:

But here's the kicker - I did a lot of games on Magicka Warden before downloading the PvPMeter addon. I may not have the same number of games as Thogard, at least in part because I've been a bit more off-and-on playing over the last year or so, but my Warden actually has significantly more games than the other two characters.

I'll end up vs the same premade team(s) multiple times, pretty much back to back, on any of my characters. Then, after a few of those super-duper-ultra-fun games, I'll be put against most/all random solo queued players, including those with names that I don't recognize. It seems very unlikely that there aren't enough of those lower MMR players to end up exclusively against each other, leaving the system to grab me after 1 minute or so in the queue.

The only way I can see that the system could be basing MMR on cumulative lifetime scoreboard points (per character), is if the range for each bracket is really big. I don't remember when exactly I downloaded the PvPMeter addon, but I wouldn't be surprised if my Warden really has > 1,000 games, but with queue times and "types" that are indistinguishable from my Magblade.

PS
The overwhelming majority of my games on all 3 characters are solo queued, and I sometimes pay attention to objectives, and sometimes don't.

But it’s interesing. I had assumed that those matches were caused by my long wait time, but you seem to be implying that they’re being caused by losing / getting pug stomped in the High MMR matches. But those games happen abojt once a day to me, and it seems like a contradiction to think that your MMR could change that fast.

I can see how you'd think that from my wording, but what I actually meant was that it just seems like a mixed bag. Sometimes I get put against premades, and sometimes I get put against random solo or duo queued players. There are times that I recognize some or all of the names in the game, and other times I don't know any of them.

My main point is that if MMR was simply a cumulative scoreboard tally that never resets, my queues on Magicka Warden should be markedly different than my queues on Magicka Nightblade. Unless of course there's one really gigantic bracket that they both somehow ended up in, despite my Warden having many hundreds of more games played. It's possible that I did some Magblade games before downloading the PvPMeter, but I doubt it was very many at all. I know for sure that I ran a bunch of games on Warden prior to knowing about the addon, though. This leads me to believe that either win rate plays a role in calculating MMR, or that it resets (or at least reduces) after some amount of inactivity, since I've taken several off-and-on breaks from the game.

Edit:
Just for the record, if I'm queueing near prime time it's generally pretty close to instant on all 3 characters.

Mojomonkeyman

Thogard wrote: »

Mojomonkeyman wrote: »

Thogard wrote: »

Mojomonkeyman wrote: »

Yeah, good post.

I personally am pretty sure that theres win/loss taken into account. It's the only difference between Cinbri's waiting times (40min+ solo) and mine (20min+ solo). We have played similar amount of matches on our chars, similar medal scores per game (since we both can heal/deal dmg/tank on our chars) and similar k/d/a per match. We usually play in the same premade, so I know our performance very well.

Yet his waiting times are double of mine, and the only difference we can relate to is winrate (mine at 65%, his at 93%). Really looking forward to learn what exactly the formula is based on.

If everything is the same and you play together then why is his win rate so much higher than yours?

One solo queues a lot, the other one plays only in group (with few exceptions). So Cinbri's W/L is basically our premades W/L.

But wouldn’t that mean that you have a much higher Number of games played?

TBH I ruled out win rate as a factor pretty early on... a lot of BG regulars on PC NA don’t play objectives, but they’re still good and have high MMR. So then I thought “what about K/D?” But then healers and tanks Would have low MMRs.

Really the only thing the high MMR players have in common is that they’ve all played a ton of games. I recognize al of them and they recognize me... we all know each other.

A 40 min queue time seems a bit long for a solo queue. We don’t get a lot of those on PC NA. 25 minutes seems to be when the queue finder gives up and lowers your MMR enough to practically guarantee a match. I don’t know of any character on PC NA that is regularly getting 40 min queues during primetime. Any premade either... and the premade queue should be even longer.

If you look at Wheem’s PvPMeter he is showing a 35-40 win rate but he’s still got a pretty high MMR even when he solo queues. I think that stat is underestimating the # of BGs he’s played because it only started counting when he installed it.

I need to find someone with the same # of BGs as me but a higher win rate and lower k/d and do sync solo queues with them to test.

Cinbri plays exclusively his main magplar, while I am 50/50 magwarden/stamsorc. So, over all my BG chars I have probably double the amount of games played compared to him. Yet he has double the waiting time. What else other than W/L could have impact on our waiting times? Really, I am not sure how it works exactly. But W/L stands out as factor that I won't rule out yet.

Maulkin

The other thing that I find a bit weird is how your MMR is character based instead of account based. What other games does that? Which game out there ranks you based on your character/class/hero choice rather than your overall average?

Isn't this encouraging people to grind up new chars to play BGs? Isn't that really dumb? If I roll a new MagSorc I get a clean start and get put against new players even though I've been playing MagSorc since launch and have the experience of probably well over 2000 BG games across all my different chars. Wouldn't that ruin the games for the new players?

Cinbri wrote: »

ecru wrote: »

Cinbri wrote: »

Well, no matter if solo/duo or full premade, with 40min of re queue during primetime:

You sure this isn't just the buggy queue system? Regardless of what I'm doing, if I'm in the queue for longer than 10 minutes I just requeue. ESO's queues aren't reliable enough to wait out a long queue like that IMO, there's no way to know if the queue will pop or if you'll be queued for the next five hours.

I requed for 40min before deciding to stop doing it and see how fast it find match. Coz re que not helping me at all and my timer after update is 40+min between matches. Played bg minly since morrowind but now literally cant do it and have to go cyro.
Idk if this is result of:

Might as well grind up a new MagPlar mate, this one's forked.

Anethum

Maulkin wrote: »

The other thing that I find a bit weird is how your MMR is character based instead of account based. What other games does that? Which game out there ranks you based on your character/class/hero choice rather than your overall average?

Isn't this encouraging people to grind up new chars to play BGs? Isn't that really dumb? If I roll a new MagSorc I get a clean start and get put against new players even though I've been playing MagSorc since launch and have the experience of probably well over 2000 BG games across all my different chars. Wouldn't that ruin the games for the new players?

Cinbri wrote: »

ecru wrote: »

Cinbri wrote: »

Well, no matter if solo/duo or full premade, with 40min of re queue during primetime:

You sure this isn't just the buggy queue system? Regardless of what I'm doing, if I'm in the queue for longer than 10 minutes I just requeue. ESO's queues aren't reliable enough to wait out a long queue like that IMO, there's no way to know if the queue will pop or if you'll be queued for the next five hours.

I requed for 40min before deciding to stop doing it and see how fast it find match. Coz re que not helping me at all and my timer after update is 40+min between matches. Played bg minly since morrowind but now literally cant do it and have to go cyro.
Idk if this is result of:

Might as well grind up a new MagPlar mate, this one's forked.

or stop play premade vs randoms. he's nice but i didn't saw him solo any time

Maulkin

Anethum wrote: »

Maulkin wrote: »

The other thing that I find a bit weird is how your MMR is character based instead of account based. What other games does that? Which game out there ranks you based on your character/class/hero choice rather than your overall average?

Isn't this encouraging people to grind up new chars to play BGs? Isn't that really dumb? If I roll a new MagSorc I get a clean start and get put against new players even though I've been playing MagSorc since launch and have the experience of probably well over 2000 BG games across all my different chars. Wouldn't that ruin the games for the new players?

Cinbri wrote: »

ecru wrote: »

Cinbri wrote: »

Well, no matter if solo/duo or full premade, with 40min of re queue during primetime:

You sure this isn't just the buggy queue system? Regardless of what I'm doing, if I'm in the queue for longer than 10 minutes I just requeue. ESO's queues aren't reliable enough to wait out a long queue like that IMO, there's no way to know if the queue will pop or if you'll be queued for the next five hours.

I requed for 40min before deciding to stop doing it and see how fast it find match. Coz re que not helping me at all and my timer after update is 40+min between matches. Played bg minly since morrowind but now literally cant do it and have to go cyro.
Idk if this is result of:

Might as well grind up a new MagPlar mate, this one's forked.

or stop play premade vs randoms. he's nice but i didn't saw him solo any time

Is there a way for him to chose to play against premades only? Can he choose his opponents?

CatchMeTrolling

I pretty much get grouped against the same people, no matter what character I'm on. Have a stamplar, stam sorc, warden and a magplar.

PS. I queue solo and have 700 plus wins, would be more if I actually tried to win still.

Thogard

CatchMeTrolling wrote: »

I pretty much get grouped against the same people, no matter what character I'm on. Have a stamplar, stam sorc, warden and a magplar.

PS. I queue solo and have 700 plus wins, would be more if I actually tried to win still.

Aren’t you on console though?

CatchMeTrolling

Thogard wrote: »

CatchMeTrolling wrote: »

I pretty much get grouped against the same people, no matter what character I'm on. Have a stamplar, stam sorc, warden and a magplar.

PS. I queue solo and have 700 plus wins, would be more if I actually tried to win still.

Aren’t you on console though?

Yeah, mostly

Thogard

CatchMeTrolling wrote: »

Thogard wrote: »

CatchMeTrolling wrote: »

I pretty much get grouped against the same people, no matter what character I'm on. Have a stamplar, stam sorc, warden and a magplar.

PS. I queue solo and have 700 plus wins, would be more if I actually tried to win still.

Aren’t you on console though?

Yeah, mostly

Ok good to know the context. So right now you aren’t in the new system that we’re discussing, but once you are it’ll be interesting to follow your data. Can you try to track your queue times and the “level” of your opponents?

We’re trhing to isolate be variables and pin down what MMR is based on.

Thogard

@Mojomonkeyman would it be safe to assume that in addition to a higher win rate, he also has a higher “average medal score” ?

I’m wondering if average medal score interacting with total games played is a factor. I assumed it was the easiest route of just total cumulative medal score over time, but it is probably more sophisticated. Medal score would have a strong correlation with win rate.

CatchMeTrolling

Thogard wrote: »

CatchMeTrolling wrote: »

Thogard wrote: »

CatchMeTrolling wrote: »

I pretty much get grouped against the same people, no matter what character I'm on. Have a stamplar, stam sorc, warden and a magplar.

PS. I queue solo and have 700 plus wins, would be more if I actually tried to win still.

Aren’t you on console though?

Yeah, mostly

Ok good to know the context. So right now you aren’t in the new system that we’re discussing, but once you are it’ll be interesting to follow your data. Can you try to track your queue times and the “level” of your opponents?

We’re trhing to isolate be variables and pin down what MMR is based on.

Queue times sure and level as in skill?

Thogard

CatchMeTrolling wrote: »

Thogard wrote: »

CatchMeTrolling wrote: »

Thogard wrote: »

CatchMeTrolling wrote: »

I pretty much get grouped against the same people, no matter what character I'm on. Have a stamplar, stam sorc, warden and a magplar.

PS. I queue solo and have 700 plus wins, would be more if I actually tried to win still.

Aren’t you on console though?

Yeah, mostly

Ok good to know the context. So right now you aren’t in the new system that we’re discussing, but once you are it’ll be interesting to follow your data. Can you try to track your queue times and the “level” of your opponents?

We’re trhing to isolate be variables and pin down what MMR is based on.

Queue times sure and level as in skill?

Look for commonalities. Right now it seems to be experience-based... the names I’m seeing are the names I’ve seen the most prior to the patch, not necessarily the best of the best players. But we don’t know for sure. Just try to see if you notice any similarities so we can try to find how MMR is determined.

Mojomonkeyman

Thogard wrote: »

@Mojomonkeyman would it be safe to assume that in addition to a higher win rate, he also has a higher “average medal score” ?

I’m wondering if average medal score interacting with total games played is a factor. I assumed it was the easiest route of just total cumulative medal score over time, but it is probably more sophisticated. Medal score would have a strong correlation with win rate.

Nope, its similarly high (medalscore) since both of our chars can reliably deal dmg, tank and heal.

Maulkin

Mojomonkeyman wrote: »

Thogard wrote: »

@Mojomonkeyman would it be safe to assume that in addition to a higher win rate, he also has a higher “average medal score” ?

I’m wondering if average medal score interacting with total games played is a factor. I assumed it was the easiest route of just total cumulative medal score over time, but it is probably more sophisticated. Medal score would have a strong correlation with win rate.

Nope, its similarly high (medalscore) since both of our chars can reliably deal dmg, tank and heal.

A bit out of context but what sets are you running on your magden atm?

Mojomonkeyman

Maulkin wrote: »

Mojomonkeyman wrote: »

Thogard wrote: »

@Mojomonkeyman would it be safe to assume that in addition to a higher win rate, he also has a higher “average medal score” ?

I’m wondering if average medal score interacting with total games played is a factor. I assumed it was the easiest route of just total cumulative medal score over time, but it is probably more sophisticated. Medal score would have a strong correlation with win rate.

Nope, its similarly high (medalscore) since both of our chars can reliably deal dmg, tank and heal.

A bit out of context but what sets are you running on your magden atm?

Full cheese. Skoria, Sloads, Overwhelming. I will go back to my old setup tho, the new one seems to perform worse than that (Skoria, Torugs / Overwhelming). Frost + s&b as always.

ak_pvp

I wish they did it over an average of your medal score of a certain game type compared to others. That would be simple and balanced, then to tone down class bias, i.e. sorcs in deathmatch, they could change medal points vs classes themselves.

Have it reset every month so you aren't stuck and separate 3+ premades from non premades. And then weigh duos a little higher than normal, but keep with soloq.

This current bad method is basically screwing over common players putting them with premades and not actually with other similar players.

Galalin

Just wanted to say thanks for the work on this @Thogard. I have no even bothered to log in the last couple days... I used to do some flower picking, questing and PvE stuff wating for queues. I hardly bother anymore. Its just not worth the frstration of the queue.

Thogard

Mojomonkeyman wrote: »

Thogard wrote: »

@Mojomonkeyman would it be safe to assume that in addition to a higher win rate, he also has a higher “average medal score” ?

I’m wondering if average medal score interacting with total games played is a factor. I assumed it was the easiest route of just total cumulative medal score over time, but it is probably more sophisticated. Medal score would have a strong correlation with win rate.

Nope, its similarly high (medalscore) since both of our chars can reliably deal dmg, tank and heal.

Good to know. Win rate makes sense from a logical standpoint. I’ll see if I can play around with it a bit to see if I can isolate that variable. It would definitely be win rate AND # of games though, because we know that people with only a few games but a high win rate still aren’t getting the ultra-MMR premades unless it’s 4AM.

Another thing to consider is that each match mode might have a different MMR. That would really make testing difficult.

Honestly I think we might just have to wait until ZOS says something.

hesobad

Thogard wrote: »

Is there a silver lining?
Yes. High MMR premades tend to fight against High MMR premades when they are in queues together. This is a good thing if there are three High MMR premades all queueing around the same time. And if there are going to be solo players fighting the high MMR premades, it makes sense to take the "best" (let's not discuss the correlation or lack thereof between being High MMR and being good) solo players and put them in the match. It just really sucks to be a solo player who is always getting put in those matches...

If the population is an issue, they should allow for smaller populated games. Such as a 2v2, 3v3, or 4v4 for a pre made game. Wow Arenas uses this system and its extremely enjoyable! BG ques for grouped players would be very quick if the system only needed to find 8 players for a 4v4 rather than 12 for a 4v4v4

Mihael

hesobad wrote: »

Thogard wrote: »

Is there a silver lining?
Yes. High MMR premades tend to fight against High MMR premades when they are in queues together. This is a good thing if there are three High MMR premades all queueing around the same time. And if there are going to be solo players fighting the high MMR premades, it makes sense to take the "best" (let's not discuss the correlation or lack thereof between being High MMR and being good) solo players and put them in the match. It just really sucks to be a solo player who is always getting put in those matches...

If the population is an issue, they should allow for smaller populated games. Such as a 2v2, 3v3, or 4v4 for a pre made game. Wow Arenas uses this system and its extremely enjoyable! BG ques for grouped players would be very quick if the system only needed to find 8 players for a 4v4 rather than 12 for a 4v4v4

I think death match, relic , and domination should be made into 4v4 to encourage more “competitive play and to help ques, this would also help make these 2 objective game modes more popular since it would be harder to win by avoiding pvp

Chaos ball and crazy king should stay 4v4v4 so that they don’t have to redo everything and the players that enjoy the three team aspect of bgs can still have fun

ShadowMonarch

I think Kill Death Ratio probably is a big factor too. I have earned about 1.8mil ap roughly off BG's on this character. I don't know what the average AP is but I just used 10k. Which would add up to 180 games, give or take. If the 10k is off you can math it with whatever you think the average is. But i have never had a super great win rate probably 30-40%(more like the south side of 10% after summerset, ik embarassing) but my KD is around 8.5.

So I think that while total games played, medal score, and win rate might matter, im sure KD is probably involved as well. My win rate has shot down wildly sense summerset but my KD has stayed largely unchanged and after about 40 games now (after summerset) Im still in the same queue. Tonight alone I had a 28min queue, then a 19min queue before I gave up 25mins into the third queue and logged out.

TLDR I think either:
A. KD also effects MMR
or
B. MMR cannot go down.

Anomanderake

All my games starts to be against premade.
Good job again ZOS. The point was to allow the premade's pugstomping right?

Thogard

ShadowMonarch wrote: »

I think Kill Death Ratio probably is a big factor too. I have earned about 1.8mil ap roughly off BG's on this character. I don't know what the average AP is but I just used 10k. Which would add up to 180 games, give or take. If the 10k is off you can math it with whatever you think the average is. But i have never had a super great win rate probably 30-40%(more like the south side of 10% after summerset, ik embarassing) but my KD is around 8.5.

So I think that while total games played, medal score, and win rate might matter, im sure KD is probably involved as well. My win rate has shot down wildly sense summerset but my KD has stayed largely unchanged and after about 40 games now (after summerset) Im still in the same queue. Tonight alone I had a 28min queue, then a 19min queue before I gave up 25mins into the third queue and logged out.

TLDR I think either:
A. KD also effects MMR
or
B. MMR cannot go down.

The issue i have with KDR affecting MMR is that healers and tanks that have low KDRs can still have really high MMRs.

But an idea:

Sum of Damage done, healing done, and damage received (all of which are tracked).

Morgul667

Quite interesting thanks !

That would explain a lot
Hope it is improved