FAQ on the new rating system for Arena

[Taknar]

I'm not a huge PvP guy, but I am a huge numbers and heuristic guy, so this caught my eye.

http://blue.mmo-champion.com/11/14910422788-arena-matchmaking-system-faq.html

I'd quote the whole thing but honestly there is too much information there.

The system is basically adding a new rating where it adapts your team rating to its own opinion of where your team should be, based on the number of wins and losses (aiming for about 50/50 winds to losses). Kalgan has a really good writeup there where someone was using an example and he actually went in, grabbed his arena team's data, put it up and walked through the heuristic to show how the hidden ranking works. It's fantastic stuff.

Read it and discuss here, mostly cause I want to see what everyone else thinks from an actual PvP perspective.

[Evanthe]

Can you copy/paste?

[Swerto]

Why am I 1300 when my record is 50/50? 70/30?
Let’s turn the question around first. Why is my record 50/50? If your record is 50/50, that means your match-making rating (MMR) is almost exactly correct for you. It means we are finding teams that as near as possible match the ability of your team. It means that you are being matched as fairly as possible. If we matched 1300 teams at 1500 instead, they would only win 30% of their matches. We would rather those teams win half their matches.

But I’m 70/30, what about that?
If you are 70/30 and we think you are 1300, your MMR will climb to roughly 1533. You will then be matched with 1500+ teams, and your record should decrease to 50/50.

Yeah, so why is my rating still going down then?
The team rating will drop until it reaches the average match-making rating of the teams you beat 50% of the time. For example, if we are matching you against 1200-rated teams, and you are winning half the time, it will drop to 1200. If you are winning 70% of the time against 1200-rated teams, then your MMR and TR should move up to 1433.

But I’m not matching 1200-rated teams, I looked these teams up and most are near 1500.
It is likely that those teams have an MMR of 1200, even though their TR is 1500. This means that even though they look like 1500 or 1400-rated teams, they have been playing more like 1200-rated teams this season.

No matter what I do, my rating dives to 1300
If your team rating keeps diving to 1300 while you win 50% of your matches, then you are winning matches against 1300 players.
Being 1300 means:
-- You go 50/50 vs 1300 teams
-- You are better than 25% of all teams
It means that the teams we are matching you against have average MMRs well under 1500.

How will I get to 1650?
-- The definition of a 1650 team in both the new and old system is “Better than 68% of all teams on this BG in this bracket”
-- To be 1650, your team needs to be better than 68% of all teams
-- This means you will beat a team with an MMR of 1500 63 out of 100 matches, and go 50/50 vs. a team with an MMR of 1650

Wait, why 63%? Shouldn’t it be 68% then?
-- Smart question! The reason it’s not 68% is because a 1650-rated team does not always perform exactly at 1650. We take this into account.

But I’m stuck at 1300, I never get to play those higher-rated teams:
-- A 1650 player will win about 4 of 5 matches against a 1300 player
-- If you are at 1300 but belong at 1650, you should be winning 4 of every 5 matches
-- If you do that, your MMR will go up steadily, and you will face harder and harder opponents
-- Once you move your MMR to 1650, your TR will follow

How long will this take?
-- This depends on how much evidence we have that you *were* a 1300 team
-- If your team *Really* is a 1650 team, average case is 40 games
-- But it can go higher and lower. Great luck = 16, Terrible luck = 150

I guess that’s correct theoretically, but it’s very discouraging:
We do believe that we can improve the experience for teams that start at 1500 and go down from there. However, the changes we prefer merit waiting for the next arena season.

Ker bla


Comments on the results...

- Your team started with the system believing that your skill as a team was around 1350, but with a low degree of confidence in that number because the system was new. Your team lost versus a team we believed to be of around 1378 skill, so the system starts trying to aggressively lower the quality of your opponents to see if you do better (the lack of confidence the system had in your rating is why it is changing your MMR quickly at ths point).

- Your team continued to lose versus teams we believed to be around 1240 rating, 1093 rating, 1313 rating, 1154 rating, and even 928 rating. That's six losses in a row to teams far below your team's currently displayed rating, your currently displayed rating of 1311 is pretty high considering the teams you're losing to and the game believes you are probably not 1311 skill.

- You win your first game, versus a team the game believes to be about 842 rating (as seen in their MMR above). Your team only gains 1 team rating, because after all your team's visible rating is already 1311, and this win was merely versus a 842 skill team. Your matchmaking rating goes up by 40 points though, as the game thinking there's definitely a chance you're better than 900 rating.

- You win 3 more in a row, versus teams the game believes to be of about 923, 689, and 1109 rating. Your matchmaking rating is going up meaningfully, but since these teams are rated so far below your team's rating, you aren't gaining much for it.

- You lose 3 in a row after this against teams of 1166, 957, and 1133 quality. These teams are significantly below your team rating, and the system is gaining confidence that your current skill is not 1300+, so it's moving you toward around a 1005 rating (this is why you're losing a lot of team rating for your losses)... the system is pretty confident based on who you're beating versus who you're losing to that you're overrated at your 1200+ rating.

- Your team wins a few versus teams of 932 and 1004 rating, teams we expect you to beat close to 50% of the time now, so the game is matching you against opponents that are probably close to your skill, but unfortunately your team rating is way higher than that already, so you aren't gaining much TR for the wins (no surprise).

- Your team loses 4 more games versus teams of around 1000 rating, indicating that maybe 1000 rating is even possibly too high, your team rating keeps moving below 1200 on it's way to around 900 pretty quickly (losing between 17 to 24 team rating per game).

- You win, lose, win versus teams of about 1050 average matchmaking rating. You're going about 50/50 versus teams that the game thinks is about your rating (1000 or so), so your team rating continues to move toward that.


As explained in the FAQ's, your team rating is moving toward a rating where you beat opponents about 50% of the time. To remain at the 1434 rating your team started at, you would have had to have been winning 50% of your games versus teams that were significantly more successful.

As some of you might see, the matchmaking rating actually adapts very quickly. By no means are you pigeonholed at a rating, it adapts more quickly in many respects than a traditional Elo system. However, you do have to be able to beat teams of a given rating about 50% of the time in order to get your team rating to a given level.

BTW, hopefully in the process of copy/pasting data and trying to get the formatting right I didn't mess up any of the data.

wonderful post, but how is this fun for anyone?

except for numbers people, this system is obscure by design and rather byzantine.

as a non arena player I just feel that the resources spent on this mini game are out of proportion to any entertainment it provides.

that of course is not my decision to make.

my decision is whether to play or not in arena, and that was decided long ago.

baring massive changes to the fun factor, I don't see any reason to reevaluate.

Part of the point of the responses I've been making are to try to illustrate a belief I have that many people are associating their lack of fun with the new math behind the rating system. I totally believe that Shadoweric (and players in similar situations) are not having fun, but not because of the math behind the new system that finds equivalent opponents for their team, but for two other reasons.

1. Teams that are below 1500 in terms of their effectiveness in arenas feel crappy because the rating they start a team with is as high a rating they'll see in a long time (until they improve, or possibly ever). Bunk.

2. Gear that most players consider to be worth their time (including the players below 1500 rating) have requirements that are too harsh relative to how hard it is to get "good" gear through other parts of the game (including by buying hateful gladiator gear for emblems).

Those two problems are solve-able, and we plan to solve them, but the problems/solutions really have nothing to do with the new fancy math itself.



There's still plenty of room for some items to be pretty hard to get without causing most players to be left out of good gear. As I mentioned earlier, with the start of the next season we expect new teams and personal ratings to start much lower than they have before (not 1500). That way, everyone's experience is to climb toward the rating they're successfully playing at. So, knowing that, here's what our current thinking is on rating requirements for season 6.

Disclaimer: all of this is subject to change!
Other notes/caveats:
- items with no rating listed would have no rating requirement

- it won't be quite as easy to get deadly items via emblems of conquest as hateful items currently are via emblems of valor

- the new Archavon boss (a new boss located in another wing of the Vault of Archavon), will have slightly different drops than the current Archavon (he can drop Furious Gloves/Legs, or any possible random honor item in addition to tier 8 pve set items similar to Archavon's t7 pve set drops, but cannot drop the Furious Chest piece)

- honor/arena costs aren't listed but Hateful costs would be similar to current Savage, Deadly costs similar to current Hateful, and Furious costs would be similar to current Deadly

- you might note that the Furious weapon is still at a high requirement, 2k, the thinking being that most players are encouraged to get their weapon upgrade through pve, although hardcore arena competitors have an outlet to get an equivalent weapon through arenas (if you're hardcore enough for 2k, you can forego raiding for a weapon)

- additional note, "deadly trinket" refers to the current battlemaster's trinkets, "furious trinket" refers to an upgraded medallion of the alliance/horde


Hateful Gloves
Hateful Legs
Hateful Chest
Hateful Helm
Hateful Shoulders
Hateful Ring
Hateful Trinket

Deadly Bracer
Deadly Belt
Deadly Boots
Deadly Neck
Deadly Ring
Deadly Cape
Deadly Trinket

1300+: Deadly Relics/Idols/Librams/Totems
1350+: Deadly Gloves
1400+: Deadly Legs
1450+: Deadly Chest
1500+: Deadly Helm
1550+: Deadly Shoulders

1400+: Furious Bracer (honor)
1450+: Furious Belt (honor)
1500+: Furious Boots (honor)
1550+: Furious Neck (honor)

1600+: Furious Gloves
1650+: Furious Ring (honor)
1700+: Furious Legs
1750+: Furious Chest
1800+: Furious Trinket (honor)
1850+: Furious Helm
1900+: Furious Cape (honor)
1950+: Furious Wands/Relics/etc
2000+: Furious Weapon
2100+: Furious Shoulders
2300+: Furious Tabard (purely cosmetic and awesome looking)

[Taknar]

Thanks Swerto. you have more time to do this than I do at work it seems.

[Swerto]

That's because I'm at school and my "class" doesn't start til 1:00 EST.

Stupid weight training.

[Swerto]

I'm done with all my classes in high school. I'm in college and a career center for my real classes.

Yay Visual Basics I and Computer Networking :D

I'm just taking Weight Training for a free credit and way to get into shape.

P.S. OFF TOPIC IS OFF TOPIC... let's move to visitor messages and delete these.

[Advurb]

http://blue.mmo-champion.com/11/14910422788-arena-matchmaking-system-faq.html


Summary: Yo dawg we herd you like ratings so we put a rating on yo rating so you can tank it while you tank it.

Blizzard needs to stop throwing ratings around IMO and just fix crap so it works. What was so bad about just having team rating? Is it really that unacceptable to allow people to play for multiple teams?

[Taknar]

Summary: Yo dawg we herd you like ratings so we put a rating on yo rating so you can tank it while you tank it.

Blizzard needs to stop throwing ratings around IMO and just fix crap so it works. What was so bad about just having team rating? Is it really that unacceptable to allow people to play for multiple teams?

The change is far deeper a change than you are implying here, although I'm always a fan of a good exibit meme.

They are moving away from the elo system to something else, adding a verification that the rating you have is the rating you deserve. The most it does to a player bouncing between teams is that the system will respond faster to the general skill increase of the team. It doesn't stop it at all. Just realize that when that guy leaves you'll bounce back down just as fast, or faster.

[Advurb]

The change is far deeper a change than you are implying here, although I'm always a fan of a good exibit meme.

They are moving away from the elo system to something else, adding a verification that the rating you have is the rating you deserve. The most it does to a player bouncing between teams is that the system will respond faster to the general skill increase of the team. It doesn't stop it at all. Just realize that when that guy leaves you'll bounce back down just as fast, or faster.

Actually with the whole multiple teams thing I was referring to PR.

Either way, though, it's like that guy said. This system seems pretty byzantine, almost dubious to me. Of course, the lead dev did say that Arena was going to be played more like poker than the old chess-based strategy. /facepalm

[Taknar]

Actually with the whole multiple teams thing I was referring to PR.

Either way, though, it's like that guy said. This system seems pretty byzantine, almost dubious to me. Of course, the lead dev did say that Arena was going to be played more like poker than the old chess-based strategy. /facepalm

With the number of times he has re-clarified that statement, people still bringing it up makes me think that they are clutching on to it in order to justify their anger at the system because it doesn't benefit them on a personal level.

Perhaps you could explain the issues to us, instead of just re-quoting wow forum hacks.

[Advurb]

I could explain it, but it would be much easier for me to just link the thread (http://www.arenajunkies.com/showthread.php?t=57992&page=2) that echoes my thoughts. To be honest, it's hardly fair to call the majority of AJ posters "hacks."

[Taknar]

Wow forums are not Arena Junkies forums. I'll read the thread when I get home. Thanks for sharing the info.

[Fhenrir]

The whole system is getting convoluted and confusing to the point where I just stopped paying attention and don't really care anymore. This reminds me of the first week of this new "third" hidden arena rating though;

http://www.imagedumpwitheasy.com/uploads/7c0eb1c4b6.png

Hilarity.

Anyway, it's my personal hope that they just can it and go back to rating and personal rating. While I'm not a fan of the limitations personal rating put on the ability to play in multiple teams a week regularly, it does allow the system to have gear require rating that people can't get by buying teams off more competent players.

[Taknar]

I disagree. This new system once it gets integrated at the beginning of a season instead of the end, and once you start at the bottom rating instead of the elo system's middle rating, will be much better. I want to be able to go into arenas, regardless of what skill I am, and be able to come out against opponents knowing I had a fighting chance. In the current system, I can't do that and that's why I don't play arenas.

Major shame on Blizzard though for springing a new system mid-season. That was just dumb.

[Marroc]

I disagree. This new system once it gets integrated at the beginning of a season instead of the end, and once you start at the bottom rating instead of the elo system's middle rating, will be much better. I want to be able to go into arenas, regardless of what skill I am, and be able to come out against opponents knowing I had a fighting chance. In the current system, I can't do that and that's why I don't play arenas.

Major shame on Blizzard though for springing a new system mid-season. That was just dumb.

QFT. Get this, since the system was put in mid-season, there are 2030+ rated teams out there that haven't played in a few weeks, and thus have 1500 hidden ratings.

Just played my games for this week and we started at 1517 (after almost 20 games the previous week to even out our rating and team so we weren't losing 30 points a game), but this week we went 9 wins to 12 losses and dropped to 1450. Now that's not that out of the ordinary on it's own, except for the fact that maybe 8 of those 12 losses were to 1800+ rated teams that just hadn't played much since the new system (and thus had hidden ratings no where near their actual skill).

I honestly have no clue what blizzard was thinking on this one, if this were a real sport there would have been player riots. You don't change a rule to a game mid way through the game season.

[Advurb]

All I know is that as a 1650 team, I'm beating other 1650 teams and getting 9 points, while they're losing 21.

That.... is just stupid.

[Yatokth]

All I know is that as a 1650 team, I'm beating other 1650 teams and getting 9 points, while they're losing 21.

That.... is just stupid.

While I was 2v2ing on my team (we started at about 1530) we shot up to 1620 with 23 point wins.

The second time we played it evened out, winning/losing about 11 or 15 points each time.

I have no idea what's up, I'm just glad my personal rating is going towards my team instead of my hidden rating so I don't miss out on gear.

EDIT: This new system doesn't look completely stupid (though why they decided the old zero-sum system wasn't good enough escapes me entirely) but it is definitely mega-borked since it was introduced IN THE MIDDLE OF A SEASON. not a good call.

[Kained]

I'm currently 2106 I'm losing -20 Pts to rouge priest and DK shaman and winning 5 pts on them.

I'm losing 7 Pts to rouge warrior.... and Winning 5 pts

Uhm wtf?

[Advurb]

I'm currently 2106 I'm losing -20 Pts to rouge priest and DK shaman and winning 5 pts on them.

I'm losing 7 Pts to rouge warrior.... and Winning 5 pts

Uhm wtf?

Yeah exactly.

[Taknar]

I'm currently 2106 I'm losing -20 Pts to rouge priest and DK shaman and winning 5 pts on them.

I'm losing 7 Pts to rouge warrior.... and Winning 5 pts

Uhm wtf?

In the current system, the amount you win or lose is dependant on the difference between your hidden rating and the enemy's hidden rating. The system is trying to adjust your shown rating to match more closely to the hidden one. So to use the example of the RP you lost 20 points when losing because your shown rating is far above the enemy's hidden rating. The system is lowering you down to where it thinks you should be. On the other hand, when you win against that team your shown rating needs less adjustment (you are already approx. where you should be vs. the other team's hidden rating)


Don't ask me to explain that image where the guy lost points by winning though. That's just messed up.

[Advurb]

Don't ask me to explain that image where the guy lost points by winning though. That's just messed up.

There was a week or so after they implemented this where something was messed up and that happened to everyone. Or, at least, it would do dumb things like only give 1 pt for wins every time.

[Taknar]

I still stand by the statement that implementing this system now was stupid.

[Ryoku]

Damnit, Blizzard.

Just can't make things easy, can you?

[Yatokth]

http://forums.worldofwarcraft.com/thread.html?topicId=14990442082&sid=1

Seeing this thread made me think:

"Blizz can't possibly be this retarded. A nearly 50% win ratio on a 2K+ team?"

But now I see the point of this new system. It's trying to match you against people where you WILL have a 50% win ratio, everyone having a near 50% win ratio means the system is -working-. It won't ever be truly 50% cause then nobody would rise or fall in ratings, but it can get close.

That team likely was two 2K+ players who then made a team and cruised back up, gaining an obnoxious amounts of rating to boost them to their hidden, and their opponents not losing much cause the hidden rating was so massively above theirs.

The new system tries to match you to people where you will have a 50% win ratio (AKA they're just as good as you) and if you do -better- than that, it adjusts you upward so you can rise in ranking. AKA, if you start mopping up scrubs, you rise in ranking, once you start hitting a place where you're fighting evenly, your point gain evens out, you can still -progress- by winning alot and jumping the 50% ratio, but you'll stay closer to where your hidden rating is.

So high rated teams SHOULD have around a 50% win ratio, cause they're fighting teams just as good as them.

The advantages and disadvantages are thus:

Pros:

- This system is less harsh on counter comping - A zero-sum system like the one before severely punished you if you got hard-countered, cause there was shit you could do about it. This system takes into account some losses, and then some wins, and matches you at a rating with players of similar 'skill'. Counter comping is still powerful, but it's not as devastating as it used to be, making queue dodging not as neccessary. (or preferred)

- The system is more forgiving in general.

Cons:

- It takes more time to adjust this system, while the rating does bounce back fast according to your wins/losses and tries to quantify your skill quickly, you still need to play a good number of games to really reach a true rating, thus arena becomes a bit more time intensive than the old zero-sum system. Not by alot, but by a bit.

- Can you really quantify skill? Shouldn't the pro teams HAVE a gigantic win ratio? It's a good question. Is the fact that the system is more forgiving mean it's less skill-based? I'm not sure, they DID adjust all rating upwards, there are more people in the 2600s now for instance, so they made it a bit more forgiving but widened the results pool.


Either way, it was really stupid of them to put this in the middle of a season, as it fucked with everyone's already established ratings and was all messed up before it had time to adjust, lowering people it shouldn't have and giving rating to people that shouldn't break 1700.

I think the new system could actually be a bit more balanced for a class-based game (the zero-sum system would undoubtedly be better from an FPS or RTS standpoint, those games are designed with full balance between just one player fighting on even terms with another, or a team, etc) as it's less harsh on single performances (which we all know can be skewed) - at the same time, could it really just be dumbing down the skill required?

Who's to say.

EDIT: TLDR - The new system might not be so bad if it hadn't been thrown in mid-season.

[Taknar]

- Can you really quantify skill? Shouldn't the pro teams HAVE a gigantic win ratio? It's a good question. Is the fact that the system is more forgiving mean it's less skill-based? I'm not sure, they DID adjust all rating upwards, there are more people in the 2600s now for instance, so they made it a bit more forgiving but widened the results pool.

I don't think pro teams should have a gigantic win ratio. They don't need it; that's what the rating is for. Instead of the rating being "this is the number of nubs (like that scrub Taknar) I've pwned", it is representative of calibur. It states who you are of equal skill with, who you are greater than and most important to most what you deserve for that.

I only have two further questions about the system:
1) Is there rating decay, or can a team work its way up to the top and maintain it be not playing anymore?
2) It's 1 AM and I can't remember it.

[Yatokth]

I don't think pro teams should have a gigantic win ratio. They don't need it; that's what the rating is for. Instead of the rating being "this is the number of nubs (like that scrub Taknar) I've pwned", it is representative of calibur. It states who you are of equal skill with, who you are greater than and most important to most what you deserve for that.

I only have two further questions about the system:
1) Is there rating decay, or can a team work its way up to the top and maintain it be not playing anymore?
2) It's 1 AM and I can't remember it.

1 - No, people can work their way to the top and sit on it, I don't think this is a bad thing, the real peeps who get their shit in gear will outrank them anyway. If you have the stuff to make it to glad, you aren't going to get stopped by other teams sitting on top of high ratings.

2 - Well fuck.

[Naheal]

Arenas last night:

Our team lost 24 points for losing to one team.
After going against that same team again, we gained 4 points.

This must be fixed.

[Kredorian]

That was actually extremely helpful! :-) I was very confused about the new system... btw... holy priest ( PvE gear) and Ret Paladin ( PvE Gear) Not a good comp hahaha