An Objective Look at Ranked Matchmaking

An Objective Look at Ranked Matchmaking

in PvP

Posted by: Acandis.3250

Acandis.3250

Hello friends, my mission on this thread is to clearly expose and explain why seemingly lopsided matches have been plaguing players these last 2 seasons.

Rather that offer speculation about why certain decisions were made, I will be very deliberate about what the actual issues are and offer some suggestions as to how to fix them for future seasons.

Problem 1: No Pip Regression in Low Divisions. This is actually at the core of why matchmaking seems so bad. Here’s why:

There are currently players that can make it to Ruby (and beyond) winning a very high percentage of their games—they are either very skilled or getting carried, the how doesn’t matter here.

On the other hand, there are people that are getting into Sapphire for free simply by spamming games, even if they have a 25% win rate. Some of those players (though, not all) can eventually make it into Ruby, again with embarrassingly low win-rates. This is due to the fact that once you cross a tier in Sapphire, you can lose an unlimited amount of games but only need to gain 5 pips to cross the next threshold.

Now, it is my understanding, that the matchmaking algorithm currently tries to match you with plays within ~15 pips of you and then looks at MMR. In a system where every division is mostly populated by similarly skilled players, this would be perfectly fine. But this is not the case.

When you mix these two very different kinds of players in the same matches, bad things are bound to happen!

DETAILS BELOW. Skip to next marker if not interested

I’ll exaggerate here, as I don’t have access to data, but Ruby tier 1 & 2 host players ranging from 33% (speculation) to 100% (factual). This is a HUGE issue. To see this, we craft the following:

Let’s say the average 10 players in Ruby have the following MMR spread (from -50 to +50):
{ -25, -43, -8, -13, +2, +14, -30, -11, -17, +39 } Case 1
I have picked these numbers as 70% are C players, 20% are B players, 10% are A players.
Now, this isn’t always the case as A players are actually much more rare than 10% of the population, so you may actually more commonly see this spread:
{ -25, -43, -8, -13, +2, +14, -30, -11, -17, +15 } Case 2

A good match for Case 1 is to split the teams into:
{ -43, -30, -25, +14, +39 } vs { -17, -13, -11, -8, +2}
with average MMRs being: (-9) vs (-9.4).
(There might be a better match, but I didn’t spend to much time thinking about this)

In this case, we’ll very likely see Team 1 win, as their A player will probably be able to 1v2 and created good opportunities for his team to take advantage of—which is necessary, because they probably wouldn’t win an even fight.

What happens when the A player isn’t there? Case 2:
{-43, -30, -13, +14, +15 } vs { -25, -17, -11, -8, +2 }
with average MMRs being: (-11.4) vs (-11.8).

Despite the fact that Team 1 is favored to win here, their top 2 players aren’t that much better than Team 2’s top 2 players. Furthermore, Team 1’s bottom 2 players will probably be double capping every node and keyboard turning.

This is the type of game where Team 2 lucked out as they’re sort of similarly matched, but Team 1 is all over the place. It really isn’t fair to the top 2 players of Team 1 that they got matched up with players that (quite frankly) should not be in their division.

As you can imagine Case 1 games are actually pretty rare, with most games being Case 2 games.

This is what’s happening in all these Ruby games that then has people convinced that Ruby is inescapable!

END DETAILS

Even certain “good” players will have a very hard time in Ruby tier 1 & 2. Not because they’re bad, but just because they’re not good enough to carry the kind of players that shouldn’t be in that division.

Solution: All that being said, my suggestion to fix this is simply add pip/tier regression to every division.

Bonus Suggestions:

  • Sufficiently large lose streaks should drop a player down 1 division.
  • Ending a losing streak should not award an extra pip.

Problem 2: No rewards for any division except Legend.
A huge reason of why people get upset that they can’t get out of Ruby, is because the only division that shows that you did something worthwhile in PvP is Legend.

Although each player might have a sense of pride and accomplishment for making Diamond (and for some people Ruby) after a PvP season, they’ll have nothing to show for it because there’s nothing worth obtaining unless you made Legend. .
And please don’t tell me that the vendor backpieces are the nice eye-candy for the lower divisions, because I don’t think I will ever see willingly display that on their character.

Solution: We need something that is pretty, that isn’t embarrassing to display on our characters, that shows our accomplishments at the end of the season.

It doesn’t have to be something unique every season, it can be something as simple as a cape colored with a scheme according to the division you achieved that season, which when combined with other copies of itself in the mystic toilet, produces a cape of the same color scheme with a nice silver/gold trim.

Examples:

  • If I reach Ruby during Season 1, I obtain a Ruby themed cape. If I reach Ruby during Season 2, I obtain another Ruby themed cape, which I can combine with my first one to create a Ruby themed cape with silver trimming.
  • If I reach Sapphire during Season 1, and Ruby during Season 2: I get a Sapphire cape, and a Ruby cape, but they cannot be combined.
  • If I collected 2 silved trimmed capes of the same PvP tier, they can be combined to get a gold trimmed cape of that tier. (Ie. If you get Legend all 4 seasons, combine all your capes to get a nice gold trimmed Legend cape).

I realize this post is a bit long, but I think it highlights some issues which are paramount to fixing matchmaking.

To re-iterate, I think the current matchmaking algorithm would work perfectly, if there was pip/tier regression in every division.

The system is almost in working order, it’s just a matter of making these small changes for the future seasons, so that PvP may be more enjoyable.

Thanks for reading,
Ac

(edited by Acandis.3250)

An Objective Look at Ranked Matchmaking

in PvP

Posted by: Acandis.3250

Acandis.3250

—Reserved for incoming TL;DR version—

An Objective Look at Ranked Matchmaking

in PvP

Posted by: Sooloo.1364

Sooloo.1364

I think a lot more thought should have been put into gaining those legendary wings and achievements. Too much is dependent upon other people and factors totally out of our control.
The fractal legendary wings are still a difficult prize but very achievable by everyone with effort. There is no random factor deciding if the fractal fails or not 25 times in a row.

An Objective Look at Ranked Matchmaking

in PvP

Posted by: Deimos Tel Arin.7391

Deimos Tel Arin.7391

Bonus Suggestions:

  • Sufficiently large lose streaks should drop a player down 1 division.

no. players already lose divisions in legendary division, the other divisions are fine.

  • Ending a losing streak should not award an extra pip.

this is only acceptable if there are other reliable methods to gain extra pips.
so, no.

currently, the other ways to earn extra pips is to get lucky with a win streak.

Solution: We need something that is pretty, that isn’t embarrassing to display on our characters, that shows our accomplishments at the end of the season.

we already have the pvp badges behind our names. that’s plenty enough.

An Objective Look at Ranked Matchmaking

in PvP

Posted by: XGhoul.7426

XGhoul.7426

If this game had a healthy population I am all for it, but the only thing your system would do is alienate the “casuals” that need to play 200+ games to get to diamond and make the queue times rise more. In your system (with the amount of people playing right now) almost nobody would get legendary because all the pip pushers are being held back and can’t eat their carrot. The best way to do the next exploit is people deliberately boosting their alt accounts to boost their mains.

Long story short: The amount of pvp knowledge to not be one of the crybabies in forums requires so many games played that they still refuse to look at their own ability to improve or efficiently play with other ppl without voip. The game has been casual for too long.

An Objective Look at Ranked Matchmaking

in PvP

Posted by: Vieteriukko.6075

Vieteriukko.6075

It is balance between two bads. Guranteed progress in lower divisions supposedly increases pool of players the match maker pulls from. This may in some circumstances improve match making. Downscaling being not possible in lower divisions decreases mm quality.

I hope Anet will carefully analyze the data before S4.

An Objective Look at Ranked Matchmaking

in PvP

Posted by: Weerus.3701

Weerus.3701

Your solution doesn’t fix the problem of joining a season late, when everything you would end up being teamed up with is..

And with no bonus pips after loose streaks you would effective be trapped in certain division being forced to grind trough xxx matches to cross that division. Nothing left for you in ranked till the end of the season.

An Objective Look at Ranked Matchmaking

in PvP

Posted by: Cuchulainn.7421

Cuchulainn.7421

Solution 3: Narrow the MMR interval . Make it {-20;+20} Anet always claims that PvP population is large , so it should actually not be a problem.

Solution 4: Show the real MMR so we can trust the calculation. Show a real time search.

After 20 second -> 2 people found -> showing their real MMR.

After 1 Minute accumulation -> 6 people found -> showing their real MMR.

After 3 minutes accumulation -> 10 people found -> showing their real MMR.

Before match starts:

Warning to the weaker team: You are aware of the negative MMR distrubtion of your players compared to enemy team, do you want to start anyways?

PS: If you play and win you get 2 pips, if u lose u lose no pips.

So, with 2 pips incentive we get the people played although they have the weaker team.

So everybody is happy

Stronger team gets 1 pip if it wins. Weaker team gets 2 pips if it dares to play and win. If they lose they lose nothing at all

(edited by Cuchulainn.7421)

An Objective Look at Ranked Matchmaking

in PvP

Posted by: Acandis.3250

Acandis.3250

Your solution doesn’t fix the problem of joining a season late, when everything you would end up being teamed up with is..

I don’t think you finished your first sentence so I’m not sure what issue you’re referring to.
Personally, I don’t see any issues with joining a season late. If anything, the later in the season you play, the easier the games in lower division will be—in the current system.

And with no bonus pips after loose streaks you would effective be trapped in certain division being forced to grind trough xxx matches to cross that division. Nothing left for you in ranked till the end of the season.

I don’t understand your issue here. Evan has explained there is no MMR skip when crossing division thresholds; your games aren’t getting any harder to win once you cross to the new division.

The only reason people think they’re losing a lot more when they get to Ruby, is because suddenly people that could safely advance a tier by getting a few lucky games in a row and then tanking a bunch of games without dropping a tier, can actually drop to a lower tier, now. So naturally, Ruby 1 & 2 becomes the mystical “MMR Hell”.

Breaking your lose-streak won’t make it any easier for you to get out of this artificial Ruby black hole, but I could see how it might help you in Emerald and Sapphire.

Either way, it is a bad mechanic seeing as someone who loses 3 in a row, and then wins 3 in a row should be at a 0 net pips, but in actuality ends up with +2 net pips.

This seems all well and good for helping players slowly advance out of Emerald and Sapphire, where there is that 4 & 5 pip safety net every tier, but it is creating artificially horrible matches for 99% of players that make it to Ruby.

However, it is important to understand that spoonfeeding people out of lower divisions doesn’t prepare them for the quality of matches they’ll be placed in in higher divisions.

It is much better to create a system in which players are placed and held at their proper division until they have improved enough to advance. That way, every division will have mostly even matches—instead of what we see now, which is mostly uneven matches, across every division.

(edited by Acandis.3250)

An Objective Look at Ranked Matchmaking

in PvP

Posted by: Highlie.7641

Highlie.7641

Acandis your math is off

You lose 3, then you end your losing steak (+2 pips)
you win 2 more games (win streak) (2 + 1 +2)
You are at +5 pips.

Then you sit in Sapphire Last division and troll the rest of the season
(you get to create salty people in both sapphire and ruby!)

An Objective Look at Ranked Matchmaking

in PvP

Posted by: Acandis.3250

Acandis.3250

Acandis your math is off

You lose 3, then you end your losing steak (+2 pips)
you win 2 more games (win streak) (2 + 1 +2)
You are at +5 pips.

Yes, you are correct. Fixed.

(edited by Acandis.3250)

An Objective Look at Ranked Matchmaking

in PvP

Posted by: Exedore.6320

Exedore.6320

While you’ve identified some of the problems well, your fixes aren’t the best.

Pip Gain/Loss
Ideally players should be distributed evenly among divisions with amber having a little less. Removing division locking in Emerald and Sapphire partially fixes the problem, but doesn’t fix it all. Regression between tiers doesn’t really need to happen.

The lingering problems are these:

  • Number of Divisions Ruby and Diamond have many more divisions than Amber, Emerald, and Sapphire. Once the player base distributes itself, the number of players in each tier will be roughly proportional to how many pips it has out of the whole.
  • Pip Creation New pips must be generated at a healthy rate. By “new pip” I mean adding pips to the total pool – games where the pips gained and lost by all players does not sum to zero. The bonus win streak pip is a “new pip” because it adds a net +1 to the pool. Division and tier boundaries also add new pips because the net pip change for that game is positive. New pips are distributed to players over time and passed up divisions and tiers. If pip creation is too slow, players feel they can’t progress. Too fast and everyone gets to legendary. In the current system, the first 3 tiers generate a proportionally large number of new pips, which allows players to cruise to Ruby. But then pip creation drops off and players feel stuck. A smoother curve is needed.
  • Ranking vs Reward Is the tier/division system supposed to show player skill or be a glorified reward track? Currently it’s more the latter. More on this soon.

Rewards
ANet clearly wants a system where everyone can obtain items, but it may take a long time for average players to get them. The current system fails on many levels. Getting the legendary backpack is an annoying time gate. It doesn’t matter how good you; it matters how many days a week you play a minimal amount. There’s no benefit for most people to push higher; they’ll typically cap out and stop playing because there’s no benefit unless you cross a boundary. This is particularly noticeable with legendary. The only reason to keep going is for the dailies for the legendary backpack.

I would recommend fixing it this way instead:

  1. Remove the league currency from crossing tiers.
  2. Reward league currency for ranked wins. Higher tiers get bonus currency. A reasonable cumulative daily cap would be needed to prevent bots from farming it in amber.
  3. Add more stuff to the league currency vendor. Maybe some of that PvE or WvW exclusive stuff. Some things may need locked behind a minimum tier.
  4. Include an item which can be converted to gold (probably indirectly through the TP) for players who routinely reach legendary and will have everything after several seasons. Maybe the ability to buy PvP reward potions (daily limit).

That way, everyone can get something; better players obtain it faster. Less skilled players need to pick and choose and save up. And there’s also no obligation to grind dailies. It also feels less frustrating to be stuck in a tier looking only at bigger goals.

By making the reward system less tied to tier progression, it allows the ranked tier system to be more skill based than a progression system.

Kirrena Rosenkreutz

(edited by Exedore.6320)

An Objective Look at Ranked Matchmaking

in PvP

Posted by: Weerus.3701

Weerus.3701

Your solution doesn’t fix the problem of joining a season late, when everything you would end up being teamed up with is..

I don’t think you finished your first sentence so I’m not sure what issue you’re referring to.
Personally, I don’t see any issues with joining a season late. If anything, the later in the season you play, the easier the games in lower division will be—in the current system.

And with no bonus pips after loose streaks you would effective be trapped in certain division being forced to grind trough xxx matches to cross that division. Nothing left for you in ranked till the end of the season.

I don’t understand your issue here. Evan has explained there is no MMR skip when crossing division thresholds; your games aren’t getting any harder to win once you cross to the new division.

The only reason people think they’re losing a lot more when they get to Ruby, is because suddenly people that could safely advance a tier by getting a few lucky games in a row and then tanking a bunch of games without dropping a tier, can actually drop to a lower tier, now. So naturally, Ruby 1 & 2 becomes the mystical “MMR Hell”.

Breaking your lose-streak won’t make it any easier for you to get out of this artificial Ruby black hole, but I could see how it might help you in Emerald and Sapphire.

Either way, it is a bad mechanic seeing as someone who loses 3 in a row, and then wins 3 in a row should be at a 0 net pips, but in actuality ends up with +2 net pips.

This seems all well and good for helping players slowly advance out of Emerald and Sapphire, where there is that 4 & 5 pip safety net every tier, but it is creating artificially horrible matches for 99% of players that make it to Ruby.

However, it is important to understand that spoonfeeding people out of lower divisions doesn’t prepare them for the quality of matches they’ll be placed in in higher divisions.

It is much better to create a system in which players are placed and held at their proper division until they have improved enough to advance. That way, every division will have mostly even matches—instead of what we see now, which is mostly uneven matches, across every division.

I thought it’s easy to understand. When you join late in the season, only the “thrash” is left in lower divisions. That doesn’t make your games particularly easy. Sure it makes you win 1v1 and possibly 2v1 easily, but thats now how you win a match, remember? Imho it even increases stomp rates (100:500 or even worst), because not everyone is equally kittened.

Considering pips, the game should do fine if they would increase ranked requirements. Devs are more likely to lock players out of ranked untill 200 games on character done, than reconfigure whole system. Think of bad matches as a symptom, and for complete cure you should focus at the root of the problem. People started massively playing pvp for the wings and thus bringing their pve or achivement kitten skills. Increasing requirements for participation in ranked would fix everything.

An Objective Look at Ranked Matchmaking

in PvP

Posted by: Yannir.4132

Yannir.4132

The thing is that you would need to come up with a totally new league system for this to actually work. When there are no safety nets in place, the starting point for all players(let’s call it 0-point) needs to be in the middle instead of at the bottom. And since legendary divisions basically climb up forever, there is no mid-point.

If we were to remove Legendary divisions beyond the first one, the 0-point would be somewhere around early to mid ruby.

PS. Someone mentioned PvP noobs in ranked. Back when I started playing PvP(around August 2015), there was a minimum requirement of PvP rank 20. Has this been removed or something?

PSS. I’m not in favor of this proposal, because MMR has a tendency to give win-lose-win-lose streaks even in amber. So some people can’t rise to even sapphire simply by soloQing. The grind is not as easy for everybody as many ppl seem to think. I do think, however, that you should be able to drop tiers in sapphire already.

Yannir for males. (guard,thief,war,ele)
Sonya for females. (necro,rev,ranger,mes,engi)
All classes lvl 80.

An Objective Look at Ranked Matchmaking

in PvP

Posted by: Zaxares.5419

Zaxares.5419

XGhoul has it. If the PvP player base was a lot bigger, it might be a good move, but with current player numbers, I suspect all that your changes would do would be to drive out all the “casual” and “bad” players who are largely there for the AP and shinies. (For proof, you just have to look at what happens to PvP when a season finishes and you can’t progress the achievements anymore.) And then you’d be left with a ghost town in Ranked where you have 45+ min queue times, only to face the same few people over and over. (In fact, to hear people at Legendary tell it, this is ALREADY what they face.)

An Objective Look at Ranked Matchmaking

in PvP

Posted by: Acandis.3250

Acandis.3250

Ideally players should be distributed evenly among divisions with amber having a little less.

I don’t quite agree with this, if you have every division holding a roughly equal number of players, higher divisions lose prestige immediately. What’s the point of being a top 5% player if you’re being queued with the top 20% of the population.

  • Number of Divisions Ruby and Diamond have many more divisions than Amber, Emerald, and Sapphire. Once the player base distributes itself, the number of players in each tier will be roughly proportional to how many pips it has out of the whole.
  • Pip Creation New pips must be generated at a healthy rate. By “new pip” I mean adding pips to the total pool – games where the pips gained and lost by all players does not sum to zero. The bonus win streak pip is a “new pip” because it adds a net +1 to the pool. Division and tier boundaries also add new pips because the net pip change for that game is positive. New pips are distributed to players over time and passed up divisions and tiers. If pip creation is too slow, players feel they can’t progress. Too fast and everyone gets to legendary. In the current system, the first 3 tiers generate a proportionally large number of new pips, which allows players to cruise to Ruby. But then pip creation drops off and players feel stuck. A smoother curve is needed.

I agree largely with this, there are too many pips being handed out for free. It’s not good for anyone.

I would recommend fixing it this way instead:

  1. Remove the league currency from crossing tiers.
  2. Reward league currency for ranked wins. Higher tiers get bonus currency. A reasonable cumulative daily cap would be needed to prevent bots from farming it in amber.
  3. Add more stuff to the league currency vendor. Maybe some of that PvE or WvW exclusive stuff. Some things may need locked behind a minimum tier.
  4. Include an item which can be converted to gold (probably indirectly through the TP) for players who routinely reach legendary and will have everything after several seasons. Maybe the ability to buy PvP reward potions (daily limit).

That way, everyone can get something; better players obtain it faster. Less skilled players need to pick and choose and save up. And there’s also no obligation to grind dailies. It also feels less frustrating to be stuck in a tier looking only at bigger goals.

By making the reward system less tied to tier progression, it allows the ranked tier system to be more skill based than a progression system.

This is a fair solution, though it doesn’t fix the main issue with the Legend division which is: there’s no prestige tied to it. What’s the point of grinding Legend x17 times when you don’t get anything out of it? I know many players that stopped queuing for Ranked once they hit Legend this season, the system holds no further allure past that point.

An Objective Look at Ranked Matchmaking

in PvP

Posted by: Acandis.3250

Acandis.3250

The thing is that you would need to come up with a totally new league system for this to actually work. When there are no safety nets in place, the starting point for all players(let’s call it 0-point) needs to be in the middle instead of at the bottom. And since legendary divisions basically climb up forever, there is no mid-point.

If we were to remove Legendary divisions beyond the first one, the 0-point would be somewhere around early to mid ruby.

This is a valid point. A slight rework would be necessary. I have proposed in some other posts that if the first few games award hidden pip so that after 5-10 games you are placed in a division according to your performance.

PS. Someone mentioned PvP noobs in ranked. Back when I started playing PvP(around August 2015), there was a minimum requirement of PvP rank 20. Has this been removed or something?

This is still a think, but rank 20 is someone anyone with enough time can achieve in a day. It is simply not a good enough mechanic to filter new players.

PSS. I’m not in favor of this proposal, because MMR has a tendency to give win-lose-win-lose streaks even in amber. So some people can’t rise to even sapphire simply by soloQing. The grind is not as easy for everybody as many ppl seem to think. I do think, however, that you should be able to drop tiers in sapphire already.

The whole point of having a “Legendary” reward after reaching Legend is that the grind isn’t easy, and the reward holds some prestige. What we’re seeing now is that the grind is actually fairly easy, and the rewards don’t hold much prestige.

An Objective Look at Ranked Matchmaking

in PvP

Posted by: Yannir.4132

Yannir.4132

This is still a think, but rank 20 is someone anyone with enough time can achieve in a day. It is simply not a good enough mechanic to filter new players.

You would need around 10 hours nonstop PvP to reach rank 20, and that’s if you win every single game(which new players will not). That takes into account 10 min games, 5 min Q times, 1500 rank points per game. Brushing the exaggeration aside, I do see your point. Ranked should be reserved for more experienced players, so the entry could be at rank 40 or even 50.

Yannir for males. (guard,thief,war,ele)
Sonya for females. (necro,rev,ranger,mes,engi)
All classes lvl 80.

An Objective Look at Ranked Matchmaking

in PvP

Posted by: BlaqueFyre.5678

BlaqueFyre.5678

While you’ve identified some of the problems well, your fixes aren’t the best.

Pip Gain/Loss
Ideally players should be distributed evenly among divisions with amber having a little less. Removing division locking in Emerald and Sapphire partially fixes the problem, but doesn’t fix it all. Regression between tiers doesn’t really need to happen.

The lingering problems are these:

  • Number of Divisions Ruby and Diamond have many more divisions than Amber, Emerald, and Sapphire. Once the player base distributes itself, the number of players in each tier will be roughly proportional to how many pips it has out of the whole.
  • Pip Creation New pips must be generated at a healthy rate. By “new pip” I mean adding pips to the total pool – games where the pips gained and lost by all players does not sum to zero. The bonus win streak pip is a “new pip” because it adds a net +1 to the pool. Division and tier boundaries also add new pips because the net pip change for that game is positive. New pips are distributed to players over time and passed up divisions and tiers. If pip creation is too slow, players feel they can’t progress. Too fast and everyone gets to legendary. In the current system, the first 3 tiers generate a proportionally large number of new pips, which allows players to cruise to Ruby. But then pip creation drops off and players feel stuck. A smoother curve is needed.
  • Ranking vs Reward Is the tier/division system supposed to show player skill or be a glorified reward track? Currently it’s more the latter. More on this soon.

Rewards
ANet clearly wants a system where everyone can obtain items, but it may take a long time for average players to get them. The current system fails on many levels. Getting the legendary backpack is an annoying time gate. It doesn’t matter how good you; it matters how many days a week you play a minimal amount. There’s no benefit for most people to push higher; they’ll typically cap out and stop playing because there’s no benefit unless you cross a boundary. This is particularly noticeable with legendary. The only reason to keep going is for the dailies for the legendary backpack.

I would recommend fixing it this way instead:

  1. Remove the league currency from crossing tiers.
  2. Reward league currency for ranked wins. Higher tiers get bonus currency. A reasonable cumulative daily cap would be needed to prevent bots from farming it in amber.
  3. Add more stuff to the league currency vendor. Maybe some of that PvE or WvW exclusive stuff. Some things may need locked behind a minimum tier.
  4. Include an item which can be converted to gold (probably indirectly through the TP) for players who routinely reach legendary and will have everything after several seasons. Maybe the ability to buy PvP reward potions (daily limit).

That way, everyone can get something; better players obtain it faster. Less skilled players need to pick and choose and save up. And there’s also no obligation to grind dailies. It also feels less frustrating to be stuck in a tier looking only at bigger goals.

By making the reward system less tied to tier progression, it allows the ranked tier system to be more skill based than a progression system.

Removing the Safety nets and allowing players to drop division will result in a better representation of the players level since thy no longer want to advertise player MMR or match make using MMR, regression in tiers does need to happen otherwise you get the bottleneck that is currently seen in Ruby where every one eventually ends up.

The only real reward tied to the the League progression is the Legendary Wings, which are not meant for everyone you either have to be an average skilled player and reach at least Sapphire twice and Ruby twice over four seasons or hit Ruby 4 seasons in a row to obtain them, it should be more skill based than it is now to acquire them, yes they should add more rewards to the pvp vendor and award pvp currency based on wins vs tier crossings to allow players to access all of the other rewards.

An Objective Look at Ranked Matchmaking

in PvP

Posted by: shion.2084

shion.2084

My understanding is very different from yours. In your example…. . { -25, -43, -8, -13, +2, +14, -30, -11, -17, +39 } The game actually would produce the following two teams in this current season. (-43,-30, -25, -17, -13) vs. (-11, -8, +2, +14, +39). * IF my understanding is right*

It basically looks for candidates within a pip range, ranks them by MMR and plays a teams of consecutive MMRs against a team of 5 folks either the next 5 up or next 5 down. So if you have few players this will be a blow out.

The reason it looks like 50% win rating, is because you ended up in the better team and half the time the worse team. However since you can gain 2 pips, and not lose 2 pips in a game, and breaking a losing streak gives you 2, in general you’ll ascened.

The 50/50 ratio is from season 1. Where they tried to match games fairly. This lead to worse win one lose one games, because doing too well would place you with worse players so that the MMRs would average out.

In season 2 they just said you play people in division and we put folks of a similar MMR in a team together. This ended up with MMR hell where once you started losing you just got matched with worse and worse people. You could have the 5 lowest MMR folks on at the time matched against the 5 top ones in the division. It was only pure chance if it was a close game.

In season 3 they did A hybrid, they matched you with your closest MMR folks, but played you against the next highest or lowest ones. This combined with the pip advantage to win streaks and benefits for breaking a losing streak, would ensure that even moderately bad folks could grind their way up and that ideally games would be closer. So you can see they put some thought into it.

(edited by shion.2084)

An Objective Look at Ranked Matchmaking

in PvP

Posted by: Astralporing.1957

Astralporing.1957

I don’t quite agree with this, if you have every division holding a roughly equal number of players, higher divisions lose prestige immediately. What’s the point of being a top 5% player if you’re being queued with the top 20% of the population.

The point? Not having 30 mins+ long queues.

Actions, not words.
Remember, remember, 15th of November

An Objective Look at Ranked Matchmaking

in PvP

Posted by: Acandis.3250

Acandis.3250

I don’t quite agree with this, if you have every division holding a roughly equal number of players, higher divisions lose prestige immediately. What’s the point of being a top 5% player if you’re being queued with the top 20% of the population.

The point? Not having 30 mins+ long queues.

This isn’t so much a lack of population problem, as much as it is people stopping to play Ranked as soon as they hit Legend, since there’s no more to achieve.

An Objective Look at Ranked Matchmaking

in PvP

Posted by: Zaxares.5419

Zaxares.5419

Isn’t that itself indicative of the reason people are playing PvP? For the AP/shinies, and once there’s nothing more to obtain, they leave? That suggests that PvP itself isn’t “fun”, and more attention needs to be given to addressing that if ANet intends on maintaining longevity for the game mode beyond enticing players with shinies.

An Objective Look at Ranked Matchmaking

in PvP

Posted by: Exedore.6320

Exedore.6320

Ideally players should be distributed evenly among divisions with amber having a little less.

I don’t quite agree with this, if you have every division holding a roughly equal number of players, higher divisions lose prestige immediately. What’s the point of being a top 5% player if you’re being queued with the top 20% of the population.

Saying the divisions will be equally distributed was an over-simplification on my part. The populations will actually follow a bell curve (assuming player skill is a Gaussian distribution). The percent of players in each tier depends on its pip size, though you’d typically find less players on the extreme tiers (assuming equal pip creation rules across all tiers).

Here’s how I would expect the system to be tuned:

  • Figure out the average number of ranked games played per day. From that metric, compute the total number of pips gained per season by all players. This is easier said than done, but you can get a rough number. The complexity stems from (at least in the current system) different divisions have different pip generation rates and players will change divisions over time. And if players become frustrated, overall pip creation falls off. A system like the current one generates significantly less pips as the season progresses.
  • Assume player skill is a Gaussian distribution. You know the total population and the total pips. Get the mean and then distribute the total pips into the population using standard deviation.
  • Cut the tier boundaries (pips per tier) based on the percentage of players you want in each tier. Fudge it a little to make pips divide into divisions nicely.
  • If the pips are too few or too many, increase or decrease pip generation, but do it evenly across the population.

The difficulty in tuning this system is the calculation of total pips per season. If you undershoot, then too many people get into the higher tiers. Overshoot and almost no one makes it to the top.

And no matter what you do, someone who plays a lot of games will have an inflated number of pips due to luck (win streaks, etc).

Those are the problems with the tier system and trying to limit it by percent of the population. You can’t fix them unless you scrap the current ranked system.

This is a fair solution, though it doesn’t fix the main issue with the Legend division which is: there’s no prestige tied to it. What’s the point of grinding Legend x17 times when you don’t get anything out of it? I know many players that stopped queuing for Ranked once they hit Legend this season, the system holds no further allure past that point.

Did they stop solely because there’s no reward left to gain?
Was lack of enjoyment with the game a significant factor?
Would increased rewards compel them to play more?
Is some kind of numeric rank system needed?

My proposal addresses the lack of rewards by linking rewards to playing games instead of crossing tiers. Legendary players can continue to earn more exclusive rewards by continuing to play. It may not work for x17 legendary players, but that’s a fraction of a percent (and they’re probably in ProLeague or could be).

Personally, I gave up after reaching legendary in season2 because I wasn’t enjoying the game. The “carrot on a stick” approach only goes so far.


Removing the Safety nets and allowing players to drop division will result in a better representation of the players level since thy no longer want to advertise player MMR or match make using MMR, regression in tiers does need to happen otherwise you get the bottleneck that is currently seen in Ruby where every one eventually ends up.

Regression in tiers is not needed. The “bottom of Ruby” problem occurs because the path leading there is too forgiving and comprises too many of the total pips created. It is not because players can’t fall out of it. By removing inter-division safeguards in Emerald and Sapphire, you’ll prevent most of those “bottom of Ruby” players from ever crossing into Ruby.

Granted that while tier safeguards do keep some people at the bottom of a tier. But with Emerald and Sapphire being meaningful, those “bottom of Ruby” players will be distributed at the bottom of many tiers, which much more accurately reflects their skill.

Kirrena Rosenkreutz

(edited by Exedore.6320)

An Objective Look at Ranked Matchmaking

in PvP

Posted by: fishball.7204

fishball.7204

I feel like the Ruby “trash players problem” is severely overstated. Even my fresh to PvP PvE friends got to diamond this season and when I was climbing on my 2nd account I saw so many people playing poorly, blaming the system when they were in fact the issue themselves.

The safeguards in Ruby seem ok to me simply because I consider Ruby<→Legend the actual spectrum of player “skill” at the end of the season where Ruby is the worst and Legend is best (excluding pro league).

Don’t really need to fix what ain’t broken imo.

I kinda tl;dr your post though so I apologize if I cover something you already said or someone else mentioned.

FOR THE GREEEEEEEEEEEEN

An Objective Look at Ranked Matchmaking

in PvP

Posted by: Saiyan.1704

Saiyan.1704

I didn’t read every post but a few things players are forgetting.

Anet added the “end lose streak +1 pip” because players were literally not going anywhere. If you think it was bad this season, last season was literally twice as bad even if you dodged the bullet this go round. There were twice as many players who could not get a break for what ever the reason.

You have to take into account, casuals.
The current system caters to both, the players who don’t play religiously, running w/e build they feel like, and not looking at their minimap compared to the players who do play religiously, knows what situation counters their toon, and more importantly, knows what counters their teamates. The elites know how to rotate because they are more experienced. The system caters for both parties and that’s what Anet tries to accomplish.

Make it more difficult, casuals will hate you. Make it too easy and the elites will think there’s no point. There’s indeed items that needs addressing but players, from both parties, are going about it the wrong way in this thread.


Overwatch is technically in the same boat; players are completely reliant on their team to win games. Not a single player can carry a team… best they can do, just like GW2, is carry that one player who’s struggling but are still reliant on the other 3. So no.. a higher population will not help matchmaking. Legendary players however, there’s several solutions that Anet can approach to have Legy’s stay in the game longer per season.

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

An Objective Look at Ranked Matchmaking

in PvP

Posted by: Sooloo.1364

Sooloo.1364

Heres an idea for the matchmaker, if you’ve played a match with the same person or people multiple times and never won with them, don’t team me with them again.
For at least 20 games.

An Objective Look at Ranked Matchmaking

in PvP

Posted by: Exedore.6320

Exedore.6320

@Saiyan.1704

The “end lose streak” pip was in for season 2.

Lol Population has a lot to do with matchmaking. If the population falls below a critical threshold, then it has to either wait longer to find matches or expand the skill difference when looking for matches.

And part of skill is being able to work as a team. Players who can work as a team expect to be matched with other players who can work as a team.

Overwatch doesn’t have ranked available yet (ETA end of June). Right now everyone is playing a rough equivalent of hot join (there’s some basic matching to keep newbies separate).

Kirrena Rosenkreutz

An Objective Look at Ranked Matchmaking

in PvP

Posted by: Saiyan.1704

Saiyan.1704

@Saiyan.1704

The “end lose streak” pip was in for season 2.

Lol Population has a lot to do with matchmaking. If the population falls below a critical threshold, then it has to either wait longer to find matches or expand the skill difference when looking for matches.

Population has been at an all time high during the beta Stronghold release (when everyone joined 1 gamemode). Also the condi patch and post HoT patch brought on a plethera of players. A “critically low” population is not the reason why matchmaking sucks.

And part of skill is being able to work as a team. Players who can work as a team expect to be matched with other players who can work as a team.

This is only accurate after 200+ games or so… if that. Then theres a contradiction because the more games you have, higher the pip. Matchmaking looks within the pip range first and it doesn’t take class types into consideration. So there’s several variables that does all but accurately measure individual performance. Matchmaking should do what your saying, but that’s not always the case because it can’t. The leaderboard has to be grindy for the minority.

Overwatch doesn’t have ranked available yet (ETA end of June). Right now everyone is playing a rough equivalent of hot join (there’s some basic matching to keep newbies separate).

Ah that makes sense. Didn’t know Ranked was eventually coming. I guess consol players have DLC to look forward to?

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

An Objective Look at Ranked Matchmaking

in PvP

Posted by: Tiffany.8576

Tiffany.8576

And please don’t tell me that the vendor backpieces are the nice eye-candy for the lower divisions, because I don’t think I will ever see willingly display that on their character.

Hey, I use the Champion’s Wings on my ele cuz they look better than the huge, overbearing wings imo. This cut me deep.

Attachments:

Tiff | [TW] Tempest Wolves | WvW Staff Tempest Guide
NA/EU sPvP Elementalist

An Objective Look at Ranked Matchmaking

in PvP

Posted by: Highlie.7641

Highlie.7641

This will be an eclectic post, bare with me.

There are some things they should remove from MMR.

Look at what fishball wrote, that is a perfect example.
His friend just started playing which basically translates to a higher MMR then an active player. this mean’s right from the get go he is going to be favored by the MMR system. He will get “carried” to a higher division before his MMR settles. (which will cause problems for people that were suppose to be in said diversion)

-They need to remove the inflated MMR on New and Inactive accounts. It didn’t work when we had solo q (remember the top tier spots where filled with player’s with a 10-0 record) It’s not working now.

Match Making:
Alot of the wacky match making is due to bad balance.
“Woah~, B-but this game is pretty balanced atm”
and you are correct. but you are overlooking one very important thing:
Ease of Use.

Reset MMR on every account when the season starts. when it end’s pause that MMR value, and use it next season. this will prevent people like me from trolling.

(edited by Highlie.7641)

An Objective Look at Ranked Matchmaking

in PvP

Posted by: Saiyan.1704

Saiyan.1704

Overwatch matchmaking as predicted.

They’re labeling it the “Worse Rank system ever” but don’t take my word for it, It’s all over the forums.

  • You don’t earn more in a 5v6 win than a 6v6 win, a bug apparently.
  • Leaving before the time limit on “an opponent left the game, you can leave in 30s” counts as a loss in Ranked regardless if you Rejoined the game immediately.
  • Multiple bugs in terms of point distributions.

But lets say they fix all the “bugs”. Matchmaking will still be at a 50% win rate in soloQ. Only difference is their personal Rank, similiar to GW2 Rank 80, they can prestige once rank 100 hit. (Something I wanted in gw2)

Rank points is based on a 1 to 100 system. Majority of people are losing points at a 50% win rate, eventually everyone will be ranked 1-10…

I bet they’ll change the league system to something similiar to what GW2 has in the following months, regardless of whether or not “bugs” are fixed.

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

(edited by Saiyan.1704)

An Objective Look at Ranked Matchmaking

in PvP

Posted by: Exedore.6320

Exedore.6320

Saiyan.1704 the ANet fanboy spreading misinformation about things he doesn’t understand. He can’t even link the correct thread.

Yes OW ranked has some bugs – it’s been out for less than a week. The big one where KotH matches do not count enough is being fixed today. Player ratings are also very volatile right now. It takes time for any rating system to order millions of people when you know nothing about them at the start.

Outside of that, the rest is just ignorant people like you running their mouth.

  • The leaver system in place is by design. The principle used is that one person leaving can’t be used to take advantage of the system. Continually leaving ranked matches has a huge penalty associated with it (unlike GW2), to the point of being automatically banned from ranked play for the remainder of the season.
  • The outcome for coming back into a game during the grace period is likely a bug (haven’t seen a dev post yet).

In OW, your competitive rank is essentially percentile rank (1 to 100) based off your MMR vs. the MMR distribution of the population. My guess without having the data is that the initial distribution has most players in the 40-60 percentile range. Give it a few weeks for them to spread out – both up and down.

There is nothing in OW which forces a 50% win rate. That’s nonsense from idiots. Forums have a lot of those. Players do end up with a 50% win rate, but that’s because you’re they’ve reached their your proper rating and aren’t continually improving. The ones who are complaining likely think they’re better than they actually are.

Kirrena Rosenkreutz

(edited by Exedore.6320)

An Objective Look at Ranked Matchmaking

in PvP

Posted by: Saiyan.1704

Saiyan.1704

Saiyan.1704 the ANet fanboy spreading misinformation about things he doesn’t understand. He can’t even link the correct thread.

Yes OW ranked has some bugs – it’s been out for less than a week. The big one where KotH matches do not count enough is being fixed today. Player ratings are also very volatile right now. It takes time for any rating system to order millions of people when you know nothing about them at the start.

Outside of that, the rest is just ignorant people like you running their mouth.

  • The leaver system in place is by design. The principle used is that one person leaving can’t be used to take advantage of the system. Continually leaving ranked matches has a huge penalty associated with it (unlike GW2), to the point of being automatically banned from ranked play for the remainder of the season.
  • The outcome for coming back into a game during the grace period is likely a bug (haven’t seen a dev post yet).

In OW, your competitive rank is essentially percentile rank (1 to 100) based off your MMR vs. the MMR distribution of the population. My guess without having the data is that the initial distribution has most players in the 40-60 percentile range. Give it a few weeks for them to spread out – both up and down.

There is nothing in OW which forces a 50% win rate. That’s nonsense from idiots. Forums have a lot of those. Players do end up with a 50% win rate, but that’s because you’re they’ve reached their your proper rating and aren’t continually improving. The ones who are complaining likely think they’re better than they actually are.

You can drop the insults so we can have a reasonable debate here.

Everything I said was in reference to what the OW forums have stated and nothing more. Mmr needing to “settle” is the exact same thing Anet have said with their system. Considering not a single person from either game can carry a team, we’re both reliant on our team and comp. These aren’t blind arrogant assumptions here…

For simplicity sake lets say their bugs were worked out as they should be. OW’s Ranked system is only going to be slightly better than GW2 because the class system, by comparison, is more balanced. Now, here’s a few things why rank 1-30 = Amber Ruby, 30-50 = Ruby Diamond and why nobody except Consistent Teams will ever reach 55-100 or Legendary division.

The random soloQ experience is just like GW2.
You have players who don’t stand in Soldier 76 Heals, behind the defenses of a tank or out of zerg line of fire. People are spread out and not working as a team, an equivalent to the “1v1 experience” of gw2. Even better, players run out to die because the team is out rotated; team members respawning 1 by 1 without regrouping.

No doubt the rank 30-50 players or Ruby Diamond equivalent will work better as a team than the lower tier players but it’s going to 1) Take a while for players who legitimatly belong in that tier group to reach the Rank, not because mmr has to settle, but because they’re completely reliant on Matchmaking. And 2) They’ll never proceed any higher than that unless they’re on a team.

I play Overwatch more than GW2 now and think it’s a darn fun game so this isn’t fanboism… I’m merely explaining how similiar it is to GW2 and how OW is going to make the same MMR Rank system mistakes.

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

An Objective Look at Ranked Matchmaking

in PvP

Posted by: Exedore.6320

Exedore.6320

Sorry, but you continue to display a lack of understanding.

  1. Picking a popular whine post from the forums does not in any way mean it’s accurate or correct.
  2. This is the first time OW has assigned rating to players. GW2 assigned initial ratings over 3 years ago. GW2 rating has stabilized years ago. OW needs a couple weeks.
  3. GW2 leagues are a progression system, not a rating system. They don’t accurately represent skill. Numerous posts have explained this already. OW is a rating system only. You can’t equate GW2 tiers with OW rating.
  4. OW has a system for determining player contribution on a team when adjusting rating. It should help to filter out the good players who are stuck with lesser teammates and get them boosted a bit to proper ratings. Ranked play has been out for about 3 days in OW. Anyone complaining now based on personal experience is clueless.
  5. In a team-based game, players who queue as a well-organized team more often should be ranked higher. However, the matchmaking in OW does account for groupings and tries to play you against similarly grouped enemy teams (less chance for a full team vs. full solo) and so far, it seems to be working a lot better than it does in GW2. That should allow good players who queue solo to climb as high as other good players who queue solo.
Kirrena Rosenkreutz

An Objective Look at Ranked Matchmaking

in PvP

Posted by: Saiyan.1704

Saiyan.1704

Sorry, but you continue to display a lack of understanding.

  1. Picking a popular whine post from the forums does not in any way mean it’s accurate or correct.
  2. This is the first time OW has assigned rating to players. GW2 assigned initial ratings over 3 years ago. GW2 rating has stabilized years ago. OW needs a couple weeks.
  3. GW2 leagues are a progression system, not a rating system. They don’t accurately represent skill. Numerous posts have explained this already. OW is a rating system only. You can’t equate GW2 tiers with OW rating.
  4. OW has a system for determining player contribution on a team when adjusting rating. It should help to filter out the good players who are stuck with lesser teammates and get them boosted a bit to proper ratings. Ranked play has been out for about 3 days in OW. Anyone complaining now based on personal experience is clueless.
  5. In a team-based game, players who queue as a well-organized team more often should be ranked higher. However, the matchmaking in OW does account for groupings and tries to play you against similarly grouped enemy teams (less chance for a full team vs. full solo) and so far, it seems to be working a lot better than it does in GW2. That should allow good players who queue solo to climb as high as other good players who queue solo.

I knew and understand all of that. It’s not a perfect system and it’s new just like how gw2 pure mmr system was at first launch. What I disagree with is their system isn’t going to be all bells & whistles 3 weeks from now.

What I didn’t know was your #4 point. If it’s a system that actually calculates individual player skill then absolutely, it’ll be eons better than gw2’s rank system but if it’s another mmr calculation that “guesses” individual skill based on team wins and “guesses” that player’s mmr, then it’s no better than gw2’s pure mmr system at launch.

Do you know if it take kills/eliminations/deaths into account in its algorithm?

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

An Objective Look at Ranked Matchmaking

in PvP

Posted by: Exedore.6320

Exedore.6320

No details have been released on how OW calculates the magnitude of rating gains/losses, other than it uses many factors besides rating difference. I’ve read a few reports that it’s under-rewarding support roles (groups which always play together have their support player end up with a lower rating after placement).

And no one is saying OW’s matching will be perfect. Just that GW2’s system has a lot of problems. OW was brought up because it has similar challenges for ranked play, rating, and matchmaking and has done a much better job addressing them.

Kirrena Rosenkreutz

An Objective Look at Ranked Matchmaking

in PvP

Posted by: fishball.7204

fishball.7204

OW matchamaking post by their dev might be interesting to you guys:

https://www.reddit.com/r/Overwatch/comments/4p5a63/jeff_kaplan_has_addressed_matchmaking/

FOR THE GREEEEEEEEEEEEN

An Objective Look at Ranked Matchmaking

in PvP

Posted by: Saiyan.1704

Saiyan.1704

OW matchamaking post by their dev might be interesting to you guys:

https://www.reddit.com/r/Overwatch/comments/4p5a63/jeff_kaplan_has_addressed_matchmaking/

Thanks for that post. I had no idea how OW calculated MMR. Only thing I had to go by was the pseudo code of GW2, the Dev talk about MMR and how both games were similiar in SoloQ. Goes to show OW matchmaking really is “smarter” than GW2.

In the long term (once OW tweaks mmr math) OW may actually succeed in the competitive soloQ Ranked scene without utilizing a Kill/Death Ratio or Resources in its algorithm, like how most FPS/Moba/Strategy games calculate individual skill.

The soloQ pugs will still suffer the random comp queues (300k+ players that needs to learn the game will do that) but matchmaking in the higher percentile is going to be 5x more accurate in doing what it’s suppose to do because the pvp game is lenient that way.

GW2’s Matchmaking
Needs to be more static and overall more smarter in comparison. Matchmaking needs something more “fixed” to pick players for a fair game. There’s too many intricate variables like class dailies in Ranked, build differences that matchmaking is blind to and class swapping that matchmaking still considers as the same MMR. It’s an MMO first and a PvP second, sure, but the PvP scene still has a long (positive) growth moving forward.

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