Register    Login    Forum    Search    FAQ Awesomenauts



Post new topic Reply to topic

Author Message
 Post subject: Can we have actual matchmaking in brawls?
 Post Posted: Sun Dec 31, 2017 9:50 pm 
Offline
User avatar

Joined: Wed May 14, 2014 12:33 pm
Posts: 489
Location: here
Brawls are often terribly imbalanced when I play.
Most times I get players with 6 games played in my team that end up ragequiting, because of the enemy team killing them over and over.
I get that the higher league players are placed with the lower leagues, but in normal matchmaking the total league difference cannot be too huge.

What you end up with is a brawl where I play 1 v 3 vs lower league players that are having no fun due to the skill difference.

video example of such a game

This seems like a rather big problem, there even was a 6:30 timer before the game started.
Surely brawls are supposed to be fun, not ragequit inducing?

_________________
Lunk The Hero wrote:
Sam! wrote:
imo.


Your opinion is one of the reasons Sentry is in it's current state


Steam
Youtube [best montages]


Last edited by Jas123c on Sun Dec 31, 2017 10:04 pm, edited 1 time in total.

Top 
 Profile  
 
 Post subject: Re: Can we have actual matchmaking in brawls?
 Post Posted: Sun Dec 31, 2017 9:52 pm 
Offline
User avatar

Joined: Wed Aug 20, 2014 1:44 am
Posts: 2004
Location: Maybe far away or maybe real near by.
The wonders of a split MMS.

_________________
Diz Hydron wrote:
Catgame21234 wrote:
Hail Hydro

I'm not sure what I think about this. :think:


Winner of the Awesomenauts Guide Contest


Top 
 Profile  
 
 Post subject: Re: Can we have actual matchmaking in brawls?
 Post Posted: Sun Dec 31, 2017 10:30 pm 
Offline
User avatar

Joined: Tue Mar 26, 2013 6:44 pm
Posts: 2683
The way ronimos matchmaking works is that when the round pops, it ignores the last 1-5 people who are in the queue to round the total number of players in queue to a multiple of 6. Then it puts every remaining player in a game using it's matchmaking algorithm, ie "the best possible match".

Say there are 8 people in queue for brawl. The first 6 people are matched together. Thats your matchmaking. Not saying brawl queue is that small but even with 12-18 people in queue, your not getting a good matchup ever lol.

_________________
Steam: Tom
Gameplay: NEW Derpl Montage 3 is out!


Top 
 Profile  
 
 Post subject: Re: Can we have actual matchmaking in brawls?
 Post Posted: Sun Dec 31, 2017 10:57 pm 
Offline
User avatar

Joined: Wed May 14, 2014 12:33 pm
Posts: 489
Location: here
Tom wrote:
Say there are 8 people in queue for brawl. The first 6 people are matched together. Thats your matchmaking. Not saying brawl queue is that small but even with 12-18 people in queue, your not getting a good matchup ever lol.


Surely a new player that has played 6 games in total, should never get matched up vs someone with as many games played as me?
I've never had a player below league 5 in normal games, not even when the wait time was 30 seconds. :derp:
Also I doubt that there would only be 12-18 players looking for a brawl at a sunday evening.

Regardless of this, shouldn't we try to find a way to prevent brawls like this in the first place?
This was a terrible experience for all 6 of the players in this brawl.

_________________
Lunk The Hero wrote:
Sam! wrote:
imo.


Your opinion is one of the reasons Sentry is in it's current state


Steam
Youtube [best montages]


Top 
 Profile  
 
 Post subject: Re: Can we have actual matchmaking in brawls?
 Post Posted: Sun Dec 31, 2017 11:24 pm 
Offline
User avatar

Joined: Tue Mar 26, 2013 6:44 pm
Posts: 2683
Jas123c wrote:
Tom wrote:
Say there are 8 people in queue for brawl. The first 6 people are matched together. Thats your matchmaking. Not saying brawl queue is that small but even with 12-18 people in queue, your not getting a good matchup ever lol.


Surely a new player that has played 6 games in total, should never get matched up vs someone with as many games played as me?
I've never had a player below league 5 in normal games, not even when the wait time was 30 seconds. :derp:
Also I doubt that there would only be 12-18 players looking for a brawl at a sunday evening.

Regardless of this, shouldn't we try to find a way to prevent brawls like this in the first place?

When the queue timer pops it MUST match everybody in the queue bar the last 1-5, even if the matchup is garbage. If there are 1000 L9 players in queue and nobody else in queue bar a single L1, that L1 will get matched with 5 L9 players.

_________________
Steam: Tom
Gameplay: NEW Derpl Montage 3 is out!


Top 
 Profile  
 
 Post subject: Re: Can we have actual matchmaking in brawls?
 Post Posted: Sun Dec 31, 2017 11:50 pm 
Offline
User avatar

Joined: Wed May 14, 2014 12:33 pm
Posts: 489
Location: here
Tom wrote:
When the queue timer pops it MUST match everybody in the queue bar the last 1-5, even if the matchup is garbage. If there are 1000 L9 players in queue and nobody else in queue bar a single L1, that L1 will get matched with 5 L9 players.


Yes I get what you are saying, but if proper matchmaking would be in place. Then surely it would put the lower leagues with other lower leagues.
I had a 6:30 min queue timer so there should have been more than enough time for proper matchmaking to take place.

As far as I know, it keeps pinging other players in the wait time and only puts you in a game when the timer is at 0. But one enemy player had 300 ping, which seriously makes me question the matchmaking in brawls.

Obviously I don't know the number of people that were queued up for a brawl, but I was playing on a sunday evening, surely there should be plenty of people looking for a brawl?

_________________
Lunk The Hero wrote:
Sam! wrote:
imo.


Your opinion is one of the reasons Sentry is in it's current state


Steam
Youtube [best montages]


Top 
 Profile  
 
 Post subject: Re: Can we have actual matchmaking in brawls?
 Post Posted: Mon Jan 01, 2018 12:01 am 
Offline
User avatar

Joined: Tue Jul 28, 2015 10:24 am
Posts: 2590
Location: In my palace, mixing my martinis with paint thinner and a healthy squirt of epoxy adhesive.
Hopefully this thread evolves sufficiently that I can reference it the next some some nonsense-person decides to suggest that splitting the matchmaking queue for a potential draft mode will produce a better matchmaking experience.

On-topic: I am incredibly surprised by precisely none of your poor matchmaking experiences. The best part of this (not for you, of course) is that this situation will only get worse as the playerbase dwindles. What we need to do is go F2P and resolve that—

—oh.

Jas123c wrote:
I had a 6:30 min queue timer so there should have been more than enough time for proper matchmaking to take place.

You could give Galactron nine hundred years to form the perfect match: the result will still be dreadful if all the system has to work with are several metres of old string, a half-eaten apple, and one L7 player whose favourite flavour is Yuri.

_________________
Say no to SAM wrote:
Give this guy a duck for making such a good job with this thread

Nekomian wrote:
Give this guy a duck lololol

DeezNauts wrote:
Nobody can see it. Maybe instead of asking for ducks, you should put it in your signature.


Top 
 Profile  
 
 Post subject: Re: Can we have actual matchmaking in brawls?
 Post Posted: Mon Jan 01, 2018 1:13 am 
Offline
User avatar

Joined: Tue Dec 09, 2014 7:35 pm
Posts: 968
Location: Grenoble, France
Jas123c wrote:
Tom wrote:
When the queue timer pops it MUST match everybody in the queue bar the last 1-5, even if the matchup is garbage. If there are 1000 L9 players in queue and nobody else in queue bar a single L1, that L1 will get matched with 5 L9 players.


Yes I get what you are saying, but if proper matchmaking would be in place. Then surely it would put the lower leagues with other lower leagues.
I had a 6:30 min queue timer so there should have been more than enough time for proper matchmaking to take place.

As far as I know, it keeps pinging other players in the wait time and only puts you in a game when the timer is at 0. But one enemy player had 300 ping, which seriously makes me question the matchmaking in brawls.

Obviously I don't know the number of people that were queued up for a brawl, but I was playing on a sunday evening, surely there should be plenty of people looking for a brawl?

What if you were the only high league player playing a brawl? Ok to match you against higher league people, but where do you find them?

_________________
Admin of the Rollcage steam group
GRIP - Intense futuristic combat racer


Top 
 Profile  
 
 Post subject: Re: Can we have actual matchmaking in brawls?
 Post Posted: Tue Jan 02, 2018 11:25 pm 
Offline
User avatar

Joined: Tue Mar 26, 2013 6:44 pm
Posts: 2683
potterman28wxcv wrote:
Jas123c wrote:
Tom wrote:
When the queue timer pops it MUST match everybody in the queue bar the last 1-5, even if the matchup is garbage. If there are 1000 L9 players in queue and nobody else in queue bar a single L1, that L1 will get matched with 5 L9 players.


Yes I get what you are saying, but if proper matchmaking would be in place. Then surely it would put the lower leagues with other lower leagues.
I had a 6:30 min queue timer so there should have been more than enough time for proper matchmaking to take place.

As far as I know, it keeps pinging other players in the wait time and only puts you in a game when the timer is at 0. But one enemy player had 300 ping, which seriously makes me question the matchmaking in brawls.

Obviously I don't know the number of people that were queued up for a brawl, but I was playing on a sunday evening, surely there should be plenty of people looking for a brawl?

What if you were the only high league player playing a brawl? Ok to match you against higher league people, but where do you find them?

Basically what potterman28wxcv said but now that I have some more time let me explain it in more detail. Technically matchmaking uses rating, but for simplicitys sake I will just use leagues, in which I will assume all people in these leagues have the same rating. I will also ignore ping matchmaking to futher simplify. So in our simple example data we have:

-In the 6:30min queue timer 57 people hit play on the main menu and select brawl.
-The last 3 players are ignored and kicked into the next 6:30min queue timer.
-The league distribution of the players is as follows:
  • L1: 1 Player (Jas123c)
  • L5:5 Players
  • L8:12 Players
  • L9: 36 Players

In your opinion, what does MM do with this? It could create matchups in 2 ways:

Solution 1 (minimize individual skill level varience):
  • 1 match containing: 1x L1 player + 5x L5 players
  • 2 matches containing: 6x L8 players
  • 6 matches containing: 6x L9 players

Solution 2 (minimize total team rating difference):
  • 1 match containing: L1 + L9 + L9 vs 3x L5 players
  • 1 match containing: 2 teams of L5 + L8 + L8
  • 1 match containing: 2 teams of L8 + L9 + L9
  • 1 match containing: 6x L8 players
  • 5 matchs containing: 5x L9 players

I think I don't need to explain that the game with the L1 in (ie your game) is going to be completely onesided regardless of how MM matches it up, but all the other games a relatively equal. Tell me, what "proper matchmaking" do you want, it has 2 options and both have 1 terrible game and 8 "perfectly balanced" ones. I don't know how Ronimo's MM works but I think we can both agree from our knowledge of skill levels that option 1 is probably better for both the game you are in, and the 8 other matches.

The main point of this is there is only so much MM can do if there is an odd high league player in the brawl queue. You could run the most insane MM algorithm off of Nasa supercomputers for 100 years and you would not get a better solution for the sample data than one of the 2 I have presented. Either of these solutions are "proper matchmaking" based on the constraints of the problem (people gathered within the 6:30min queue timer).

What about ping (which I've ignored up until now), why did you face that 300ping player? Imagine that instead of 5 L5 players, we had 4 L5 players and 1 L3 player. Everybody has 100ping except the L3 player who has 300 ping to you, and 250 ping to everybody else. ie Our data looks like this:

  • L1: 1 Player (Jas123c)
  • L3: 1 Player (zorking 300 ping lagger!!11!!11!)
  • L5:5 Players
  • L8:12 Players
  • L9: 36 Players

In this case, our first solution looks like this:

  • 1 match containing: L1 + L5 + L5 vs L3 (300 ping) + L5 + L5
  • 2 matches containing: 6x L8 players
  • 6 matches containing: 6x L9 players

And our second solution looks like this:

  • 1 match containing: L1 + L9 + L9 vs L3 (300 ping) + L5 + L5 players
  • 1 match containing: 2 teams of L5 + L8 + L8
  • 1 match containing: 2 teams of L8 + L9 + L9
  • 1 match containing: 6x L8 players
  • 5 matchs containing: 5x L9 players

The 250-300 ping to everybody basically means there is nobody good to match ping wise. Your rating means that the only choice is to match the L3 300 ping player against you. ie There are soo few very high rating players queueing for brawls that your high rating is going to offsett the value of ping to an extent; Any player over L4 is going to get dragged into your games almost regardless of ping in this queue due to your high rating.

----

Ok example over, now onto the question of why does this dataset happen in brawls? I can answer this question with 2 statements which I'm not going to prove:
-A small fraction of the playerbase plays brawls, and the playerbase is bottom heavy.
-The hardcore playerbase cares less for "fun" modes than the casual playerbase.

I could go into a lot of detail but im going to keep this one quick: for arguments sake lets say 90% of the active player base queue for standard and the remaining 10% queue for brawl. Also your really highly ranked, so only the top 10% of the playerbase are sutible to be matched against you. Thus if 600 people were in queue, 1% of them are sutible to be in your brawl, so that gives you a match of 6 players (inc you). Thing is, only 15-20% of the playerbase is in queue at any one time, and at peak we have just under 1000 players atm, which is 200 players. So at even peak times you would struggle to get a single opponent near your skill level. Not only am I using extremely generous numbers here, but I am ignore so many factors such as ping and the likelyhood of a top tier player queueing for brawl at all.

In all honestly I think the matchmaking in brawls is probably fine for most people. I think we both agree that the skill descrepency between people in the top 10% of the playerbase is similar to the skill descrepency between the people in the bottom 50% of the playerbase. In other words, it's probably ok to put L1-2 in the same matchs as each other, and it's ok to put L5-9 in the same matches as each other. Not only are L5-9 the bigger leagues but the types of players in those leagues are more likely to play brawls. TL:DR; It's likely most brawls are balanced matchs comprising mostly of L5-9s with the occasional L3-4, and the ones YOU are in are the outliers where there is a stray L1-2 that MM can't do anything with.

----

I've already written an absolute wall, but I just want to finish this off for completeness sake as to what can be done about this. First off, if your really badly want to play the brawl, you can get them off of the workshop and get 5 high league friends to play it with you. Alternatively if you have 1 or 3 friends, you could all queue during the same round for the brawl, and so long as none of you are in the last 1-5 you should all get in the same match and split across teams with 1-2 noobs each.

Lastly, there's something that I really want to get off my chest, which is in your words "If proper matchmaking would be in place, then surely it would put the lower leagues with other lower leagues". I want to highlight "proper matchmaking" here. In my earlier example I stated how the solutions were the best MM could do with the constraints. However you can do better matchup wise if you change the constraints. Ie if the queue was 10min long you would get a larger dataset, ie better matches at the extent of queue time.

However there is another constraint: the absolute need to create matchs, ie if you have no latacy issues and queued before the cutoff for the last 1-5 ppl, you WILL be matched almost regardless of how bad the matchup is. In pretty much every other PvP game I play, this is not the case, you will frequently skip matchmaking rounds if there is no sutible match that fits strict constraints. Instead matchmaking rounds are far, far more frequent, ranging form 10-30sec per round. In other words, you queue until you have a good matchup (ie "proper matchmaking"). What this causes is shorter queues for the bulk of the playerbase and the extremes have huge queue times. I've had an average 20-25min LoL queue time back when I played (when most people had sub 1min queue times), because my MMR in the dominion game mode got so damn high that there was very few players to match me with. If this system was in place, you would be able to get balanced brawl matchs, but you would have to wait around 30-60min to get them (whilst waiting 10-15min for solo q normals, and double that for duo L1). However most people (L3 and below) queueing for brawl will have shorter queue times of around 3min (whist waiting 1min for solo q and 2-3 min for duo q).

Why Ronmio went for their system instead of the method most of the bigger games use is probably down to some of the following reasons:
  • They're amateurs. Ronimo are not malicious devs in any way, they are geuinely nice people and have always wanted what is best for the playerbase and the game. They're just still very new to PvP video game development (nauts being thier first and only pure PvP title), and do things you would expect from a first try, and are a bit unprofessional at times.
  • They didn't have a similar example MM system to take inspiration from and wanted to play it safe. Most matchmaking systems only consider skill, because they just match all the players on the same server together. Naut's matchmaking has to also consider ping between 6 players on peer to peer, vastly increasing the complexity of the project, so they didn't want to overshoot and simplified parts of it to make it more manageable.
  • They didn't want to risk making the game unapproachable for new players. There are very few beginners trying nauts outside of F2P launch, ie if F2P failed and new player intake remained slow, new players would be faced with huge queue times and this game would face a certain death and become unprofitable.
  • They didn't want to lose the non-NA/EU playerbases. Basically queue times for the smaller regions would increase to over an hour with this system, pretty much guarenteeing these playerbases a certain death. (I mean even with the current system, some regions took a hit)
  • They didn't want to upset the high end of the playerbase who tend to be more vocal. I think most of L1 would tolerate 10min queues, but I don't think they wanted to risk such a huge jump in queue times to thier biggest critics.

I'd take the 10-15min queue times and the inability to play brawl over playing against lag, crushing noobs and having to 1vs3 premades. However, It's a fine line between what's too much queue time and with the playerbase size of nauts, what I consider "proper matchmaking" could easily kill the game. I respect Ronimo's decision to play it safe with the queue method, but I still would have prefered regular rounds which could be skipped if there wasn't a sutible match.

_________________
Steam: Tom
Gameplay: NEW Derpl Montage 3 is out!


Top 
 Profile  
 
 Post subject: Re: Can we have actual matchmaking in brawls?
 Post Posted: Thu Jan 04, 2018 12:57 am 
Offline
User avatar

Joined: Tue Dec 09, 2014 7:35 pm
Posts: 968
Location: Grenoble, France
Tom, that is one of the greatest post I've read explaining matchmaking in a long time. Quite a wall of text, but worth a read for sure. :ayla:

_________________
Admin of the Rollcage steam group
GRIP - Intense futuristic combat racer


Top 
 Profile  
 
Display posts from previous:  Sort by  
 
Post new topic Reply to topic