Jump to content
You need to play a total of 20 battles to post in this section.
KS_s6i6x

Yep MatchMaker Again

62 comments in this topic

Recommended Posts

14
[CDW]
Members
41 posts

 Why cant the MM pick out teams with the same or close to same player skill level ? As  I understand it in Random matches,  it choses ships that are roughly equal to each other. When they must be able to have it more on players skills equal. It says on the que page of Brawl that teams are picked by player skill level. So I guess it can..... Well I can 100% guaranty that ranked battles are NOT matched by any means equally by player skill .    For example in a ranked battle lets say we added up one teams individual number of Random games played together with the other 6 players number of random battles played, and did the same for the other team to get that total. Now with those totals we divided the numbers to get the average number of Random battles played per player. So lets say one teams average battle per player was 1799/player and the other teams was 7177/player. Which means on average that's 5378 more battles per player........... which team are you gonna put your money on. You can see you don't have to see in the future to know what teams gonna win if you knew these numbers pre-match. Would you say there's a problem with the matchmaking ??

 

I've had al lot of response to this . But no one has commented  on what my main question  about  example of rank battle  avg. game/player etc.

Edited by KS_s6i6x
  • Cool 1
  • Haha 3
  • Boring 5

Share this post


Link to post
Share on other sites
1,757
Members
6,889 posts
14,875 battles

Please read the thousands of posts already on this subject.  MM does NOT know the future, does NOT know how one player much less 12 or 24 will play together, and there is no real way to determine "skill".  WR is piss poor as it can be manipulated pretty easily.

  • Thanks 1

Share this post


Link to post
Share on other sites
14
[CDW]
Members
41 posts
21 minutes ago, CylonRed said:

Please read the thousands of posts already on this subject.  MM does NOT know the future, does NOT know how one player much less 12 or 24 will play together, and there is no real way to determine "skill".  WR is piss poor as it can be manipulated pretty easily.

I have read many of the posts on this.  Im not sure  about the WR. I asked nothing about it. And I agree MM does NOT know the future. Im pretty sure I already said that.  What I'm not sure about is  how I could ask my questions differently because an answer to them would be greatly appreciated.

 
  • Meh 2

Share this post


Link to post
Share on other sites
Wiki Editor
275 posts
8,136 battles
1 hour ago, KS_s6i6x said:

 Why cant the MM pick out teams with the same or close to same player skill level ? As  I understand it in Random matches,  it choses ships that are roughly equal to each other. When they must be able to have it more on players skills equal. It says on the que page of Brawl that teams are picked by player skill level. So I guess it can..... Well I can 100% guaranty that ranked battles are NOT matched by any means equally by player skill .    For example in a ranked battle lets say we added up one teams individual number of Random games played together with the other 6 players number of random battles played, and did the same for the other team to get that total. Now with those totals we divided the numbers to get the average number of Random battles played per player. So lets say one teams average battle per player was 1799/player and the other teams was 7177/player. Which means on average that's 5378 more battles per player........... which team are you gonna put your money on. You can see you don't have to see in the future to know what teams gonna win if you knew these numbers pre-match. Would you say there's a problem with the matchmaking ??

If teams were sorted by skill, it would become even harder to rank out. The expected wr for everyone will be 50% and someone with skill equivalent to a 65% wr will win the same amount of games as someone with skill equivalent to a 40% wr.

Edited by rhulkb27

Share this post


Link to post
Share on other sites
14
[CDW]
Members
41 posts
1 hour ago, The_Ruleset_Difference said:

because then people get que times like this thread:

 

So are you saying  as long as you don't have to wait in que  for too long its doesn't matter whether matches  are evenly matched or not ?  As well I sortuv don't think that would  end up being the case in Random

Share this post


Link to post
Share on other sites
1,401
[KAG2]
Members
2,248 posts
37,022 battles

queue time argument is just dumb, you do the balancing after the group of 24 is picked 

  • Cool 2

Share this post


Link to post
Share on other sites
Supertest Coordinator
7,223 posts
14,436 battles
9 minutes ago, vikingno2 said:

queue time argument is just dumb, you do the balancing after the group of 24 is picked 

You can do balancing all you want, but when you have a div of 2 Satsumas, for whose average win-rate is 42% faced off against two solo queue Satsumas whose win-rates are 55%, you're going to be in for a bad time. As long as people of approximately equal skill levels are allowed to div, you either have terrible divisions, or all-star divisions.

  • Thanks 1

Share this post


Link to post
Share on other sites
14
[CDW]
Members
41 posts
8 minutes ago, Compassghost said:

You can do balancing all you want, but when you have a div of 2 Satsumas, for whose average win-rate is 42% faced off against two solo queue Satsumas whose win-rates are 55%, you're going to be in for a bad time. As long as people of approximately equal skill levels are allowed to div, you either have terrible divisions, or all-star divisions.

That's what balancing means. You  wouldn't have the scenario you speak of  if it was properly balanced.

Share this post


Link to post
Share on other sites
7,159
[WOLFG]
Members
17,121 posts
20,520 battles
Just now, KS_s6i6x said:

That's what balancing means. You  wouldn't have the scenario you speak of  if it was properly balanced.

How?  forcibly split the divisions?  Make one team wait.

A 55% vs. 42% div is likely to imbalance the game no mater what the makeup of the other 20 players in the game.

  • Cool 1

Share this post


Link to post
Share on other sites
Supertest Coordinator
7,223 posts
14,436 battles
1 minute ago, KS_s6i6x said:

I find it a bit hard to understand how I can get an answer to my questions out of this graph.  Especially when the site referred to hasn't been developed for five years. Graphs from 2017 wont help me.

You claim that number of games played is indicative of "skill level." The API is old, but the data is fresh as of this morning.

Each dot represents a player. The number of games they've played, and their acocunt win rate.

image.png.c0a2474c80bdd4d28cc8c2311976861f.png

The average player with 10K games or more has a 50% win rate at best. Playing this game a lot does not correlate at all with game skill, especially when many of us have the ability to start over from scratch.

For example, Pikachu_lvl23 has a 78% WR with less than 1000 games, and I'd prefer him over almost every player with more than 8000 games under their belt.

https://na.wows-numbers.com/player/1037894430,Pikachu_lvl23/?

Share this post


Link to post
Share on other sites
5,927
[WOLF1]
Beta Testers
16,312 posts
23,671 battles
2 hours ago, KS_s6i6x said:

Why cant the MM pick out teams with the same or close to same player skill level ?

It can. WG has chosen not to do so for reasons explained in hundreds of threads going back to day one World of Tanks. 

Share this post


Link to post
Share on other sites
1,137
[NMKJT]
Members
4,287 posts
9,133 battles
5 minutes ago, KS_s6i6x said:

That's what balancing means. You  wouldn't have the scenario you speak of  if it was properly balanced.

Yeah you would, because you wouldn't be able to balance the divs. Even the 42% div wouldn't appreciate being split up to satisfy a bunch of people they don't know.

Share this post


Link to post
Share on other sites
14
[CDW]
Members
41 posts
2 minutes ago, Compassghost said:

You claim that number of games played is indicative of "skill level." The API is old, but the data is fresh as of this morning.

Each dot represents a player. The number of games they've played, and their acocunt win rate.

image.png.c0a2474c80bdd4d28cc8c2311976861f.png

The average player with 10K games or more has a 50% win rate at best. Playing this game a lot does not correlate at all with game skill, especially when many of us have the ability to start over from scratch.

For example, Pikachu_lvl23 has a 78% WR with less than 1000 games, and I'd prefer him over almost every player with more than 8000 games under their belt.

https://na.wows-numbers.com/player/1037894430,Pikachu_lvl23/?

Okay. I hear ya. So your thots is that it needs to be balanced on  WR percentage ?

Share this post


Link to post
Share on other sites
Supertest Coordinator
7,223 posts
14,436 battles
5 minutes ago, KS_s6i6x said:

That's what balancing means. You  wouldn't have the scenario you speak of  if it was properly balanced.

Yeah, those players would still be in queue after 20 hours because there's almost no opponents who they can match up against. There's 15 people on the server with a recorded win rate of less than 45% in Satsuma.

https://na.wows-numbers.com/ship/4178523856,Satsuma/

 

Is that an acceptable solution? Don't let them play the ships they want, ever, because there's no suitable opponents?

Share this post


Link to post
Share on other sites
1,757
Members
6,889 posts
14,875 battles
35 minutes ago, vikingno2 said:

queue time argument is just dumb, you do the balancing after the group of 24 is picked 

How?  Team is picked already - can't even remotely balance anything.

Share this post


Link to post
Share on other sites
Supertest Coordinator
7,223 posts
14,436 battles
Just now, KS_s6i6x said:

Okay. I hear ya. So your thots is that it needs to be balanced on  WR percentage ?

If I had all the data and all the effort in the world, it would be Experience + Win Rate by ship sub-type, and anyone with less than 100 data points for that type of ship would just be considered "not enough data" and be thrown into battle as randomly as possible. Once the data is collected, THEN you can adjust the parameters. Win-rate is a causal outcome that doesn't necessarily mean a player is good. Experience earned is more indicative of actual player effort, since it arises from all battle actions, including capping and combat.

Share this post


Link to post
Share on other sites
14
[CDW]
Members
41 posts
1 minute ago, Compassghost said:

Yeah, those players would still be in queue after 20 hours because there's almost no opponents who they can match up against. There's 15 people on the server with a recorded win rate of less than 45% in Satsuma.

https://na.wows-numbers.com/ship/4178523856,Satsuma/

 

Is that an acceptable solution? Don't let them play the ships they want, ever, because there's no suitable opponents?

No that isnt a suitable solution. But  its a team. Meaning  I think  an overall average of the team needs to be taking into concideration. An average of what I'm not sure. I just feel its worth discussing,

Share this post


Link to post
Share on other sites
1,757
Members
6,889 posts
14,875 battles
7 minutes ago, paradat said:

It can. WG has chosen not to do so for reasons explained in hundreds of threads going back to day one World of Tanks. 

Because it can't really determine what defines "skill" and no one plays at one "skill" level over time.  Anything that can be used is a rabbit hole that you can spend hours going down and try and prevent manipulation that things like WR are bad to use for "skill".

Share this post


Link to post
Share on other sites
14
[CDW]
Members
41 posts
6 minutes ago, Compassghost said:

If I had all the data and all the effort in the world, it would be Experience + Win Rate by ship sub-type, and anyone with less than 100 data points for that type of ship would just be considered "not enough data" and be thrown into battle as randomly as possible. Once the data is collected, THEN you can adjust the parameters. Win-rate is a causal outcome that doesn't necessarily mean a player is good. Experience earned is more indicative of actual player effort, since it arises from all battle actions, including capping and combat.

Wow.  A real  non argumentative answer. Thank you for that.  Back later with reply.

  • Haha 1

Share this post


Link to post
Share on other sites
14
[CDW]
Members
41 posts
32 minutes ago, MnemonScarlet said:

Yeah you would, because you wouldn't be able to balance the divs. Even the 42% div wouldn't appreciate being split up to satisfy a bunch of people they don't know.

Nobody would have to get split up. Just averaged in with the other 10 players on the team no ?

Share this post


Link to post
Share on other sites
5,927
[WOLF1]
Beta Testers
16,312 posts
23,671 battles
32 minutes ago, CylonRed said:

Because it can't really determine what defines "skill" and no one plays at one "skill" level over time.  Anything that can be used is a rabbit hole that you can spend hours going down and try and prevent manipulation that things like WR are bad to use for "skill".

And that is a good example of why WG does not do it. 

Problem one - What metric do we want to use?

Problem two - No matter what metric we decide to use, in short order that metric will regress toward the mean and so it becomes ineffective and we are all back in the same pool.

WG is very unlikely to ever willing go down this rabbit hole. 

I always prefer variety of game play I like my Random battles as random as possible. The fewer variables controlled the better.

  • Cool 1

Share this post


Link to post
Share on other sites
1,137
[NMKJT]
Members
4,287 posts
9,133 battles
19 minutes ago, KS_s6i6x said:

Nobody would have to get split up. Just averaged in with the other 10 players on the team no ?

That still wouldn't average it out, because divs are worth more in impact than solos. With divs of less skilled players that is still high potential impact, it's just high negative potential impact.

You can have triple BB or triple DD divs where those ships are the only ones of that type a team gets...and they frequently stick together. You can see triple or double DD divs concentrate on one side of the map leaving the other without DDs completely. In the case of BBs, that leaves the other flank without any good HP/armor units. It's almost impossible for MM to balance that just by queuing in two or three 42% players (using the example given before), because at least those hulls will be more spread out, so their value gets more evenly distributed to the flanks.

In other words the MM would have to wait for another div just like that to show up, and then make sure both teams are equally screwed.

And this is just talking about divs of lower skill. High skill players might have to wait a long long time, period, because there are a lot fewer of them.

The only way to balance these divisions would be to split players out of them, and like I said even lower skill players aren't going to appreciate being split from their friends they wanted to game with to balance a match.

Edited by MnemonScarlet

Share this post


Link to post
Share on other sites
Guest
This topic is now closed to further replies.

  • Recently Browsing   0 members

    No registered users viewing this page.

×