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

Tier 10 Matchmaker for Random Battles

22 comments in this topic

Recommended Posts

2
[DENY]
Members
5 posts
9,029 battles

I wanted to make this post to highlight a real issue with the matchmaking I've noticed occurring since the CV rework patch. I play ships of all tiers and right now I have 13 tier X ships. 

What I'm talking about is a significant disparity in skill between players in tier 10 ships in Random battles. In my opinion, this is a fault of the matchmaking system. To be clear, this skill gap exists in every tier for all of the players on the server, but it doesn't become a serious problem until tier 10. Tier X have the highest damage potential and are the most advanced ships in the game. When you're in a Tier X ship, the matchmaker assumes you are "Elite". There is no "tier 11".

Let me make 2 examples. You join a tier 10 random Battle in your Des Moines. You're pretty average, with your 49% winrate in Random battles. Noticing you have 4-5 other tier 10s on your team as follows:

Montana (lets say, 38% winrate and average 29000 damage per battle in Montana)

Gearing (44% WR)

Midway (46% WR, average 42000 damage per battle)

Zao (41% WR, average 23000 damage)

Then, looking at the enemy team you might commonly see a scenario like this:

Yamato (55% winrate 93,000 AVG)

Des Moines (51% WR 71000 AVG)

Grozovoi (54% WR 45000 AVG)

Hakuryu (56% WR 88000 AVG)

Like I said, everyone gets to play their random battles and the ship you choose has nothing to do with this. The matchmaker simply doesn't account for any of the visible player statistics and this is painfully apparent at tier 10 where teamwork and team support is paramount. 

In my opinion, the random matchmaker is doing very well when you consider tier 7 battles and lower. At tier 10, the matchmaker is like a malicious roulette that throws anyone together and has a tendency to create terribly unbalanced games. I've seen this often at tier 10, where a team will fold and die in 6-7 minutes with the enemy usually taking 0-1 losses. This is toxic and it wastes both teams time, and limits their rewards for playing. I've been on the giving and receiving end of this unbalanced garbage, and it started happening more often recently. 

World of Warships is like a boxing match, not a firing squad. The tier 10 Matchmaking needs to account for this and make incremental adjustments to improve the tier 10 experience.

Wargaming should do the following:

  • Assign each player a visible Tier 10 ELO matchmaking value that is influenced by all the public stats on your account such as winrate, damage, team contribution, etc etc.
  • Better player = Higher number
  • When you queue Random Battle with any tier 10 ship, the matchmaker normalizes players together, putting closer Tier 10 ELO ships into the same battle.
  • After 90 seconds in the queue, if nothing within the margin of error can be found (for example, if only 10 players are queued), then the battle starts as it normally would with no consideration of Tier 10 Elo values.

 

 

  • Cool 2
  • Boring 1
  • Meh 1

Share this post


Link to post
Share on other sites
3,310
[MUDDX]
Beta Testers
8,144 posts
22,441 battles

Perhaps you should submit this eloquent suggestion for skill based matchmaking directly to WG Staff and Developers.

Share this post


Link to post
Share on other sites
Members
5,263 posts
9,965 battles
21 minutes ago, Wobblebomber said:

Assign each player a visible Tier 10 ELO matchmaking value that is influenced by all the public stats on your account such as winrate, damage, team contribution, etc etc.

How much of each and over what period of time?  Overall stats or stats only in the ship being played?  There are a crap ton of possibilities with pluses and minuses of each and each questions open up even more questions.  Determining 'skill' is a huge rabbit hole once thought about for longer than  a few minutes.  None of that even takes into consideration the time to process.  Since the 'skill' levels are so different at any game and since we do not have 20-30K players at any given time on the NA server - I would really bet that the 4th rule would be hit most of the time completely negating the first 3 rules and creating more anger when people think the MM is not working right.  Much like we saw in the forums after the last rule change which made sure less uptiering happened and people took that to mean virtually to no +2 up-tiering at all.

23 minutes ago, Wobblebomber said:

Better player = Higher number

Better player by what stats?  At any time I can be unicun or super unicum according to WoWs Numbers - even in ships with low WR and the overall WR is average - 50.8% as a solo player.  Recently I began playing the Charles Martel where my numbers are not great and my WR is about 45% but yet I was unicum in it recently.

 

Share this post


Link to post
Share on other sites
Members
3,757 posts
23,311 battles

Another skill based MM thread where the "skill" is never defined. Please oh please tell us what skills players would be rated on? Still waiting on one credible suggestion that WG could implement over all these years.

Share this post


Link to post
Share on other sites
9,269
[SALVO]
Members
24,823 posts
25,888 battles
3 minutes ago, 1SneakyDevil said:

Another skill based MM thread where the "skill" is never defined. Please oh please tell us what skills players would be rated on? Still waiting on one credible suggestion that WG could implement over all these years.

Average base XP in the readied up ship?  Average damage in the readied up ship?  Those are two possible options.  Definitely not WR, because if they switched to a SBMM based on WR, your WR from the previous (i.e. current) MM model would become obsolete.  They'd really need to use some stat that wasn't linked to WR.

Share this post


Link to post
Share on other sites
Members
691 posts
7,037 battles

One issue is that ELO is designed specifically for chess and makes a couple assumptions which don't apply to team games. I think the other posters are also correct that you really need to be more specific in how player skill is quantified and presented. So you'll need to find a different and more specific rating system. Something that is simple for players to understand with some kind of easily remembered visual component for them to keep track of. So let's try that:

You said that you want to consider a host of statistics when calculating ratings, but many statistics aren't comprable across different ship classes and types. DDs consistently do less damage and get fewer kills than BBs. Will players have to sacrifice their rating in order to play DD? Conquerors consistently out-tank and out-damage many other ship classes. I don't think you want to encourage Conqueror spam with this system. So I suggest we restrict the system to just winrate; the most universal statistic, with maybe some consideration of base XP. Almost all tier X ships are quite well balanced within these two statistics.

So let's say that with each win a player's rating increases, and with each loss a player's rating decreases. We'll say that if a player is top of a losing team then he gets to keep his rating, so base XP is considered but not as important as winrate, to avoid encouraging selfish play and farming.

And you say you want players with higher ratings to be matched together, while other players get separate matches. That could be very easily implemented with a league system, perhaps with higher leagues getting some snazzy tag or symbol next to their names which show prestige, plus special rewards, both to avoid high skill players abusing this system to sealclub.

And let's just represent our rating system as a set of stars that you gain and lose with wins and losses.

Boy, this sure is starting to sound an awful lot like ranked battles

  • Cool 1

Share this post


Link to post
Share on other sites
2
[DENY]
Members
5 posts
9,029 battles

I just want to clarify for all the people who have responded:

I'm not trying to define skill. The World of Warships client already keeps stats on every game that you play. It's all there and its black and white, publicly viewable stats. For the Tier X MM ELO that I suggested, here is how it should work in theory:

 

  • Only applies to solo players and Random Battle (Not applied to divisions)
  • Only applies to tier X ships
  • Takes into account all relevant player statistics for that ships, the most important being Damage done and Win rate in that ship 
  • Seperate Tier X ELO applied to each Tier X ship that the player has

 

Share this post


Link to post
Share on other sites
Members
3,757 posts
23,311 battles
17 minutes ago, Wobblebomber said:

I just want to clarify for all the people who have responded:

I'm not trying to define skill. The World of Warships client already keeps stats on every game that you play. It's all there and its black and white, publicly viewable stats. For the Tier X MM ELO that I suggested, here is how it should work in theory:

 

  • Only applies to solo players and Random Battle (Not applied to divisions)
  • Only applies to tier X ships
  • Takes into account all relevant player statistics for that ships, the most important being Damage done and Win rate in that ship 
  • Seperate Tier X ELO applied to each Tier X ship that the player has

 

You can't call for this without defining skill...some stats kept are worthless to this request. Every time I call this out everyone runs from answering the question.

If you don't then you're at the mercy of WG to define it and will be very unhappy...lol

Share this post


Link to post
Share on other sites
572
[PEED2]
Beta Testers
2,016 posts
25,613 battles
2 hours ago, Wobblebomber said:

..................................................

Wargaming should do the following:

  • Assign each player a visible Tier 10 ELO matchmaking value that is influenced by all the public stats on your account such as winrate, damage, team contribution, etc etc.
  • Better player = Higher number
  • When you queue Random Battle with any tier 10 ship, the matchmaker normalizes players together, putting closer Tier 10 ELO ships into the same battle.
  • After 90 seconds in the queue, if nothing within the margin of error can be found (for example, if only 10 players are queued), then the battle starts as it normally would with no consideration of Tier 10 Elo values.

 

 

WG has consistently said that they will not use any "skill" based MM system for this game.

Share this post


Link to post
Share on other sites
1,388
Members
3,097 posts
12,268 battles
11 minutes ago, Wobblebomber said:

I don't run any mods 

For reference, I looked at data from this website: 
https://na.wows-numbers.com/

Yet, without skill based MM, skilled players manage to maintain above the average winning rates.  You should think about it.  What you are asking is the system where players with above the average metrics will be penalized consistently until they are average.

Share this post


Link to post
Share on other sites
2
[DENY]
Members
5 posts
9,029 battles
4 minutes ago, Ramsalot said:

Yet, without skill based MM, skilled players manage to maintain above the average winning rates.  You should think about it.  What you are asking is the system where players with above the average metrics will be penalized consistently until they are average.

I disagree, what I'm asking for is more consistent and challenging tier 10 gameplay. Maybe I'm a minority, but this is a competitive game and I like to be challenged when I play it. Wargaming can correct these imbalances with Tier X potatos. Player statistics indicate a lot of things about how consistent a player is:

 They aren't AFK. They aren't playing on a wooden PC from 2004. They can effectively communicate and support the team. They know how to aim. 

Share this post


Link to post
Share on other sites
1,388
Members
3,097 posts
12,268 battles
21 minutes ago, Wobblebomber said:

I disagree, what I'm asking for is more consistent and challenging tier 10 gameplay. Maybe I'm a minority, but this is a competitive game and I like to be challenged when I play it. Wargaming can correct these imbalances with Tier X potatos. Player statistics indicate a lot of things about how consistent a player is:

 They aren't AFK. They aren't playing on a wooden PC from 2004. They can effectively communicate and support the team. They know how to aim. 

You assume there's huge number of players in T10 queue so that MM can put good players with good players, and poor players with poor players.  This is not going to happen.  Currently there's limited pool of players, and generally there are more poor players that good players.  So if "the matchmaker normalizes players together" it will throw ALL of the poor players on the team with unicum player, and put the remaining good players together on the other team.  The system will be penalizing the better players, the better the player the bigger the penalty.

Share this post


Link to post
Share on other sites
224
[STAR]
SuperTest Coordinator
373 posts
11,017 battles

Skill-based MM won't save anyone. Even a unicum is capable of making mistakes or being incredibly unlucky by tripping over a torpedo or getting detonated. Even the most die-hard potato can be in the right place and by doing absolutely nothing, he'll be doing the right thing (eg going dark and not being killed, denying the enemy team most-needed points to get victory). Long story short: selection of players for match-making based on skill will keep giving the same (frustrating) result we're getting right now.

Please, don't get me wrong. I can fully relate with the discontent MM is creating. I've just had a few horrible team lineups in a row that made me leave the game for the rest of the day. I'll rejoin the game in the wee hours of Monday, when most Sunday drivers are gone. And the frustration unbalanced teams are probably the main source of toxicity in this game. I decided to suspend random matches for today after a sh*tstørm of curses being exchanged by both teams in three different languages. ...comical if that wasn't tragic.

But you have a point: MM for higher tiers should consider other things than just class and tier, especially during prime time, when the player base is much broader and allows MM to select players base on a few extra conditions with little impact on the waiting time (besides, I'd gladly accept a 1 minute increment in waiting time for more balanced matches). If personal skill isn't a reliable stat, then MM should take into account the experience one player has in that ship (not talking about avg. XP, just the plain number of matches in a given ship). A player in a t8 ship with just 2 matches is obviously running his ship in stock configuration and has measurable disadvantage when paired with another player with 46 games in that ship. The second has already upgraded the ship - at least partially - and has an advantage even if both are running different ships. One team has a stock ship, wich will perform worse than a fully upgraded ship. Furthermore, it is expected that the first player is still adapting to his ship (hence a lower that usual performance) than the second one. Although to measure how much better is one player than the other is disputable (and will lead back to the deadlock of skilled-based MM), it isn't wrong to assume that the second player will be more comfortable in his ship than the first one.

In t10, there's hardly any upgrades, but there are legendary upgrades that can be achieved after ~100 matches. And even though some of these upgrades don't give much of an advantage, some really do and are, again, a measurable edge over on player that has it and other that doesn't. You can argue that a player with 100+ matches in his t10 won't necessarily mount that LM, but it is expected of that player to be much more confident in driving his ship than one with only 5 matches in his record.

Having MM to consider the number of matches will also balance out unicums and (what I call) hopeless potato (HP). Don't get me wrong, I defend that even the hopeless potato has the right to join a random match, in his t10 for the 1000th time (and have a pitiful 37% WR in that ship) and be in the same team as I do. Both unicums and *special* potato have long records (300+) in a single ship. The problem is when you get 3 of the HP in one team and none in the other (and I have in mind exactly the last, ill-fated match I faced today, in which my team had 3 of the *very special* kind of potato, the avg. Nr. of games of the team was over 300, while the opposing team had an average of 92, but WR was merely 0,5% above the opposing team). Unicums are harder to come by (although, in higher tiers, they are more present than in mid- and lower tiers), but would be sorted in similar way, avoiding the concentration of 2 unicums on just one side, what usually has devastating effects on the enemy team.

I short, MM would not only pair class and tier (e.g. a Chapayev with an Edinburgh), but would also try to pair players with similar number of matches in that ship. Considering the player base, both player with similar number of games will have also similar stats in that same ships. One HP will probably be pair with another HP, while average players will be pair with average players and a unicum would be pair with a HP but - hey! - we already have this in the current MM, so no big difference here. Besides, unicums usually come in divisions (I wonder why) and would bypass all this being pair with just another division, just like current MM. The altered MM I have explained is to balance matches during prime time when the player base was much more players and the quality of team is perceived as declining by most players. 

The model will still be open for unbalance, as it's impossible to rule out every unbalance without making everybody wait half an hour in queue for just one 20-minutes- match. You could have one HP on one side and a unicum in the other team. But considering how seldom (real) unicums (not talking about average or good players) come by alone and how frequent HP are in prime time, the system might just work.

Share this post


Link to post
Share on other sites
7,846
[GWG]
[GWG]
Supertester
26,431 posts
14,165 battles
1 hour ago, Wobblebomber said:

I don't run any mods 

For reference, I looked at data from this website: 
https://na.wows-numbers.com/

MMM is a program you run that grabs the information for you but the principle of what I said still applies. Following that path leads to madness.

Share this post


Link to post
Share on other sites
5
[90THD]
Members
67 posts
6,806 battles
1 hour ago, Ramsalot said:

You assume there's huge number of players in T10 queue so that MM can put good players with good players, and poor players with poor players.  This is not going to happen.  Currently there's limited pool of players, and generally there are more poor players that good players.  So if "the matchmaker normalizes players together" it will throw ALL of the poor players on the team with unicum player, and put the remaining good players together on the other team.  The system will be penalizing the better players, the better the player the bigger the penalty.

You see, it doesn't matter how many are in the pool. Potato players get evenly placed on both team, so does average, good and unicum players. This means by the end, each team should have the same amount of potato, average, good and unicum players. The team will be more balanced this way.

Share this post


Link to post
Share on other sites
1,388
Members
3,097 posts
12,268 battles
7 minutes ago, 13579bc_bc said:

You see, it doesn't matter how many are in the pool. Potato players get evenly placed on both team, so does average, good and unicum players. This means by the end, each team should have the same amount of potato, average, good and unicum players. The team will be more balanced this way.

For the purposes of “normalization” that OP is talking about number of people in the pool matters.  Most of the time you will not have matching number of people with high skills.  Random distribution is good enough of a solution in the current system.

Edited by Ramsalot

Share this post


Link to post
Share on other sites
5
[90THD]
Members
67 posts
6,806 battles
2 minutes ago, Ramsalot said:

For the purposes of “normalization” that OP is talking about number of people in the pool matters.  Most of the time you will not have matching number of people with high skills.  Random distribution is good enough of a solution in the current system.

I think I might have misunderstood OP, sorry about that

Share this post


Link to post
Share on other sites
Members
74 posts
16,051 battles

I stopped playing solo because of poor MM.  I div up now almost exclusively.  55% WR solo, almost 80% in last month with Div mates.  I enjoy the game more with other people to play with on discord too.

  • Cool 1

Share this post


Link to post
Share on other sites

  • Recently Browsing   0 members

    No registered users viewing this page.

×