MatchMaking - THE un-fun component

[CV_Jeebies]

Like clock work, MM is once again stacking teams to fail.  Bring in machine learning already so we can have balanced teams that get to actually get in a good fight.  It's long overdue.  Daily missions need changed -- if you hit the XP mark REGARDLESS if the team drops the ball, It should give you credit.  Last 5 matches ALL losses due to CrapMaking stacking teams.  I've exceeded 1400 XP every game and I'll be damn'd if my daily missions are unable to be completed due to potato stacking.  If the XP mark is hit, win or lose, it should be awarded.  Losses can't stack to hit XP, fine, but if XP marks are exceeded on a loss, don't penalize the player for lack of a proper team.  That's already its own punishment in itself. 

We're doing our part WG, do yours. 

[Ramsalot]

1 minute ago, NoSoMo said:

Last 5 matches ALL losses due to CrapMaking stacking teams. 

Working as intended comrade.  Ever watched roulette table in a real casino?  It’s easy enough to lose five times with perfectly equal odds.

[CV_Jeebies]

2 minutes ago, Ramsalot said:

Working as intended comrade.  Ever watched roulette table in a real casino?  It’s easy enough to lose five times with perfectly equal odds.

I'm not looking for opinions comrade, I'm giving them.  Roulette has nothing to do with how 28 humans of varied skill are placed in 2 teams against each other.  It's a poor excuse, for an even poorer system. 

[Doombeagle]

Just dodge

[CAPTMUDDXX]

Machine learning. Skill based MM. Same thing. Not happening.

@NoSoMo yes it is the same. You want MM to look at and learn from player stats so it picks from the queue players whose stats are comparable to yours to make up the two teams. 

Stats used in this manner makes for Skill Based MM.

[CV_Jeebies]

2 minutes ago, CAPTMUDDXX said:

Machine learning. Skill based MM. Same thing. Not happening.

Not the same thing - not even close.  Humans are incapable of spotting the nuances and variables that make balanced teams.

[Ramsalot]

19 minutes ago, NoSoMo said:

I'm not looking for opinions comrade, I'm giving them.  Roulette has nothing to do with how 28 humans of varied skill are placed in 2 teams against each other.  It's a poor excuse, for an even poorer system. 

Roulette is an example of statistically balanced system.  If you want balanced teams, roulette has more to do with it than you are willing to admit.

[Landsraad]

How many times are we going to have to repeat that MM only looks at ship type and tier when determining match-ups; and that win/loss streaks, stats, and consumables do NOT matter to it? Ramsalot hit the nail on the head with the roulette comparison, though I'd have gone for a hand being dealt in a card game, it's all chance where the unaccounted for variables are concerned.

[CV_Jeebies]

11 minutes ago, Ramsalot said:

Roulette is an example of statistically balanced system.  If you want balanced teams, roulette has more to do with it than you are willing to admit.

Roulette is not influenced by the numbers -- If each slot was larger or smaller, the odds of a ball landing in it would change.  Your assumption that MM is roulette and that all players represent equal odds, which we know to not be true, shows the fallacy of your comparison.    Trying to make MM equate to a random game of chance where each position has nearly equal potential, shows just how much one doesn't understand what it is that makes MM broken.

[Landsraad]

3 minutes ago, NoSoMo said:

Your assumption that MM is roulette is that all players represent equal odds, which we know to not be true, shows the fallacy of your comparison.

The players themselves all influence the game differently, true, but the odds of getting a particular player are always determined by the same thing: How many people of that tier and ship type are in queue. The player is the outcome, not the means of getting it in this scenario.

[CV_Jeebies]

6 minutes ago, Landsraad said:

How many times are we going to have to repeat that MM only looks at ship type and tier when determining match-ups --- 

Which is the problem.  Humans are incapable of keeping running tabulations on the 1,000+ variables that go into each player's play styles.  Humans are incapable of changing the influence they give to each variable on the fly.  You all seem to be assuming that stas are what I suggest -- stats is the dumb way of going about it.  I suggest a system that looks at each individuals play style in all aspects, their ability to play off of a ship, their inabillity to handle any number of ships, their position on the map, the map, the game style, the divisions, and any other variable that a learning system deems important.  We quickly leave the ability for a human to categorize in importance, let alone know which variables even have any affect in the process.  Achieving a truly balanced team is far outside of the realm of any human's capabilities, especially when it comes to adapting to things that while may be held trivial to any one person, may be of great importance.  I think we have a disconnect between what exactly machine learning is, and what it could do for balancing teams.  This has absolutely NOTHING to do with stats that many hold so close to their hearts.  Balancing on stats is the last thing that's needed.

[Umikami]

50 minutes ago, NoSoMo said:

Last 5 matches ALL losses due to CrapMaking stacking teams.

If your last 5 games have all been bad, maybe you should look at the common denominator; YOU! Maybe you should just get better and quit blaming MM for your dump-trucking 5 matches in a row.

[CV_Jeebies]

3 minutes ago, Landsraad said:

The players themselves all influence the game differently, true, but the odds of getting a particular player are always determined by the same thing: How many people of that tier and ship type are in queue. The player is the outcome, not the means of getting it in this scenario.

Which is the point -- see the humans assume that each ship needs a 1:1 opponent.  That's an assumption, one I believe to be very inaccurate given the limitless variables that go into how each match can play out.   An intelligent system combined with variable Q hold times (the system would predict the likelihood of another ship entering Q within a certain threshold) would provide the ultimate balance.  In the end the simple "how happy were you with that match" Q/A would provide a simple reporting feedback to be taken into consideration.

[CV_Jeebies]

1 minute ago, Umikami said:

If your last 5 games have all been bad, maybe you should look at the common denominator; YOU! Maybe you should just get better and quit blaming MM for your dump-trucking 5 matches in a row.

Go eat a stick -- if I want your opinion, I'll give you one.

[Umikami]

1 minute ago, NoSoMo said:

Go eat a stick -- if I want your opinion, I'll give you one.

Sure thing, killer! 

[Ramsalot]

11 minutes ago, NoSoMo said:

Roulette is not influenced by the numbers -- If each slot was larger or smaller, the odds of a ball landing in it would change.  Your assumption that MM is roulette and that all players represent equal odds, which we know to not be true, shows the fallacy of your comparison.    Trying to make MM equate to a random game of chance where each position has nearly equal potential, shows just how much one doesn't understand what it is that makes MM broken.

If you want truly balanced teams, the odds of winning for such teams would be 50/50, that’s what true balance is.  50/50 would turn the MM into roulette.

[Landsraad]

So your proposal is more feasible because, rather than simply saying "skill-based MM", you're asking that they put together a deep-learning AI and teach it to not only determine the skill of every single player in the game without looking at a single number but also predict who's going to queue up with what and when?

I think you OVER-estimate what machine-learning can do, and the spare computational processing power that WG has at their disposal.

[kgh52]

23 minutes ago, Ramsalot said:

Roulette is an example of statistically balanced system.  If you want balanced teams, roulette has more to do with it than you are willing to admit.

I do not believe the OP is willing to listen to reason at this time.

[CV_Jeebies]

1 minute ago, Ramsalot said:

If you want truly balanced teams, the odds of winning for such teams would be 50/50, that’s what true balance is.  50/50 would turn the MM into roulette.

No -- there's nothing about having balanced teams that'd turn it into a coin toss, or roulette. I think you just really like roulette. 

[CV_Jeebies]

3 minutes ago, Landsraad said:

So your proposal is more feasible because, rather than simply saying "skill-based MM", you're asking that they put together a deep-learning AI and teach it to not only determine the skill of every single player in the game without looking at a single number but also predict who's going to queue up with what and when?

I think you OVER-estimate what machine-learning can do, and the spare computational processing power that WG has at their disposal.

No, this can be done on a very low-powered system  -- it's already analyzed everyone from their recent games.  Each game someone plays, it goes back and looks if the players metric needs updated.  It's already got this info compiled and is simply comparing those who are in Q and what the odds are of another ship / player of a certain metric entering Q because it already knows what the running distribution is.  You don't analyze those in Q, you analyze everyone's last 30-50, maybe 100 games in various ships.  There's always going to be a level of uncertainty in the system because it's humans after all.  The benefit of machine learning is that it can go back over every game and analyze them so that the second it's put into control over balancing, it already knows a substantial amount about each player.  It's not going in blind, and once it starts balancing teams, it will begin to prioritize things that matter and change the outcome.   Every game that gets played, it gets better at matching opponents. 

[Ramsalot]

2 minutes ago, NoSoMo said:

No -- there's nothing about having balanced teams that'd turn it into a coin toss, or roulette. I think you just really like roulette. 

Okay, let me give you an honest example without mentioning the roulette.

MM finds enough players to make a game.  One of them is amazingly exceptional player, couple of decent ones, few horrible players, rest is average.  In order to properly balance this game, MM is going to place most of horrible players on the team with one strong player, and put both decent players on the opposite team.  Such MM would consistently punish exceptional players until they either quit or become average.  That’s what your balanced MM would mean.

[CV_Jeebies]

Just now, Ramsalot said:

Okay, let me give you an honest example without mentioning the roulette.

MM finds enough players to make a game.  One of them is amazingly exceptional player, couple of decent ones, few horrible players, rest is average.  In order to properly balance this game, MM is going to place most of horrible players on the team with one strong player, and put both decent players on the opposite team.  Such MM would consistently punish exceptional players until they either quit or become average.  That’s what your balanced MM would mean.

No, not at all.  Nobody is penalized.  It's analyzing play styles, not stats.  It's distributing skill among each teams.  As it stands now, there is no distribution.  Rarely do games go into an oh-so-close finish.  One gets steam rolled.  People enjoy the matches that do come with an even spread.  Potatoes are eaten and those who know what's up are left to fight each other.   Right now the teams with the most potatoes gets rolled as they melt, and then there's an imbalanced distribution of final players.   Even teams with unis get steam rolled because one or two players can only do so much.

[Landsraad]

3 minutes ago, NoSoMo said:

No, this can be done on a very low-powered system  -- it's already analyzed everyone from their recent games.  Each game someone plays, it goes back and looks if the players metric needs updated.  It's already got this info compiled and is simply comparing those who are in Q and what the odds are of another ship / player of a certain metric entering Q because it already knows what the running distribution is.  You don't analyze those in Q, you analyze everyone's last 30-50, maybe 100 games in various ships.  There's always going to be a level of uncertainty in the system because it's humans after all.  The benefit of machine learning is that it can go back over every game and analyze them so that the second it's put into control over balancing, it already knows a substantial amount about each player.  It's not going in blind, and once it starts balancing teams, it will begin to prioritize things that matter and change the outcome.   Every game that gets played, it gets better at matching opponents. 

That's still hundreds of games each with two-dozen players all happening at once, and what metric is it looking at anyway? With the parameters you mentioned before it seems to me like the system would have to be constantly watching every players' performance simultaneously, that's a LOT more to do than just aggregating post-game info like some websites do. By definition every single game is going to have to be looked at from every single perspective, because at one point or another every game is going to be the last one each person has played.

And I'm still trying to wrap my head around how you intend to teach this AI to determine someone's "playstyle" when even players have trouble discerning that about themselves and others, especially WITHOUT using stats. It's an already difficult task, and you're asking the computer-a device made and specialized in tracking massive amounts of numbers-to assess player performance without using numbers, but by "feel" instead. You may as well ask a person to watch a replay while blindfolded!

[Ramsalot]

16 minutes ago, NoSoMo said:

People enjoy the matches that do come with an even spread

Oh yes.  Even spread is what happens at higher levels of ranked play, when players get close to rank 1.  People enjoy it so much, they can’t stop expressing how much they are enjoying it each ranked season.

[CaptainKiwi_2016]

Am I the only one who feels that if Match Making was changed so that it picks players of even skill levels and places them in teams that the waiting time to get into a match would be increased dramatically?