44 replies to this topic

[MrMadguy]

There were 3 major issues in MWO:
1) Hitreg
2) Map design
3) MM

We will get EU servers on the next week, that should fix many hitreg issues. Also PGI given up on CW, that only minority of players need, and started to fix public queue maps with most terrible design instead. So... Only one major issue left to be fixed - MM.

Formally, I quited MWO, but I don't give up and try to return from time to time. But what I see, every time, I try to return? 10 roflstomps in a row? No, thanks. Yeah, may be I'm not the best player and sometimes I can underperfom, so may be this roflstomps are caused by me, but isn't MM exist for only one reason - to provide teammates and enemies, whose skill level matches my own?

So. MM should be completely revamped. Now it's ELO-based, which is totally wrong, cuz ELO is good only for 1vs1 or premade vs premade situation. Average ELO, that is currently used to measure skill level of PUG teams - is just like average temperature across the hospital. Also currently MM tries to control our W/L - which is also totally wrong, cuz W/L - it team-based factor and MM should control K/D or other personal-based factor.

MM should take into account your PERSONAL performance in specified 'Mech variant - not performance of PUG team as a whole, while not taking into account, how viable your 'Mech's build is. For example: if your team wins, but your match score is way too low (due to dying early for example) - then your team and enemies skill level is too high for you! Your ELO should not be risen in this case, as it happens now (simply cuz your team won) - IT HAS TO BE LOWERED instead. Cuz otherwise you're getting stuck in endless loop of roflstomps and so called "unpleasure matches" (matches, where your or enemies' performance it way too low).

Think about it - just one step to finally make this game enjoyable.

Edited by MrMadguy, 18 July 2015 - 08:29 AM.

[EgoSlayer]

Unfortunately you are wrong on most of your assumptions about the match maker.
1) Neither elo nor the match maker care, or track your win loss ratio. They don't track it, don't know it, and don't use it. That is a fact. Elo only looks at if you win or lose your current match to determine your elo rating change. MM only looks at your current elo score to slot you into a match.
2) Elo does work in groups because of the law of large numbers. Also a fact that has been proven in many studiies. Google is your Friend.
3) Stomps happen because of combat loss grouping. And happen more frequently in *evenly matched teams*. because once they are down units they don't have the skill reserve to overcome the remaining enemy numbers.
4) Elo skill matching is going to be undergoing some changes to reflect individual contributions rather than just win or lose. Think like the 'TrueSkill' system by MS. Again google is your friend.

But with better skill matching expect more stomps.

[Gattsus]

They mentioned they are going to implement a better ranking system that looks at individual performance after a game. PGI didn't give many details beyond this, though I'll give it a couple months.

[Cyborne Elemental]

I wish MM would take ECM into account when building a game.

To at least put an equal # on each side.

[FitzSimmons]

I'm an OK player but certainly not great. Occasionally, I will notice *very* good or *very* poor/new players. For the most part, people I play against are of similar quality. Stomps happen, but usually the case is that one of the teams loses a few players or puts themselves into a bad spot early, and it spirals out of control - not because of some insurmountable skill of one side or the other.

Most who I see complain about match-making are in group queue - which I suspect has difficulty matching #s of players and ELO score at the same time, and sacrifices in preference of 12 and 3/3/3/3 rather than skill levels.

Also: Confirmation Bias

Edited by FitzSimmons, 18 July 2015 - 09:53 AM.

[MechWarrior5152251]

He is right about MM, I am sick of getting top score in 80% of my matches. Maybe in mechs where you have had over 100 matches it may place you correctly, but not until a good 100 matches.

[Mystere]

Mister D, on 18 July 2015 - 09:28 AM, said:

I wish MM would take ECM into account when building a game.

To at least put an equal # on each side.

How about gauss rifles? Dire Wolves and King Crabs? Laser vomit? Artillery and air strikes? Jump jets? Where does it end?

[MrMadguy]

EgoSlayer, on 18 July 2015 - 08:49 AM, said:

Unfortunately you are wrong on most of your assumptions about the match maker.
1) Neither elo nor the match maker care, or track your win loss ratio. They don't track it, don't know it, and don't use it. That is a fact. Elo only looks at if you win or lose your current match to determine your elo rating change. MM only looks at your current elo score to slot you into a match.
2) Elo does work in groups because of the law of large numbers. Also a fact that has been proven in many studiies. Google is your Friend.
3) Stomps happen because of combat loss grouping. And happen more frequently in *evenly matched teams*. because once they are down units they don't have the skill reserve to overcome the remaining enemy numbers.
4) Elo skill matching is going to be undergoing some changes to reflect individual contributions rather than just win or lose. Think like the 'TrueSkill' system by MS. Again google is your friend.

But with better skill matching expect more stomps.

1) It care, but indirectly. You win - your ELO rises, so you start to be matched against stronger enemies => you win less. You lose - your ELO drops, so you start to be matched against weaker enemies => you win more. At some point amount of wins and losses becomes balanced.
2) No. Because XYZ team skill level =/= simple X + Y + Z sum. That means, that MM's initial assumptions are wrong => comparisons of this assumptions with real match results are also wrong => final ELO adjustments are wrong too => matchmaking is unpredictable and in most cases has just random results - ELO is randomly fluctuating or sticking in extreme position.
3) Chance that you will be put into team, that will "loss grouping", when teams are "evenly matched" - is 50/50. Right? Then why this is always MY TEAM? Nonsense.
4) That's what I'm talking about. In a PUG world only personal skill matters - any kind of team skill has no sense.

Edited by MrMadguy, 18 July 2015 - 10:17 AM.

[Kubernetes]

FitzSimmons, on 18 July 2015 - 09:28 AM, said:

I'm an OK player but certainly not great. Occasionally, I will notice *very* good or *very* poor/new players.

Yeah, this is the frustrating bit. Most of my games I see a number of recognizable names, players who are clearly much better than me (I avoid them like the plague once the match starts), and yet those same games also feature complete noobs (guys who do not target, shoot small lasers at distant targets, fire every weapon at once regardless of heat or range, etc.). Is it a problem of a small player base such that the MM has to throw all of us in together?

[EgoSlayer]

MrMadguy, on 18 July 2015 - 10:15 AM, said:

1) It care, but indirectly. You win - your ELO rises, so you start to be matched against stronger enemies => you win less. You lose - your ELO drops, so you start to be matched against weaker enemies => you win more. At some point amount of wins and losses becomes balanced.

No W/L ratio is an indicator of how good or bad the match maker is doing it's job, but it's an output of the match maker not a criteria to be used for making matches. W/L ratio isn't a factor that has any effect on the creation of matches -period.
And with elo all wins are not equal, and neither are all losses. You can win a match or lose one and have almost no change on your elo. It all depends what the expected results were. If you are at the point where you are getting matched properly and your W/L ratio becomes near equal that is a side effect of good matching, not an input.

MrMadguy, on 18 July 2015 - 10:15 AM, said:

2) No. Because XYZ team skill level =/= simple X + Y + Z sum. That means, that MM's initial assumptions are wrong => comparisons of this assumptions with real match results are also wrong => final ELO adjustments are wrong too => matchmaking is unpredictable and in most cases has just random results - ELO is randomly fluctuating or sticking in extreme position.

No it's not that simple, and it's why the law of large numbers applies. Google it.

MrMadguy, on 18 July 2015 - 10:15 AM, said:

3) Chance that you will be put into team, that will "loss grouping", when teams are "evenly matched" - is 50/50. Right? Then why this is always MY TEAM? Nonsense.

This assumes we have the player base to always created perfect matches. We don't. But even if we did, it's a probability not a certainty. It's not impossible to flip a coin and have it streak several times on heads.
Dunning Kruger and selective memory have an effect here. People are hardwired to remember bad things better than good things. Your stat history probably tells a different story, specifically your W/L record.

MrMadguy, on 18 July 2015 - 10:15 AM, said:

4) That's what I'm talking about. In a PUG world only personal skill matters - any kind of team skill has no sense.

And when people get their wish, and hopefully we have the player numbers for it; it's not going to change much if at all for the better. If anything it will be worse because more stomps happen when the teams are better matched.

[Mystere]

After all this time, it is quite obvious people still do not know how the matchmaker works. PGI should probably display how it works more prominently.

[Tarogato]

I guess you're in luck, devs told us that a major rework of the matchmaker is in the works.

(btw, people who talk about ELO automatically lose their right to try to talk about Elo)

[Cyborne Elemental]

Mystere, on 18 July 2015 - 09:54 AM, said:

How about gauss rifles? Dire Wolves and King Crabs? Laser vomit? Artillery and air strikes? Jump jets? Where does it end?

baby steps dude.. baby steps.

[Summon3r]

well OP you gave me my LMFAO for the day thx

if there is one major issue left in MWO it is MWO itself

[Mystere]

Mister D, on 18 July 2015 - 10:52 AM, said:

baby steps dude.. baby steps.

Yeah right. I heard that before.

[MrMadguy]

EgoSlayer, on 18 July 2015 - 10:38 AM, said:

No it's not that simple, and it's why the law of large numbers applies. Google it.

I really love those people, who use some "rocket science"ish words as arguments, assuming, that opponent doesn't know anything about it, so this he'll be beaten really easily, while not understanding, what they are talking about, by themselves. I know, what theory of large numbers is. And behind this scary words lays simple statement - "MM doesn't guarantee every single match or some streak of matches to be balanced, but overall your results will be balanced" or just "MM insures, that your W/L = ~1".

Read my post more carefully. I said, that ELO or W/L, that reflects your ELO level - are team-based ratings. But in PUGs team's win or loss =/= your personal "win" or "loss" (I mean enjoyment here - not actual win or loss). Team's performance =/= your personal performance. Team's enjoyment =/= your personal enjoyment. Team could win 100500 matches in a row and you could just hide somewhere at spawn point or be one-shot on a first minute during all of this matches. So when W/L = ~1 - it doesn't mean your personal "W/L", performance or enjoyment are balanced. Actually they can be random. And they will be. Because PUG - is just a bunch of RANDOM PLAYERS with RANDOM ELOS.

"Balancing overall result", i.e. insuring, that one bunch of random numbers is equal to other bunch of random numbers - isn't what MM intended to do. Providing enjoyment via competitive gameplay, i.e. insuring, that every number is balanced - is what MM is existing for.

Edited by MrMadguy, 18 July 2015 - 12:52 PM.

[EgoSlayer]

MrMadguy, on 18 July 2015 - 12:42 PM, said:

I really love those people, who use some "rocket science"ish words as arguments, assuming, that opponent doesn't know anything about it, so this he'll be beaten really easily, while not understanding, what they are talking about, by themselves. I know, what theory of large numbers is. And behind this scary words lays simple statement - "MM doesn't guarantee every single match or some streak of matches to be balanced, but overall your results will be balanced" or just "MM insures, that your W/L = ~1".

If you think they are "rocket science"ish or scary words you have little understanding of them. They are statistical analysis, nothing more.

MrMadguy, on 18 July 2015 - 12:42 PM, said:

Read my post more carefully. I said, that ELO or W/L, that reflects your ELO level - are team-based ratings. But in PUGs team's win or loss =/= your personal "win" or "loss" (I mean enjoyment here - not actual win or loss). Team's performance =/= your personal performance. Team's enjoyment =/= your personal enjoyment. Team could win 100500 matches in a row and you could just hide somewhere at spawn point or be one-shot on a first minute during all of this matches. So when W/L = ~1 - it doesn't mean your personal "W/L", performance or enjoyment are balanced. Actually they can be random. And they will be. Because PUG - is just a bunch of RANDOM PLAYERS with RANDOM ELOS.

This is exactly why the law of large numbers appiles. There is one constant in all that randomness - that individual player.
And the only way your scenerio plays out is if the player who is AFK all the time is continually put on teams with better and better skill levels that can carry them to a win every time. Which won't happen.

And nice moving of the goal posts by the way, now W/L doesn't mean the actual numbers it means your enjoyment. Sorry, not how statistics work. But congrats on proving your own OP post wrong that team elo and the match maker are broken.

MrMadguy, on 18 July 2015 - 12:42 PM, said:

"Balancing overall result", i.e. insuring, that one bunch of random numbers is equal to other bunch of random numbers - isn't what MM intended to do. Providing enjoyment via competitive gameplay, i.e. insuring, that every number is balanced - is what MM is existing for.

And to you that means the match maker should have you winning every match it sounds like, or at least stacking the deck in your favor so that you have a better chance to win than the other team. That's just a ridiculous proposal. The match maker is, and should be, completely unbiased in how it creates the matches. It tries for the best balance it can with the parameters and number of players it has in the pool. It's not deliberately stacking the deck for or against you.

Fact is that A) In equal skill level matches you are going to see winning AND losing streaks. B ) The closer the skill ratings in the match the *more* likely a win is going to be a lopsided stomp. Go check out the long 'Paging Karl Berg' thread for the details. It's all there. But I know you won't so I'm going to stop beating my head against the wall.

Edited by EgoSlayer, 18 July 2015 - 01:20 PM.

[Rushmoar]

Mystere, on 18 July 2015 - 09:54 AM, said:

How about gauss rifles? Dire Wolves and King Crabs? Laser vomit? Artillery and air strikes? Jump jets? Where does it end?

I agree. ECM has nothing to do with the match. If you have the "proper" amount of good players in your elo on your team, ECM can be over come in any game. The player is more important than ECM in any game.

From what I can tell, elo will be 50% your ability to win a match, if you win or lose. And 50% of your match score, how well you did in the match.

[bad arcade kitty]

EgoSlayer, on 18 July 2015 - 08:49 AM, said:

1) Neither elo nor the match maker care, or track your win loss ratio.

any proper elo is based on the win loss ratio
for each win or loss you get or lose a bit of rating (i.e. elo) where its amount is weighted by the elo of your teammates and your enemies

the proper elo doesn't care how exactly you won too, it doesn't care of your kill ratio, damage dealing etc stuff... not sure if it's true for mwo though, but if it is not it's a major flaw

also russ said that soloq elo doesn't raise properly... so it's kind of meh in mwo right now

mm is different from elo, it's an algorithm which applies elo/mm rating to find a match for you, mwo mm definitely uses a way too wide spectrum of the players' skill, that's for sure, it's probably due to low population of the game

Edited by bad arcade kitty, 18 July 2015 - 03:06 PM.

[Kiiyor]

EgoSlayer, on 18 July 2015 - 08:49 AM, said:

3) Stomps happen because of combat loss grouping. And happen more frequently in *evenly matched teams*. because once they are down units they don't have the skill reserve to overcome the remaining enemy numbers.

#1 fact.

Most players find it difficult to grasp because it isn't nearly as tangible as the numbers on the scoreboard itself. After first contact, your team may have taken 10 or 20 tonnes of armour from the enemy team, and traded 60, without any kills appearing - but your team has almost certainly already lost.

It's these figures that control a fight, not necessarily the MM.