69 replies to this topic

[Bonzai VI]

Played some games today. All played ~ 8 pm EST (NA prime time).
Used the Jarl's list to compare stats and look at the averages of both teams.

This REALLY shouldn't be happening.

(Didn't cherrypick anything btw - those are simply the games I played)

[Formatted everything so the first team is always the victor]


Game 1

Score: 12 - 2



Full stats

Spoiler

Game 2

Score: 12 - 1



Full stats

Spoiler

Game 3

Score: 12 - 6
(Nascaring on HPG ftw >->)

Spoiler

Game 4

Score: 12 - 6

Spoiler

At least one of them is close (at least the average is) amirite?

Edit: Thrown in some more info.

Edited by Bonzai VI, 06 March 2018 - 02:30 AM.

[MrMadguy]

PSR is at least 2x biased towards increasing, MM isn't functional at all - are well known facts. Nothing new.

[Bonzai VI]

Well known facts yes, but (I at least) didn't think for the MM to screw up THAT hard.

[Oldbob10025]

Problem is that when it cant get a decent team it throws MM out the window and puts players based on how many are playing not on skill level or Tier. So you will get wonky teams

[Jackal Noble]

Game 3.

[Bonzai VI]

@Oldbob10025

Sry didn't specify in the post: I played during NA prime time. Started playing at like 8 pm.

@JackalBeast

Wonderful, isn't it?

Don't quite understand how that should be fun to play :')

[MrMadguy]

Bonzai VI, on 06 March 2018 - 01:59 AM, said:

Well known facts yes, but (I at least) didn't think for the MM to screw up THAT hard.

Several bad games - isn't screwing up. Having bad game for months and years - that's, what is screwing up.

See the difference between season 19 and 20:

[Bonzai VI]

@MrMadguy

Meh, sadly those aren't even worse games than I normally have.

[Scyther]

The MM is a lie.

However TBH given the size of the player queue, and making some sort of equivalence between mechs on each side (which it apparently does try to do), I don't find those stat comparisons to be out of line.

Of course, my primary stats-based comparison is World of Tanks, where instead of getting a 72percentile team vs an 85, you can end up with a 48 percentile against an 80, not rare either. And that's working with a population that is easily 10x bigger than MWO, although the tank tiering throws a fair curve into it.

[Tarl Cabot]

Bonzai VI, on 06 March 2018 - 01:59 AM, said:

Well known facts yes, but (I at least) didn't think for the MM to screw up THAT hard.

Team of Solo queue and was your team communicating or all silent? Based on average match scores and games played all of the players are likely T1, may a T2 thrown in. There are no additional levels within PSR and that is all that the MM is using. And you have shown that within a Tier stomps will definitely happen for a variety of reasons.

Again, MM is doing its job within its parameters and likely did not even have to open up. So what is the issue other that there are not enough tiers with a PSR that has upward movement for most of the player base. If we did have a much larger player base there could be more tiers, or more parameters could be used when putting together a match, etc...

But I can definitely see why some would say it is a lie, an illusion since its primary purpose is to separate the truly new players from the hard hitting vets but once those who do improve and move up that is where the distinction, at least with the PSR/MM disappears.

Edited by Tarl Cabot, 06 March 2018 - 04:29 AM.

[Bonzai VI]

@Tarl Cabot

Game 3

My team had a matchscore ranging from 270 - 364 while 7 were over 300
Enemy team was ranging from 200 - 324 while 1 was over 300

Averaged it's 311 vs 256

See my problem with it now?

Edit: I'm actually not really talking about tiers, just the MM. It would be great if tiers were actually good for anything^^

That stomps happen isn't even the problem, it's the vastly unbalanced matches that are generated.

Edited by Bonzai VI, 06 March 2018 - 04:22 AM.

[Sjorpha]

On a positive note it looks like the jarls list is a good predictor of match results, adding further evidence that it's a pretty good ranking system. Kudos to the creators.

[Khobai]

Quote

It would be great if tiers were actually good for anything

they are good for something. keeping new players from playing vet players.

which is exactly what tiers were designed to do. so working as intended.

youre just expecting the tier system to be something it isnt. its not an ELO based matchmaking system. its an experience bar that keeps less experienced players from playing more experienced players. thats all it is.

PGI did say solaris will probably have ELO based matchmaking. And that they might add that to quickplay at some point. But that remains to be seen...

Edited by Khobai, 06 March 2018 - 04:54 AM.

[Asym]

Demographics don't support any match making system anymore....

I'd bet you'd see a huge right side curve instead of a bell curve. So, if all of the players playing are on one side of the curve, there's nothing to work with !!! You'll only get 12-1/2/3 or 3/2/1-12 games.

MWO is at a point of becoming "something else" because there is absolutely no player balance anymore.... Weapons balance came before player balance and now, you are reaping the rewards of that decision: a departed player base with crappy weapons.... Good luck with that. Remember that silly saying: "People First, Mission Always?"

What a mess.

[Novakaine]

After an extensive and intensive scientific study based on science.
I have indeed determined that the damned thing has gone hoopty.

[KodiakGW]

Notice how many sub 70th percentiles are in the same game with 95th+ percentiles.

Nope, nothing wrong with PSR. Stomps will happen!

Nope, no need to think about going to 8v8 to help even out teams. Looks fine to me!

Yep, stay he course. We’ll have a new developer, or no game at all, before the end of next year. Kudos!

[Marquis De Lafayette]

We all know the matchmaker is a mess, as everyone ends up in tier 1 eventually. So, those matches are likely full of tier 1’s (the average jarl’s rank isn’t exactly rock bottom for the losing team), but the MM thinks a 75% jarls percentile tier 1 is the same as a 95% percentile tier 1...and we all know better than that. The matchmaker just has no hope of figuring that out as it’s currently designed.

[Bud Crue]

While the MM is bonkers, broken, hoopty, or whatever other label folks may seek to pin on it, the reality is that the population at many times is apparently so low that even if the MM was functional and perfect in every way the poor thing would still be stuck making significantly imbalanced matches for extended periods. I’m not going to get into specific anecdotes, but suffice to say that this past weekend both group queue and solo queue were dominated for hours by many of the same names. We were all playing each other over and over and over. Add in significant skill disparity within that confined population set and you will get some real painful and statistically absurd results...that were often no fun for anyone. Nothing for it. And its going to get A LOT worse when Solaris arrives and when BT drops...


...but hey, I am sure the clever Solaris rebranding effort, and sudden upsurge in interest of this IP due to BT’s release and the vast marketing push of MW5, will soon draw in vast numbers of old and new stompy robot game fans/players to restore the MM to functional glory! A golden age of MWO interest and enthusiasm is nigh!

I for one can hardly contain my excitement.

[Bombast]

Event has started, I suggest trying not to think too hard about the matchmaker. The influx of 'seasonal' players and Trial mechs is really screwing it up.

[Bonzai VI]

@Bud Crue

I actually doubt a perfectly functioning MM would put two teams together that have an average matchscore disparity of >50.

A MM shouldn't force wins to go = 1, but put equal teams against each other.

And having 6 guys on one side equal or above rank 20k and on the other side 2 isn't really a good thing. (The first game I posted)

Edited by Bonzai VI, 06 March 2018 - 09:08 AM.