Please upgrade your browser for the best possible experience.

Chrome Firefox Internet Explorer
×

Solo Que Matchmaking has a Serious Flaw - Data Inside

STAR WARS: The Old Republic > English > PvP
Solo Que Matchmaking has a Serious Flaw - Data Inside
First BioWare Post First BioWare Post

Elfa's Avatar


Elfa
11.21.2013 , 04:52 PM | #21
4V4 Arenas have done little but to show how utterly unbalanced the classes truly are.
I'd like to sit down with the guy responsible for having Busta Rhymes fight Michael Myers in HALLOWEEN RESURRECTION. Then, before he could speak, I would slap him so hard his face would explode and his dog would die." - Jim Law

JP_Legatus's Avatar


JP_Legatus
11.21.2013 , 05:20 PM | #22
Quote: Originally Posted by EricMusco View Post
We have seen similar posts on what you have reported in this thread. The rating disparity you have shown in your results is entirely possible, although very unlikely to happen on any consistent basis. I wanted to assure you that our team is aware of these reports and is tracking data behind our matchmaking to look for any potential issues.

Thanks!

-eric


Dev response in pvp forums what is this I don't even...

Thank you for visiting us and please do feel welcome here, it's always good to see new faces here on the missing-persons forum

EDIT: inb4 I get banned
Plaje - 65 PVP Assassin
http://www.twitch.tv/plajje

DarrelK's Avatar


DarrelK
11.21.2013 , 05:55 PM | #23
Another thing that should be taken into consideration is class and maybe spec. You know, it is ridiculous when you get 3 shadows vs 3 smash sentinels with a random healer in both teams.
It can easily get balanced to at least 2-1 1-2.
[TRE] Oberei - occasionally a decent shadow

AdrianDmitruk's Avatar


AdrianDmitruk
11.21.2013 , 06:31 PM | #24
Quote: Originally Posted by DarrelK View Post
Another thing that should be taken into consideration is class and maybe spec. You know, it is ridiculous when you get 3 shadows vs 3 smash sentinels with a random healer in both teams.
It can easily get balanced to at least 2-1 1-2.
I had a 4 merc vs. 3 sorc match once (I think we had a random sin thrown in for lulz).

We Healed to Full and Made Them Pay.

But of course mercs aren't exactly an OP class, as DPS they're somewhere in the middle and as heals, well welcome to the sorc world of constant interrupts.
Adrian-Defense Guardian | Kerrik-Vigilance Guardian | Matari-Scoundrel Healer | Treston-Pyromaniac Vanguard | Eamen-Sage Healer
Kendrew-Op Healer | Andreus-Sorc Healer | Matarion-Juggernaut Smasher | Darrien-Veng Juggernaut
The Makaryk Legacy of Healing to Full - Begeren Colony
Referral Link - Click for 7 days' sub time and other free stuff

stan_stilpleeze's Avatar


stan_stilpleeze
11.21.2013 , 07:46 PM | #25
Quote: Originally Posted by cashogy_reborn View Post
While I agree with the basic premise that the matchmaking sucks, you need a lot more than 1 game in order to prove your theory. If you had maybe 100 games, and it showed that there was a discrepancy between average team ratings then it would hold water.

But based on 1 single match, no. The data here is far too little to accurately represent anything.
Quote: Originally Posted by EricMusco View Post
We have seen similar posts on what you have reported in this thread. The rating disparity you have shown in your results is entirely possible, although very unlikely to happen on any consistent basis. I wanted to assure you that our team is aware of these reports and is tracking data behind our matchmaking to look for any potential issues.

Thanks!

-eric
I don't think the rate of occurrence actually matters that much. The mere fact this can happen once means it can happen any other number of times (given whatever circumstances lead to this). The problem I think that's being illustrated is that because this happened at least once, there's nothing preventing this from happening again and that's a big problem.

Speaking only from personal experience (which is really the only thing any player can do since we don't have access to the metric the devs do) there seems to be poor matchmaking. I can only offer up my own ranked queue history (as a healer, mind you) as a data point in the sea of player data points, but the fluxuation in my rating, I think, is far greater than it should be. The time between each match is consistently lengthy (save for the first days of the season), which would indicate the pool of players queueing is low, yet the average rating of each team is never the same between matches nor close to each other within the match (I check everyone's ranking before the match, but you don't have to take my word for this). Though this propels me to think that the matchmaking system throws the first 8 players of compatible roles together if there's not a match up in rating, this is purely conjecture.

As queue times are getting longer with each day (on my server, at least) something is clearly disenfranchising or dissentivizing people from queuing with the frequency they did at the start of the season. If the goal of rating is to create a normal bell curve with most players being average (which is presumably 1200, but would need developer insight to know their adaptation of the Elo rating system), the current implementation seems horrendous. If I'm in a queue population of similarly skilled players, my rating shouldn't change very much over the course of the day (~50% win-loss ratio), but being able to drop (or gain) 150 points in one sitting should only be possible if the I'm against players well above (or below) their relative skill. Statistically speaking, a random person has a ~68% chance of being one σ away from μ. But out of the 911 players on The Harbinger, 568 are below 1200, which means either the μ is lower than I thought (for instance, the median rating of all USCF members is 657) or the Elo isn't normally distributed (which is certainly possible, as USCF uses a logistical distribution). Any way about it, the system of scoring seems to have been a significant factor in discouraging people from playing. Elucidation on the part of a dev or community manager seems paramount in order to either clarify how their Elo rating system works, or possibly give examples as to how players might work their way up in rating (which actually gets harder the more games you play in the USCF system).

I have a lot of problems with the current state of pvp beyond just these issues, but I think the OP is bringing light to something that needs to be addressed.

Mursie's Avatar


Mursie
11.21.2013 , 08:18 PM | #26
what a joke - pot5 republic. here are my last two ques:

Match 1:
republic: 1161, 1271, 1205, 1394 (this was a recent transfer sage from SL that did less than 30k dmg)
imperial: 2106, 1956, 1928, 1534

Match 2:
republic: 1161, 1271, 1126, 1394 (same transfer sage)
imperial: 2256, 1956, 1784, 1534

Your game is garbage. With over 80 ranked games played on republic pot5 it has been literally the worst gaming experience of my life. Why I continue to play I do not know. I am literally that bored I guess. Regular wz's are meaningless and I am clearly insane thinking that next match around the corner will be a fun competitive one. But look at the above, this is a joke.

THE RICH GET RICHER.

orryko's Avatar


orryko
11.21.2013 , 08:37 PM | #27
Quote: Originally Posted by EricMusco View Post
We have seen similar posts on what you have reported in this thread. The rating disparity you have shown in your results is entirely possible, although very unlikely to happen on any consistent basis. I wanted to assure you that our team is aware of these reports and is tracking data behind our matchmaking to look for any potential issues.

Thanks!

-eric
musco, you know i got love for ya, but lesbehonest

forget matchmaking for a second...the ELO coding is completely trash

example: Achido(Skillx) was in solo queue the other night, on stream, @2200 rating. he was matched up against people in the 1200ish range. his team lost the match and he only lost 13 points. 13!!!!! Meanwhile, he beats other 1200ish people and still gets 7-9 per win!!!!

Does that look right to you or any of the devs? I mean really.

A loss between that disparity should be 40-50 points and a win should only be 1. Yes, I know there are averages and blah blah...but come on. 13 and 9 are very, very far off.

Same thing in grouped ranked really...the win and loss points are really messed up in some cases.

Mursie's Avatar


Mursie
11.21.2013 , 08:38 PM | #28
rofl.. and the last match classic:

republic: 1161, 1271, 1205, 1224
imperial: 2106, 1957, 1956, 1534

Is this a joke? THREE matches where my average group rating is 1200 vs. an average group rating of about 1900.

HOW DO YOU THINK THIS IS LEGIT?

HaLeX's Avatar


HaLeX
11.21.2013 , 09:09 PM | #29
A 1900 team going against a 1200 team would be the equivalent of a 50 point spread in college football.
GatorCountry GatorNation GatorUniverse POT5 SithUnited

JP_Legatus's Avatar


JP_Legatus
11.21.2013 , 09:29 PM | #30
Quote: Originally Posted by Mursie View Post
rofl.. and the last match classic:

republic: 1161, 1271, 1205, 1224
imperial: 2106, 1957, 1956, 1534

Is this a joke? THREE matches where my average group rating is 1200 vs. an average group rating of about 1900.

HOW DO YOU THINK THIS IS LEGIT?
Could be due to imbalance in faction ratings.
IE if all the pubs on that day average 1200 and the imps 1900, well this is the matchup you'll always see.

That's why I said if u get 2 or 3 like this in a row, just go do something else and come back later.
Plaje - 65 PVP Assassin
http://www.twitch.tv/plajje