134 replies to this topic

[Kunae]

tuffy963, on 23 October 2013 - 08:55 AM, said:

to OP:
Seems like a reasonable assertion. Could be true.

Unfortunately, your solution is probably not a good fit for where the game is today. This would have been more reasonable for a game carrying the "Beta" tag, but might produce to disruptive an experience in a production environment.

Oh, I agree. I'd actually prefer it if they left their "Public test server" up all the time, so we could reasonably testing things like this.

The sad part about this, is that it didn't used to be this way. For whatever reason Elo had been working reasonably well for 3-5 months... it's only recently, especially since "launch" that it's really gone to hell.

LordVanquish, on 23 October 2013 - 09:15 AM, said:

Having said all that, I wouldn't be opposed to test out your idea. The only issue with this is to balance out the 4man premades and those PUG drop-syncing scum as well. If i can make a good 4-man premade and have the rest of the teams are random, we will stomp much much more than we will lose.

That could be ameliorated by PGI actually allowing 1-12 people to drop into the "12-man" queue, like they said they were going to do before they put our current, idiotic, queue system in.

[FerretGR]

Kunae, on 23 October 2013 - 08:46 AM, said:

No, I am absolutely correct about how matchmaking works, you are just misunderstanding. Within the ranges that are set, which grows as the "search" runs, it will balance a side that has a high Elo player/s on it with the bottom of the ever widening range, to reach the "average target" that it's looking for.

No, you really are not correct. You're basing this thread, and your entire anti-Elo stance, on this, and it's incorrect. It was discussed to death back in the day and it's simply not the way it works. There's no averaging of Elo. There's no bringing in a low-Elo player to average out the high-Elo player. It's not doing this by design. It's not doing this at all.

Now I'm not saying you're not seeing low-Elo players in your matches. What I'm saying is that they're not landing there by design, they're not there because the matchmaker is trying to average values, they're there because with a player pool as small as ours, the net gets cast widely enough in many cases to take in a low-Elo player with a high-Elo player. It's a last resort in order to obtain matches, not a first resort in order to maintain an average.

Kunae, on 23 October 2013 - 08:46 AM, said:

I also never said I was experiencing the beginner bracket. It's merely that players in or just out of that area are being dropped into the same matches as people playing at a moderately high level for over a year.

Yes. But that happens only when the matchmaking net gets widened by necessity. Without Elo, that could happen in every match.

That said, the fact that you admit there exists a "beginner bracket" means that Elo must be working!

ETA: here's the dev quote that I'm basing all of this on (ie. it's not coming out of my ***, and it's explicitly said that there's no averaging when selecting teams):

Quote

How does the match maker compose a teams Elo rating, is it average rating or closest to a target?

It's closest to a target value, so the match maker starts trying to make a match for an Elo of say 1300 and will pull in players to those teams closest to those values; however, as mentioned earlier within growing thresholds and those curves will be tuned. Currently it may be a bit 'sloppy' about how it's filling those buckets but over time it will be tuned to be much more precise.

http://mwomercs.com/...-making-update/

As an aside, we've been having this stupid debate for like 8 months now... it'd be a nice if a dev would come in here and set the ******* record straight for once and for all.

Edited by FerretGR, 23 October 2013 - 10:24 AM.

[Ghogiel]

Kunae, on 23 October 2013 - 08:51 AM, said:

Both of your underlined "points" are wrong. The range is much wider than you'd imagine.

Actually those 2 things you underlined as incorrect, is actually how it works. It does try to pick players closest to the target Elo for the match. Yes it does widen the net as the search goes on, originally it only did this after 2mins of searching, but with the recent update that may have changed.

I quote dev>
"It's closest to a target value, so the match maker starts trying to make a match for an Elo of say 1300 and will pull in players to those teams closest to those values; however, as mentioned earlier within growing thresholds and those curves will be tuned"

There are some parts we don't really know what it's doing though> Like as more players click launch, if there is a game waiting for players with an Elo target close to their Elo rating, do they get dropped into it and that accounts for some short wait times, or are they pulled into a game that's been searching for a player for a while and the range has expanded enough to just catch that player, even though they haven't been waiting long. Or both.

I want to think it's only doing the former with short wait time games because every time the MM goes to 2mins and beyond, the match quality is patchy. And the dev posts kinda say thats how it is. Some nights I just X out of MWO if the MM is taking long because I know it's not a match I want to really play in XD. The quick search times means I usually get in with good players. But I don't think we really know about that part of the search.

It's not even really trying to even out Elo on the teams either, so it's not actually trying to bring scrubs or good players in to make up the difference. What happens is once the total average Elo rating for the teams are calculated, it predicts which one is more likely to win. Say the one with the higher total Elo is predicted to win 67% of the time against that other team, due to Elo difference. The thing is your rating doesn't go down a lot if you had a higher predicted chance of losing.


Anyway what we can all agree on is Bryan needs to make a new Elo command post and address the questions raised about it, and update us on how it currently works, if they changed anything that contradicts the command posts on matchmaking.

Edited by Ghogiel, 23 October 2013 - 10:00 AM.

[Alexandrix]

At this point,the whole ELO matchmaking thing is really just a sad joke.Unfortunately,we are the butt of that joke.The other day I had a match where the enemy team had at least 1 4 man premade with tricked out meta mechs....our team had a newb in a trial dragon who couldn't figure out how to power his mech up.He had some how managed to power down...and then sit there unpowered while the enemy team hunted him down to finish the round because he couldn't figure out how to turn it back on.Also couldn't see us saying "HIT P" in the chat box either evidently.

Anyways,how does that even happen in a system that is supposed to balance teams by skill level? how can any match making system possibly consider that a "fair" match?

Edited by Alexandrix, 23 October 2013 - 10:00 AM.

[Crockdaddy]

Kunae, on 22 October 2013 - 06:26 AM, said:

This isn't about ego. This is about throwing new players, in trial mechs, in with people who've played 7000 matches... on purpose. This is what the current, broken, Elo system is doing. And, as Roland pointed out, it's doing this in preference to balancing the weight-classes between sides.

At least with no Elo, you will get a much more varied and mixed distribution of players in any given match.

I agree with this point. I am over 6000 matches. Like many of you, I see the same faces in most of my drops ... however it is luck of the draw as to whom gets the better pugs ... so to speak. I often am either stomping hard on a team which includes a few high ELO guys or I am getting stomped by a nicely run 733C group. My win percentage is around 60% ... I have no clue what that translates into for ELO score, but I do know I see most of the so called top end players often with a few unlucky new guys placed in the middle.

[Jman5]

Kunae, on 23 October 2013 - 08:51 AM, said:

While "perfect unicorn world" may act that way, that's not how it's working in matches. Both of your underlined "points" are wrong. The range is much wider than you'd imagine.

And as for #3? It's not see-sawing anything. It's placing one on a side and then trying to fit one on the other side, and then balancing with solo's Elo's to make it all work out. Weight matching is barely a factor at all.

The broken Elo system is king.

Please explain to me with evidence how you know the Elo range of the people you are matching with and against? Those numbers are hidden to everyone outside PGI HQ. Have you broken into their offices and looked up Elo scores? Or are you just assuming you're matching against low Elo players because of lopsided matches?

Look, opening up the floodgates to allow people who are at the top 5-10% to drop 4 man groups against anyone would be a disaster. They would just annihilate the competition. It would be great for them, but a completely horrible for 90% of the playerbase that they were unleashed upon. The odds of being faced against a team that was at their level would be so small that their win/loss ratio would climb dramatically. I mean, have you played against SJR, or HBA, or former KAOS/Dv8 guys? They routinely crush me and my Elo is apparently high enough to occasionally match against them. Imagine what they would do to a team of 12 average players?

I too remember what it was like pre-matchmaker and it was easy. I wasn't even that great at the game and it was ridiculously easy to solo assault mechs in my hunchback. It was extremely rare to meet a mech that outplayed me because there were so many noobs out there the odds were low I'd play against a better player.

I'm not going to presume your skill level, but if after thousands of games you are matching against people fresh out of noob-queue, perhaps it's not the matchmaker that needs work.

[Asakara]

FerretGR, on 23 October 2013 - 09:54 AM, said:

No, you really are not correct. You're basing this thread, and your entire anti-Elo stance, on this, and it's incorrect. It was discussed to death back in the day and it's simply not the way it works. There's no averaging of Elo. There's no bringing in a low-Elo player to average out the high-Elo player. It's not doing this by design. It's not doing this at all.

Now I'm not saying you're not seeing low-Elo players in your matches. What I'm saying is that they're not landing there by design, they're not there because the matchmaker is trying to average values, they're there because with a player pool as small as ours, the net gets cast widely enough in many cases to take in a low-Elo player with a high-Elo player. It's a last resort in order to obtain matches, not a first resort in order to maintain an average.

...

http://mwomercs.com/...30#entry2633230

Matthew Craig, on 07 August 2013 - 11:10 AM, said:

Technically the match maker will consider it a 'good' game if both teams were matched by having 1 high elo and 1 low elo player on both teams just as much as putting 2 average players on both teams.

The match maker can only work within the ranges given to it though and currently the ranges can be too large i.e. the matchmaker can drag in a high elo player to balance out a few low elo players, when we start to reduce the range what we should see is that the match maker will wait longer to create a 'good' game i.e it will have to wait for a player with a more average elo to show up (as the high elo player will be out of range). In the extreme case the match maker will simply give up and say it couldn't find a good game, this can potentially be seen as a good thing as you didn't want to play that game anyhow.

As mentioned we'll be monitoring carefully and we suspect we can tighten the ranges without having a dramatic impact on average wait times and failed matches, we can also adjust the timeout up from 2 minutes to compensate. Hopefully that answers your question.

[MooseofDoom]

Unfortunately, arguing the nature of the matchmaker being 'broken' or 'working is a rather moot point. None of us can actually see the ELO of our teammates or the enemy. To call a player a low ELO player because of a silly maneuver could be misidentifying a troll. Someone walked off the cliff straight into the enemy team?...Distracted by the phone or once again...a troll. You think they are a baddie because their loadout sucks or they don't know how to use it?...Could be someone trying out their assault for the first time and are having issues adjusting to the low speeds involved / firing angles...or a troll. Unless someone is going to sit down to record all of their matches, who was in them, w / l, inquire into the w / l of both teams, mechs and loadouts used and then post their results, I don't see how anyone outside PGI HQ is going to have anything better than hyperbolised anecdotal evidence for a broken matchmaker.

What it all really comes down to is this, if you aren't having fun stop playing and take a break. Nobody is forcing you to play this game, there is no contract signed or subscription paid for, you are free to come and go as you please.

This incessant whining about matchmaking and other such things by people that feel like they are being cheated in some way is pretty unlikely to force PGI's hand in development and if the last year hasn't already shown that to you well I don't know what help can be provided to you. It also detracts from helpful conversation that helps to direct new players or less skilled ones towards concepts that will make them into the type of player you don't despise for dulling your shiny W/L record.

[Alexandrix]

Jman5, on 23 October 2013 - 10:36 AM, said:

Please explain to me with evidence how you know the Elo range of the people you are matching with and against? Those numbers are hidden to everyone outside PGI HQ. Have you broken into their offices and looked up Elo scores? Or are you just assuming you're matching against low Elo players because of lopsided matches?

Look, opening up the floodgates to allow people who are at the top 5-10% to drop 4 man groups against anyone would be a disaster. They would just annihilate the competition. It would be great for them, but a completely horrible for 90% of the playerbase that they were unleashed upon. The odds of being faced against a team that was at their level would be so small that their win/loss ratio would climb dramatically. I mean, have you played against SJR, or HBA, or former KAOS/Dv8 guys? They routinely crush me and my Elo is apparently high enough to occasionally match against them. Imagine what they would do to a team of 12 average players?

I too remember what it was like pre-matchmaker and it was easy. I wasn't even that great at the game and it was ridiculously easy to solo assault mechs in my hunchback. It was extremely rare to meet a mech that outplayed me because there were so many noobs out there the odds were low I'd play against a better player.

I'm not going to presume your skill level, but if after thousands of games you are matching against people fresh out of noob-queue, perhaps it's not the matchmaker that needs work.

Gonna have to call out this one Jman.We've played together more than a few times.I regularly see you in my matches,either with or against.I also regularly see guys from the forums here that claim to be the bestest of the bestest.sometimes I kill them,sometimes they kill me.So I Have to assume we are in the same general area of ELO

So,assuming I'm correct and we are in the same general ELO,being fairly high up in that whole ELO bracket deal....then how can i end up with a trial dragon that doesn't know how to hit P and power up his mech on my team?

The matchmaker definitely needs work.Lots of it.

It's pretty obvious that the matchmaker grabs people of low ELO to balance out the ELO scores of high elo players on a team.Why anyone would think saddling a couple of high elo players with a bunch of steering wheel under hivers...and then placing them against a premade voice com team of average to fairly good players is a fair fight...I have no idea.

Funny enough,the matchmaker seemed to be in much better shape just a few months ago.Sure,you'd still get those guys that did 30 damage in an atlas,we all have a bad match sometimes,but the general skill level at least seemed to be a smidge tighter.Now? not so much.

Edited by Alexandrix, 23 October 2013 - 12:00 PM.

[Kunae]

Asakara, on 23 October 2013 - 10:40 AM, said:

http://mwomercs.com/...30#entry2633230

Beat me to it, Asakara.

@Those who were questioning my assumptions, what Asakara linked is what it's based off of, with some other bits from random PGI coding staff.

Edited by Kunae, 23 October 2013 - 11:35 AM.

[Kunae]

Jman5, on 23 October 2013 - 10:36 AM, said:

Please explain to me with evidence how you know the Elo range of the people you are matching with and against? Those numbers are hidden to everyone outside PGI HQ. Have you broken into their offices and looked up Elo scores? Or are you just assuming you're matching against low Elo players because of lopsided matches?

Look, opening up the floodgates to allow people who are at the top 5-10% to drop 4 man groups against anyone would be a disaster. They would just annihilate the competition. It would be great for them, but a completely horrible for 90% of the playerbase that they were unleashed upon. The odds of being faced against a team that was at their level would be so small that their win/loss ratio would climb dramatically. I mean, have you played against SJR, or HBA, or former KAOS/Dv8 guys? They routinely crush me and my Elo is apparently high enough to occasionally match against them. Imagine what they would do to a team of 12 average players?

I too remember what it was like pre-matchmaker and it was easy. I wasn't even that great at the game and it was ridiculously easy to solo assault mechs in my hunchback. It was extremely rare to meet a mech that outplayed me because there were so many noobs out there the odds were low I'd play against a better player.

I'm not going to presume your skill level, but if after thousands of games you are matching against people fresh out of noob-queue, perhaps it's not the matchmaker that needs work.

I know, as any good player does, who else is good. Mostly.

Also, when I spectate in matches, and I see people firing LRMs at point blank, and "sniping" in 3x zoom at things over 600m away with medium lasers, and SRMs, I can pretty well determine that those are newbies.

With 12v12, one 4-man makes much less of a difference, than it did in 8v8.

Alexandrix, on 23 October 2013 - 11:16 AM, said:

Funny enough,the matchmaker seemed to be in much better shape just a few months ago.Sure,you'd still get those guys that did 30 damage in an atlas,we all have a bad match sometimes,but the general skill level at least seem to be a smide tighter.Now? not so much.

Yeah, not sure what's causing that, but I had noticed the same thing. That's a good part of the reason for this post.

MooseofDoom, on 23 October 2013 - 11:15 AM, said:

Unfortunately, arguing the nature of the matchmaker being 'broken' or 'working is a rather moot point. None of us can actually see the ELO of our teammates or the enemy. To call a player a low ELO player because of a silly maneuver could be misidentifying a troll. Someone walked off the cliff straight into the enemy team?...Distracted by the phone or once again...a troll. You think they are a baddie because their loadout sucks or they don't know how to use it?...Could be someone trying out their assault for the first time and are having issues adjusting to the low speeds involved / firing angles...or a troll. Unless someone is going to sit down to record all of their matches, who was in them, w / l, inquire into the w / l of both teams, mechs and loadouts used and then post their results, I don't see how anyone outside PGI HQ is going to have anything better than hyperbolised anecdotal evidence for a broken matchmaker.

When you've played over 8300 matches, you get a pretty good feel for when someone is a newbie or just a bad. This is especially true if you happen to be spectating them.

[FerretGR]

Here's the thing. I don't interpret that as saying exactly what you guys claim it's saying, and coupled with the post I quoted from the original dev posts about Elo matchmaking, I think it just muddies the waters. Averages ARE used at one point in the Elo process: when determining the "who should win" and "who should lose" in a match, the average for the team is the determining factor. And what Matthew Craig is saying there is true: if you match two average players vs. one high and one low, you'll get two Elo averages that are close to each other, which is seen by the matchmaker as a "good" match despite that probably not being the case.

Now, he does go on to talk about the matchmaker pulling in a high Elo player to balance out a few low Elo players, but it's not explicitly said that the matchmaker is making the selection in order to average out the Elo... I read this as the end result, not as the deciding factor. It's at best a confusing addition to the discussion. It might be that you guys are correct. If so, it's a ****** piece of code. It would mean that the guy with the top Elo is almost always paired with a total scrub given a matchmaker that starts with an "in-between" Elo target.

Don't get me wrong, the original quote (the one I quoted above) was confusing too, but it certainly seemed to contradict what Matthew Craig says in the second quote. There's never been a super good explanation for how the matchmaker selects and sorts players. Which is why I suggest that we need a dev to, once and for all, give us a straight answer written in clear, well-worded sentences

[Asakara]

FerretGR, on 23 October 2013 - 11:52 AM, said:

Here's the thing. I don't interpret that as saying exactly what you guys claim it's saying, and coupled with the post I quoted from the original dev posts about Elo matchmaking, I think it just muddies the waters. Averages ARE used at one point in the Elo process: when determining the "who should win" and "who should lose" in a match, the average for the team is the determining factor. And what Matthew Craig is saying there is true: if you match two average players vs. one high and one low, you'll get two Elo averages that are close to each other, which is seen by the matchmaker as a "good" match despite that probably not being the case.

Now, he does go on to talk about the matchmaker pulling in a high Elo player to balance out a few low Elo players, but it's not explicitly said that the matchmaker is making the selection in order to average out the Elo... I read this as the end result, not as the deciding factor. It's at best a confusing addition to the discussion. It might be that you guys are correct. If so, it's a ****** piece of code. It would mean that the guy with the top Elo is almost always paired with a total scrub given a matchmaker that starts with an "in-between" Elo target.

Don't get me wrong, the original quote (the one I quoted above) was confusing too, but it certainly seemed to contradict what Matthew Craig says in the second quote. There's never been a super good explanation for how the matchmaker selects and sorts players. Which is why I suggest that we need a dev to, once and for all, give us a straight answer written in clear, well-worded sentences

Your quote was from February 2013. Matthew's quote was from August 2013.

Personally, I think the more recent one is more applicable than the one 5 or so months prior to it, but that is just me.

Edited by Asakara, 23 October 2013 - 12:01 PM.

[PEEFsmash]

Kunae, on 23 October 2013 - 07:35 AM, said:

You are over-stating this, as others have.

It would not change hardly anything, other than make the chance of a newbie on your team random, rather than by design.

There will not be anyone on the opposing team to balance a top competitive premade. As of the present, they balance eachother. Without that limitation, it will be facerolls every time, and that isn't overstating it. Some players had K/Ds above 30:1 before Elo, and we would go right back to that or worse, because there are even new more new players to beat up on.

[Kunae]

FerretGR, on 23 October 2013 - 11:52 AM, said:

Which is why I suggest that we need a dev to, once and for all, give us a straight answer written in clear, well-worded sentences

Yeah... like that's ever going to happen... You new here?

PEEFsmash, on 23 October 2013 - 12:01 PM, said:

There will not be anyone on the opposing team to balance a top competitive premade. As of the present, they balance eachother. Without that limitation, it will be facerolls every time, and that isn't overstating it. Some players had K/Ds above 30:1 before Elo, and we would go right back to that or worse, because there are even new more new players to beat up on.

That was in 8v8. Back then, a good 4-man on one side could totally dominate.

It's no where near as much of a factor, with 12v12.

[Sarsaparilla Kid]

Part of the problem I think is that noobs are given an ELO of 1300, which is smack in the middle of the bell curve for ELO, and averaging ELO for all on a team to affect an individual's actual ELO just makes that middle-ground such a huge magnet that is hard to escape...like trying to fly out of the gravity well of a black hole!

http://mwomercs.com/...ost__p__2265319

Edited by Sarsaparilla Kid, 23 October 2013 - 12:08 PM.

[Sarsaparilla Kid]

On second look...it appears the ELO of 1300 that a newbie would start out with is slightly higher than the average ELO...probably not a good thing.

[FerretGR]

Asakara, on 23 October 2013 - 12:01 PM, said:

Your quote was from February 2013. Matthew's quote was from August 2013.

Personally, I think the more recent one is more applicable than the one 5 or so months prior to it, but that is just me.

Well, given that they're about the same thing, they shouldn't contradict, should they? The same MM is in place now and in August and in February. And to be fair, there are plenty of situations where the first word on a subject is the official one. Not to mention the first quote coming from an "official" Command Chair post, while the second comes from a reply in a forum thread.

But who cares which one should take precedence, they appear to contradict each other, and they were both posted by Matthew Craig. Doesn't make sense.

Sarsaparilla Kid, on 23 October 2013 - 12:14 PM, said:

On second look...it appears the ELO of 1300 that a newbie would start out with is slightly higher than the average ELO...probably not a good thing.

IMHO, this is the likely cause of experienced players riding with noobies. The starting Elo should be much lower IMHO. it'll right itself quickly enough if the player is good.

Edited by FerretGR, 23 October 2013 - 12:26 PM.

[Feetwet]

PEEFsmash, on 23 October 2013 - 12:01 PM, said:

There will not be anyone on the opposing team to balance a top competitive premade. As of the present, they balance eachother. Without that limitation, it will be facerolls every time, and that isn't overstating it. Some players had K/Ds above 30:1 before Elo, and we would go right back to that or worse, because there are even new more new players to beat up on.

Alot of those ultra high KDRs were also achieved BEFORE the drop limits (or at least began in that period). Before the 8 man queue (4 man drop limit) some of the matches were REALLY lopsided. I pugged back then and it could be brutal.

I think the root of the problem is playerbase. Once the MM starts expanded its criteria to pull in more players. This and the high starting ELO for new guys might be what we are seeing. One way or the other some of the matches are just downright horrible.

S

Edited by Feetwet, 23 October 2013 - 12:46 PM.

[Kunae]

FerretGR, on 23 October 2013 - 12:26 PM, said:

IMHO, this is the likely cause of experienced players riding with noobies. The starting Elo should be much lower IMHO. it'll right itself quickly enough if the player is good.

I suggested that very thing, to Matt Craig, in the beginning of Oct, and he brushed it off, basically saying that was how it was now... which didn't really make any sense.

FerretGR, on 23 October 2013 - 12:26 PM, said:

Well, given that they're about the same thing, they shouldn't contradict, should they? The same MM is in place now and in August and in February. And to be fair, there are plenty of situations where the first word on a subject is the official one. Not to mention the first quote coming from an "official" Command Chair post, while the second comes from a reply in a forum thread.

That was just their position, at that time...