205 replies to this topic

[Targetloc]

FerretGR, on 14 March 2013 - 12:35 PM, said:

5 matches is far too few to draw a conclusion from. You could roll snake-eyes 5 times in a row, but that doesn't mean it's a likely outcome, that's just how probability works. You ended up on the snake-eyes side of the Elo matchmaker 5 games in a row. That's as likely an explanation as any other.

But that's enough to draw at least one conclusion from. That you can roll snake-eyes on the matchmaker. I think what most players were hoping for was more segregation, so that if you're playing in the high range you won't randomly get guys that run around on 3x zoom the entire match and don't know how the R button works.


Now, for Mr Zeh...

My win rate in heavies since they started tracking per-mech stats is 65%. My win rate during seeding and the Hero tournament was 70%.

Elo has performed pretty much exactly as expected. The first couple days after per-mech stats were being recorded I was holding a 75% win rate on my heavies. (Still have that on my K2). It has gradually gone down as matches have become more and more challenging.


Where I think most people are surprised is that the 'more challenging' progression seems to be: "I need to get 3 kills and 500 damage for a win" to "I need to get 4 kills and 650 damage for a win" up to "Damn. Only did 800 damage and 5 kills... not quite enough."

Now I'm a traditional Clanner, so I feel there's no point in blaming your team. If you really wanted that win you'd have killed more enemies. Do a better job of protecting your PUGs, Mr. Top-5%.


Though it is much less frustrating at times to have team-mates you can count on... or who won't all die in the middle of the lake 30 seconds into the match. But you know... if you keep losing you'll find yourself back in matches you can win again.

The only thing that really, really hurts is sometimes having to watch. You know, when there's a Stalker, an Atlas and a Dragon all standing in place, slowly turning around in circles trying to kill that last remaining one-armed Trebuchet... The Stalker has overheated and shut down his last 3 consecutive shots, and the Dragon has shot the Atlas twice. 30 seconds in the Atlas slowly pitches over onto it's back. A horrifically anti-climatic death at the hands of a limping TBT. Finally one of your olive drab lance-mates lands a hit, mercifully ending the match and the echos of Yackity Sax playing in your head.

[FerretGR]

shabowie, on 14 March 2013 - 02:06 PM, said:

FerretGR, Von pilsner just posted something new that appears to prove you wrong.

I don't think so. His post is about how Elo is used to calculate "chance of winning" in match, which is a factor used in further calculation of Elo rating. It doesn't speak to how the matchmaker makes the matches initially.

Targetloc, on 14 March 2013 - 02:38 PM, said:

But that's enough to draw at least one conclusion from. That you can roll snake-eyes on the matchmaker. I think what most players were hoping for was more segregation, so that if you're playing in the high range you won't randomly get guys that run around on 3x zoom the entire match and don't know how the R button works.

I haven't spectated a single player that sounds like you describe since the first tournament. I'm not that good at this game. How can someone with a better Elo than me be meeting up with worse players, given that the Elo system is working in my instance? I mean, maybe there's a major glitch in the matrix and that's what's happening, but I think there might be a simpler explanation.

Edited by FerretGR, 14 March 2013 - 05:30 PM.

[Devil Fox]

FerretGR, on 14 March 2013 - 05:26 PM, said:

I don't think so. His post is about how Elo is used to calculate "chance of winning" in match, which is a factor used in further calculation of Elo rating. It doesn't speak to how the matchmaker makes the matches initially.




I haven't spectated a single player that sounds like you describe since the first tournament. I'm not that good at this game. How can someone with a better Elo than me be meeting up with worse players, given that the Elo system is working in my instance? I mean, maybe there's a major glitch in the matrix and that's what's happening, but I think there might be a simpler explanation.

It depends on how fast your being matched... you might be lucky in that you're of the average ELO where there is plenty of numbers. At present if I 4man I get a much worse experience then soloing because the matchmaker builds around 4man's... so a high ELO 4man will pull in low ELO players to "average" the group ELO. This averaging means you tend to pull 1-2 premades on the opposing team of average ELO score.

Now go solo dropping and I average 30s-1min on finding a match, and am the last one into the game... that means I'm a filler whether high or low ELO we don't know, half the time it's a close match, the other half it's a stomping one side or the other.

Now throw in mismatched weights such that a 4man of medium mechs with 1 or 2 high ELO players, with seasoned pilots playing with a 4 day old player in a sub-par machine... and it throws out 3-4 assault, ecm light's and cheese heavies... not a good experience. Equipment makes alot of difference in fight's, because it's harder to make pilots fall for traps without being obliterated by their firepower first. Still it's doable as my Cent and a Cicada drew over 6 enemy into urban fighting where we piece mealed them, but is it enjoyable to have watch LRM boats zoomed x3 firing what small lasers they have at the light mech running circles around them? No.

[FerretGR]

Yeah, that's what this thread is about. The average thing you're describing is not happening. Read back a couple of pages.

[shabowie]

FerretGR, on 14 March 2013 - 06:18 PM, said:

Yeah, that's what this thread is about. The average thing you're describing is not happening. Read back a couple of pages.

It's either using an average or an aggregate within a certain range, if it doesn't it can't possibly match any team with another team. If You think that isn't the case explain how it could be that way, and use math as proof. What you quoted never has definitively said what you claimed it said, it was actually extremely vague and now newer more complete information intended to clarify things is contradicting what the old statement said.

FerretGR, on 14 March 2013 - 05:26 PM, said:

I haven't spectated a single player that sounds like you describe since the first tournament. I'm not that good at this game. How can someone with a better Elo than me be meeting up with worse players, given that the Elo system is working in my instance? I mean, maybe there's a major glitch in the matrix and that's what's happening, but I think there might be a simpler explanation.

Simple. You say you aren't that good. You are much easier to fit into a team where the average or aggregate skill number is equal to a larger number of teams. If somebody has a really high score, they either raise the average or the aggregate, and require a bad player to balance out the team compared to the other.

We don't know if they are using average or aggregate, we have conflicting and incomplete information.

Edited by shabowie, 14 March 2013 - 06:36 PM.

[Targetloc]

FerretGR, on 14 March 2013 - 05:26 PM, said:

I haven't spectated a single player that sounds like you describe since the first tournament. I'm not that good at this game. How can someone with a better Elo than me be meeting up with worse players, given that the Elo system is working in my instance? I mean, maybe there's a major glitch in the matrix and that's what's happening, but I think there might be a simpler explanation.

I don't know. Maybe it has something to do with the "grows those thresholds over time to ensure a match is made within currently a 2 minute time period."

There is an elaborate scenario I constructed to convince myself I'm super good at robots...

[FerretGR]

shabowie, on 14 March 2013 - 06:31 PM, said:

It's either using an average or an aggregate within a certain range, if it doesn't it can't possibly match any team with another team. If You think that isn't the case explain how it could be that way, and use math as proof. What you quoted never has definitively said what you claimed it said, it was actually extremely vague and now newer more complete information intended to clarify things is contradicting what the old statement said.

Despite the further discussion we've had, I thought we settled that it's clearly not doing what the poster I responded to is saying, given that the Command Chair post did explicitly say that it doesn't. Do you still think that the target number is being changed after players have been matched? Because that's the only way "lower Elo" players would be added to a "higher Elo" group to average the teams.

The devs have stated, as noted above, that average Elo values are used for scoring the "chance to win" within game, so it's perfectly reasonable to posit that they might be used at some point in the matchmaking process, as we discussed above. Simply not in the way that Apostal describes.

I'm not sure why you think the matchmaker can't possibly match up teams without using the averaging of teams described by Apostal. I'm not clear on what math you'd like me to provide, but here's one way it could work, with the understanding that I really don't know anything firm. Let's say a group of 4 drops. Their Elos are averaged and they're stuck in the matchmaker as "one big player", kind of like a weapon that takes up multiple crits... a player that takes up four spots in a game. Ie. an 800, 1000, 1200, and 1400 Elo group would look like an 1100 Elo player from the perspective of the matchmaker. As we've taked about before, I don't know how the target number is established, but let's say it's established when the first player enters the room. This group would enter, setting the target number at 1100. Then, a second 4-man clicks launch, consisting of a 900, 1100, 1300, and 1500 Elo player, for an average of 1200. This "big player", from the perspective of the matchmaker, is within range of the target number, so it gets added.

Of course, I have no inside knowledge, but it's plausible that it could work this way, no? It could possibly match team vs. team without violating what we've been told about Elo in this game so far.

Edited by FerretGR, 14 March 2013 - 06:47 PM.

[FerretGR]

Targetloc, on 14 March 2013 - 06:34 PM, said:

I don't know. Maybe it has something to do with the "grows those thresholds over time to ensure a match is made within currently a 2 minute time period."

Indeed; I'm just thinking that my threshold would necessarily include anyone who a person with a higher Elo than mine encounters. If I'm middle of the pack, my threshold should dip down further than someone at the top of the bell curve, shouldn't it?

[shabowie]

FerretGR, on 14 March 2013 - 06:45 PM, said:

Despite the further discussion we've had, I thought we settled that it's clearly not doing what the poster I responded to is saying, given that the Command Chair post did explicitly say that it doesn't.

It does not explicitly say that. In fact I said from the beginning I thought it was a team aggregate number being looked at, not an average. I also said I don't believe it says what you think it says. It says a target value is used but doesn't say what that target value is or where it comes from, then it uses a very vague hypothetical which isn't really an explanation but is more of an analogy.

FerretGR, on 14 March 2013 - 06:45 PM, said:

Do you still think that the target number is being changed after players have been matched? Because that's the only way "lower Elo" players would be added to a "higher Elo" group to average the teams.

I think poor players are being teamed with 4 man groups of good players to balance things out on a pretty regular basis. How or in what order is hard to say. I am pretty sure either an average or aggregate rating is used for matchmaking purposes though. The programming has to involve math at the team level, not individual. There's no other way to explain the lopsided skill levels seen so consistently by so many people.

Edited by shabowie, 14 March 2013 - 07:09 PM.

[FerretGR]

shabowie, on 14 March 2013 - 06:51 PM, said:

It does not explicitly say that. In fact I said from the beginning I thought it was a team aggregate number being looked at, not an average.

We'll have to agree to disagree, because from my perspective, if the devs are asked, straight out, "is it an average or is it something else," and they say "it's something else," that means "it's not an average," and again, though it might not be precisely what you're describing, it is what Apostal is describing. TBH, though, I think you're just being disagreeable, because I've agreed with you that averages/aggregates are certainly used at some point.

shabowie, on 14 March 2013 - 06:51 PM, said:

I think poor players are being teamed with 4 man groups of good players to balance things out on a pretty regular basis.

I think if folks are consistently seeing poor players in their matches, they should look at themselves as closely as they're looking at the matchmaker.

Edited by FerretGR, 14 March 2013 - 07:06 PM.

[BFett]

http://mwomercs.com/...49#entry1985349

Vote!

[Targetloc]

FerretGR, on 14 March 2013 - 06:50 PM, said:

Indeed; I'm just thinking that my threshold would necessarily include anyone who a person with a higher Elo than mine encounters. If I'm middle of the pack, my threshold should dip down further than someone at the top of the bell curve, shouldn't it?

Middle of the pack is also the fattest part of the curve. Maybe it's easier for the MM to find matches so it's much, much rarer that it runs into the 'okay, gotta find something, anything goes' mode? Depends how they've prioritized "who's been waiting the longest" versus "who would be the closest match?"

[FerretGR]

You might be right... the matchmaker could be panicking and just picking whoever clicked launch last for the folks at the peak of the bell curve, but it would seem that they'd be, again, based on the fact that there are a lot of players in the fat part of the curve, someone relatively close to me rather than someone at the bottom end of the bell curve, on average, at least.

Oh god, maybe I'm one of the scrubs everyone's talking about

Edited by FerretGR, 14 March 2013 - 07:10 PM.

[shabowie]

FerretGR, on 14 March 2013 - 07:03 PM, said:

We'll have to agree to disagree, because from my perspective, if the devs are asked, straight out, "is it an average or is it something else," and they say "it's something else," that means "it's not an average,"

Average and aggregate are not the same thing.

FerretGR, on 14 March 2013 - 07:03 PM, said:

I think if folks are consistently seeing poor players in their matches, they should look at themselves as closely as they're looking at the matchmaker.

This thinking on your part stems from a false belief that there are levels that people are being segregated into based on an individual score.

What's really happening is a bunch of individual scores are being thrown into a pool and then the matchmaker is trying to create two teams with either an average or aggregate score within a certain range, then expanding those ranges if it can't find them.

Edited by shabowie, 14 March 2013 - 07:17 PM.

[FerretGR]

shabowie, on 14 March 2013 - 07:11 PM, said:

Average and aggregate are not the same thing.

Well, average can be a part of the aggregate, to be fair, since aggregates are by definition collections of statistical data. But you edited out the key part of that sentence: "though it might not be precisely what you're describing, it is what Apostal is describing" (ie. though the aggregate you're describing might consist of something more, Apostal is talking about average Elo). I might slip between average and aggregate in my posts because I'm lazy, not trying to obfuscate. You're in hardcore "prove Ferret wrong" mode, though! I'm not trying to pick a fight, we just disagree.

Edited by FerretGR, 14 March 2013 - 07:16 PM.

[Targetloc]

FerretGR, on 14 March 2013 - 07:09 PM, said:

You might be right... the matchmaker could be panicking and just picking whoever clicked launch last for the folks at the peak of the bell curve, but it would seem that they'd be, again, based on the fact that there are a lot of players in the fat part of the curve, someone relatively close to me rather than someone at the bottom end of the bell curve, on average, at least.

Oh god, maybe I'm one of the scrubs everyone's talking about

Even if it is just picking randomly from the middle at a certain point, all new players start in the middle, so there's a statistically high chance you will pick up a bunch of new players.

But we're all just guessing.

You could be right... maybe we have met the noobs, and they are us.

[FerretGR]

Targetloc, on 14 March 2013 - 07:17 PM, said:

Favorite new MWO meme

[Dracol]

With Elo, one team is set up to be statistically the most likely team to lose. Maybe you've just had a string of bad luck and was on the disadvantaged team. I was stuck on River City 3 out of my first 4 matches tonight. Streaks happen.

[shabowie]

FerretGR, on 14 March 2013 - 07:14 PM, said:

But you edited out the key part of that sentence

Because what somebody else says isn't the crux of the disagreement. So I focused on other things. Also there's no tone on the Internet, so to make it clear, I'm not mad, or trying to pick a fight.

Edited by shabowie, 14 March 2013 - 07:23 PM.

[FerretGR]

shabowie, on 14 March 2013 - 07:22 PM, said:

Because what somebody else says isn't the crux of the disagreement.

It is when the quote you disagreed with was directed at the other guy. Can we stop this now?