117 replies to this topic

[Deathlike]

OK... TBH, there are issues are seriously magnified with the MM when it comes to lopsidedness, especially in a 12v12.

Newbie ELOs initially start at 1300, but use 1100 for the duration of their cadet bonus. "Elite" ELOs seem to be at 1600 or so (I could be wrong, but IIRC that was the scale shown in an aging Command Chair post.

http://mwomercs.com/...ted-april-19th/

If we look at the distribution, "average" ELO looks like less than 1300... it's closer to 1250 or so... and for the sake of the discussion, we'll just use the graph AS IS since we have no actual references to the current #s for ELO and the game.

Let's say for the sake of argument that MM generally picks around 1250 to 1350 ELO for each team, where the difference between ELOs is at most 100.

I'll just spit out some #s for the sake of discussion:

Team 1:
900
1000
600
500
1800
1600
1500
2000

Avg ELO: 1237.5

Team 2:
1150
1200
1250
1300
1200
1000
1400
1150

Avg ELO: 1206.25

According to the MM, the Team #1 is supposed to win. Team #1 kinda assumes at least a premade who has played together and often... and Team #2 may have a premade (but not an elite group, just a bunch of friends), and generally a lot of average PUGs.

Unless the Premade IS as good as their ELO numbers would indicate (remember, its the ELO with respect to the MECH WEIGHT CLASS that they are fielding), they literally have to carry the rest of the weaker ELO teammates... at least 1 more than each player on the field. It is in my opinion that the players that carry the "average" ELO would do better... since they very likely solo PUG and took their share of failure for a longer period of time during their time as newbies... and work their way up.

This discrepancy doesn't quite get made up in 12v12.. where if the the elite players used to use the 4 PUG players as meatshields.. they cannot do so anymore.. they'll probably be a greater discrepancy if two 3-man teams with high ELO are on one team.. or many solo PUGs that have been in premades at one time or another are playing regular matches... where losing 6 "terribad" players has a far greater consequence in the outcome of the match.

The basic idea of "balancing" high ELO players with low ELO players can be a serious problem PARTICULARLY if the RANGE (a basic statistics term) is vastly different between two teams. The gaps themselves do not help... so having an elite player/group will tend to generate a bad player/group and as more of the discrepancies increase, the more lopsided the match is more likely to be. It wasn't as bad in 8v8, but it now is much worse in 12v12 because of this behavior.

I'm sure this is not a perfect nor accurate representation, but this can give you can idea who is affected more by the MM and WHY a tutorial would be benefit a newbie. The gaps in skill make MM a nightmare... and unless the MM gives "equal #s of bad ELO players" on each side, the system will penalize you more when you have a smaller premade AND/or if you solo PUG with a high ELO.

Also, remember that the initial newbie ELO is in play as well, so actual newbies tend to hurt the system in which this was meant to be balanced against (initial ELO is 1300 internally, newbie ELO is 1100).... their ELO should start FAR lower than they are currently if we go by PGI's last known distribution.

Edited by Deathlike, 10 August 2013 - 04:13 PM.

[DeaconW]

Hauser, on 10 August 2013 - 03:38 PM, said:

Those are just more permutations.

Permutations the ELO system was never designed for.

Quote

There are many variables but Elo only looks at three. The outcome of the match, your teams Elo and your opponents Elo. The Elo ratings are under control of the algorithm so the only free variable is the match out come. Anything you do that influences the outcome will be reflected in your Elo rating.

But it's not only you that influences the outcome. This is the crux of the problem. That and the fact that the MM will throw ELO out the window if it takes too long, further skewing and ELO determinations from that match.

[Deathlike]

DeaconW, on 10 August 2013 - 04:12 PM, said:

But it's not only you that influences the outcome. This is the crux of the problem. That and the fact that the MM will throw ELO out the window if it takes too long, further skewing and ELO determinations from that match.

I don't think I'm a great player by any stretch... but when I'm the last one into the match (or at least one of the last ones), I'm almost confident that I would lose the match outright, regardless of my mech choice.

[Livebait]

12 V 12 has made wining for me next to impossible. Can't say why. My W/L ratio was about 50/50 but know it is mega lose. In 2 days of playing a won maybe 10 games but scoring has not changed. 300-700 damage 1 or 2 kills. Strange but thats MWO.

And just for the record, I'm in four man drops mostly. PUGing a little here and there.

Edited by Livebait, 10 August 2013 - 04:33 PM.

[Hauser]

DeaconW, on 10 August 2013 - 04:12 PM, said:

Permutations the ELO system was never designed for.

But it's not only you that influences the outcome. This is the crux of the problem. That and the fact that the MM will throw ELO out the window if it takes too long, further skewing and ELO determinations from that match.

Elo doesn't care. It is statistically speaking just noise on your ability to win matches.

[DeaconW]

Hauser, on 10 August 2013 - 04:34 PM, said:

Elo doesn't care. It is statistically speaking just noise on your ability to win matches.

What you call "noise", I call "relevant".

[Hauser]

DeaconW, on 10 August 2013 - 04:37 PM, said:

What you call "noise", I call "relevant".

I don't think we're using the same concept of noise.

On the long run the maps are a constant influence on your performance. The individual influence of each map on your ability to win averages out to some constant in the long run. The teams composition is variable but on the long run their influence also averages out to another constant that influences your performance.

The individual maps and teams are just noise on those constant influences. Your Elo will go up and down because of it but it will average out to a certain constant (assuming you don't get better). The influence of players and teams is only relevant when his Elo deviates far from its average (e.g. lots of noise, little signal).

Edited by Hauser, 10 August 2013 - 05:04 PM.

[DeaconW]

Hauser, on 10 August 2013 - 05:01 PM, said:

I don't think we're using the same concept of noise.

On the long run the maps are a constant influence on your performance. The individual influence of each map on your ability to win averages out to some constant in the long run. The teams composition is variable but on the long run their influence also averages out to another constant that influences your performance.

The individual maps and teams are just noise on those constant influences. Your Elo will go up and down because of it but it will average out to a certain constant (assuming you don't get better). The influence of players and teams is only relevant when his Elo deviates far from its average (e.g. lots of noise, little signal).

By that logic, *nothing* should matter because it all "evens out in the end", right? So why even have ELO. At some point that kind of logic doesn't make practical sense.

Edit: perhaps a better way to say it this is...the noise signal is too large all the time so that it affects the signal.

Edited by DeaconW, 10 August 2013 - 05:12 PM.

[scJazz]

1 post... 1 screen shot... lots of words...

why is this a new thread? not subsumed into one of the approximately 10000 other threads on the same subject?

why have roughly 100 people posted in this thread which contains no data of statistically significant relevance?

Exactly how many times a week must we do this stupid dance?

[Hauser]

DeaconW, on 10 August 2013 - 05:06 PM, said:

By that logic, *nothing* should matter because it all "evens out in the end", right? So why even have ELO. At some point that kind of logic doesn't make practical sense.

Edit: perhaps a better way to say it this is...the noise signal is too large all the time so that it affects the signal.

No, it evens out to a constant. Including your skill. That constant, how good you are at winning games, is what your Elo rating reflects.

And signal to noise would be easy to check. You'd just look at the Elo progression of a high Elo player. Or even simpler, start a new account and you'll see what it like in the bottom of the barrel. I tried a while ago. Difficulty wise it barely is a step up from the training grounds so the match maker is working.

Edited by Hauser, 10 August 2013 - 05:29 PM.

[Deathlike]

scJazz, on 10 August 2013 - 05:11 PM, said:

1 post... 1 screen shot... lots of words...

why is this a new thread? not subsumed into one of the approximately 10000 other threads on the same subject?

why have roughly 100 people posted in this thread which contains no data of statistically significant relevance?

Exactly how many times a week must we do this stupid dance?

Well, MM isn't getting much better.

But, I think everyone likes to play "TheoryWarrior" and "GuessWhatPaulIsThinkingWarrior".

ForumWarrior won't evolve... it keeps replicating.

[DeaconW]

Hauser, on 10 August 2013 - 05:19 PM, said:

No, it evens out to a constant. Including your skill. That constant, how good you are at winning games, is what your Elo rating reflects.

And signal to noise would be easy to check. You'd just look at the Elo progression of a high Elo player. Or even simpler, start a new account and you'll see what it like in the bottom of the barrel. I tried a while ago. Difficulty wise it barely is a step up from the training grounds so the match maker is working.

You keep focusing on the "progression of ELO". I don't care if ELO is working as they intended it...it sucks because it doesn't truly balance properly because too many matches are too lopsided. THAT is the part that I think is not OK and that ELO is not addressing properly.

[Screech]

scJazz, on 10 August 2013 - 05:11 PM, said:

why is this a new thread? not subsumed into one of the approximately 10000 other threads on the same subject?

Subsumed you say? Well played sir, well played.

[Corvus Antaka]

i dunno. pugging is it's own challenge. after spending a ton of time pugging the game since last year, yes it can be frustrating, but when i play with lancemates or organized 12 mans it's like easy mode compared

[Training Instructor]

Had another 700+ damage losing match today.

I think that the math isn't at fault, but maybe it's simply a lack of a large player base that is the problem. I've done better in 12 mans than 8 mans, because the opposing team having a good four man is diluted somewhat by the bigger numbers. I'm still carrying the team, win or lose, in far too many matches. My ELO shouldn't be very high, because I'm not an elite player at all. So how do I keep getting put on teams where so many of the players are new, or simply terrible?

[Hauser]

Deathlike, on 10 August 2013 - 05:21 PM, said:

But, I think everyone likes to play "TheoryWarrior" and "GuessWhatPaulIsThinkingWarrior".

Most people I talk with about the matchmaker seem to misunderstand how it works. It is important they do so when they complain they are complaining about the right thing.

See below:

Deathlike, on 10 August 2013 - 04:11 PM, said:

The basic idea of "balancing" high ELO players with low ELO players can be a serious problem

I'm afraid you are mistaken. The match maker does not do that intentionally. Though people often complain about this, so that does suggest the match maker is putting in new players with veterans.

Now it seems that there are not enough similar players in the queue to make a match. So the match maker has to be less picky about who it puts into a match.

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

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.

We need to do this carefully over time as generally the cost of precision is time to find a match we want to slowly find a very nice balance between time to find a match and the number of matches that are correctly composed.

Training Instructor, on 11 August 2013 - 01:41 AM, said:

I think that the math isn't at fault, but maybe it's simply a lack of a large player base that is the problem.

This, or perhaps the time to compose a match is too small to find enough players.

DeaconW, on 10 August 2013 - 05:26 PM, said:

You keep focusing on the "progression of ELO". I don't care if ELO is working as they intended it...it sucks because it doesn't truly balance properly because too many matches are too lopsided. THAT is the part that I think is not OK and that ELO is not addressing properly.

I don't think match outcome is representative for the balance of two teams. I've seen organized teams roll each other first one way and then the other. But otherwise I agree the match maker is making matches were players are too far apart skill wise. We settled that pages ago.

Edited by Hauser, 11 August 2013 - 02:51 AM.

[ztac]

Any ranking system based on wins/losses in a team based game is doomed to fail .

The reason is that any win/loss is the result of a teams effort, not an individual. They need to start basing the elo system on player performance.

The next problem with matchmaking is actual mech balance , its not just the weight that comes into play though as the mechs are customisable. Therefore you would need to at the very least take a mechs actual firepower into consideration, as well as it's potential for Alpha strikes.

LRM count may also be an issue in 12 vs 12 games. As on certain maps where you can be out in the open, or have to advance in the open the number of LRM's can be very overpowering, at the very least they should lose the effect of any mech being hit by LRM's being effectively blind. But then it's not just the LRM's is it? (This problem could simply be down to the sheer number of LRM's in a game, and their ability to target people no matter what their position is. Direct fire people need to actually see and be in a position to hit the target, with LRM's they just need to see a target lit up by someone else)

As it would appear that PGI once again in their infinite wisdom (or lack of), have been doing the rounds on buffing and nerfing.

And with reference to Lanchesters square law:- it would only be true assuming that all mechs are created equally, which I can assure you they are not. As firepower varies with all mechs as well as the time that they can lay down said firepower and the actual player skill of being able to hit the enemy. And as the matchmaker does not take any of this into consideration......... We get very poor matches!

[Hekalite]

I've skimmed through this thread since my last post and I find much of the discussion about as frustrating as the ELO hole I've been stuck in. If you don't believe there is such a thing it's just because you have not been unlucky enough to fall into it. Period. End of discussion. You can see from several of the posters here that their experience obviously differs from yours and the system is a complete black box. If you still refuse to even entertain the possibility that you can be stuck in a viscous losing streak then you are drinking too much coolaid. I'm talking 100+ matches where your w/l is well below 50/50. If I had somehow gone above my natural skill level surely it shouldn't take that long to equalize. If it does, that needs to change asap.

Lastly, do you know how I know this is real? Immediately after my last post yesterday I started to have reasonable results again in my medium mechs. No more hole. The matches are night and day different from what I was seeing before, even when I lose. And although I have seen some stomps they are once again rare, as they should be. Something in the system is broken. Be thankful if you have not experienced it.

[Hauser]

ztac, on 11 August 2013 - 03:06 AM, said:

Any ranking system based on wins/losses in a team based game is doomed to fail .

The reason is that any win/loss is the result of a teams effort, not an individual. They need to start basing the elo system on player performance.

Nope. You're wrong.

Simple fact is that if you have a random match maker like we used to have you'd get people with great win loss ratios like say 10.0 and people with terrible ratios like say 0.1. So clearly some people are better at winning random matches then others. In other words your individual performance is significant when it comes to winning matches.

Hekalite, on 11 August 2013 - 05:50 AM, said:

Lastly, do you know how I know this is real? Immediately after my last post yesterday I started to have reasonable results again in my medium mechs. No more hole. The matches are night and day different from what I was seeing before, even when I lose. And although I have seen some stomps they are once again rare, as they should be. Something in the system is broken. Be thankful if you have not experienced it.

It's weekend. More people are online. More people means the match maker can make better matches. Same thing happens when playing in the European morning compared to playing in the American evening.

Edited by Hauser, 11 August 2013 - 06:05 AM.

[Hekalite]

Hauser, on 11 August 2013 - 05:57 AM, said:

It's weekend. More people are online. More people means the match maker can make better matches. Same thing happens when playing in the European morning compared to playing in the American evening.

I played most of my losing streak during the same time frame last weekend. Broken. Try again.