202 replies to this topic

[Joseph Mallan]

cleghorn6, on 08 April 2014 - 06:04 AM, said:

Yeah, that's the only benefit I can see. The mythical "premade stomp" would be a thing of the past. Like it was when they stopped limited 8s dropping in the PUG queue.

Oh, wait ...

Thats better...

[Bhael Fire]

StaIker, on 07 April 2014 - 11:10 PM, said:

1 Great player cannot move a 50/50 result up to a 90/10. That doesn't mean a great player has no influence at all however. It would not be unreasonable to expect a top 1% player to be winning PUGs at 65% or better, even though he is placed with 11 other random folks. For people in the middle of the skill range, which is most folks who play regularly, they are not going to move the needle that much, perhaps just a few points up or down from 50%. That's because there is a huge gap in capability between the "average" player and a player in the top 1%. That guy is worth 2 or 3 or 4 "average" team mates.

The real issue is that an average solo player (like myself) can get caught in an Elo funnel where they get artificially pulled up or down into an Elo bracket they don't belong in because of a string of lucky (or unlucky) matches...and get stuck there for an inordinate amount of time because of the actions of the 11 other random players. I was pointing out the fallacy that a player's Elo score will eventually settle into its rightful place, when in fact it can (and does) fluctuate wildly on a regular basis ...especially if you play solo a lot.

For example my current w/l ratio is about 1.3...and it's rising. I can assure you, this has nothing to do with my skill as a player. And as such, my matches lately have been brutal...I'm completely out of my element.

Edited by Bhael Fire, 08 April 2014 - 07:58 AM.

[JudgeDeathCZ]

I feel your pain OP.
I played ~10k games since they reseted ELO so no1 can say my ELO is not where it should be.
Problem is it counts average TEAM ELO and not individual player ELO...

[Roadkill]

Roland, on 07 April 2014 - 04:05 PM, said:

As I pointed out, you don't really have any actual math to back up your claims here.

That's because I don't care enough to prove Abivard wrong again. He refuses to acknowledge facts when they're presented and simply keeps re-stating his erroroneous claim over and over.

There was another very, very long thread in which multiple people presented multiple research papers that all concluded the same thing - Elo works fine in a multiplayer environment. Abivard ignored all of them, even going so far at one point as to claim that he's smarter than a PhD from MIT.

I've implemented Elo before in a multiplayer environment. I've run simulations to determine whether or not it works. It does.

Do other systems work better? Some yes, some no. But it's a relative thing - Elo works just fine, but another system might work a little bit better.

Abivard claims that Elo doesn't work in a multiplayer environment. He's wrong. But he'll twist words and change his claim constantly to avoid having to admit that he's wrong. It's just not worth arguing again since the previous 30+ pages of evidence did no good.

[wintersborn]

So you don't like PUG Stomping because its either too easy or too hard?

[Joseph Mallan]

wintersborn, on 08 April 2014 - 10:39 AM, said:

So you don't like PUG Stomping because its either too easy or too hard?

I'm an equal opportunity stomper... PUGs are no better or worse than groups.

[Abivard]

Roadkill, on 08 April 2014 - 10:34 AM, said:

That's because I don't care enough to prove Abivard wrong again. He refuses to acknowledge facts when they're presented and simply keeps re-stating his erroroneous claim over and over.

There was another very, very long thread in which multiple people presented multiple research papers that all concluded the same thing - Elo works fine in a multiplayer environment. Abivard ignored all of them, even going so far at one point as to claim that he's smarter than a PhD from MIT.

I've implemented Elo before in a multiplayer environment. I've run simulations to determine whether or not it works. It does.

Do other systems work better? Some yes, some no. But it's a relative thing - Elo works just fine, but another system might work a little bit better.

Abivard claims that Elo doesn't work in a multiplayer environment. He's wrong. But he'll twist words and change his claim constantly to avoid having to admit that he's wrong. It's just not worth arguing again since the previous 30+ pages of evidence did no good.

Roadkill just insists he is right, and he has 3 or 4 friends that will tell you he is right. But they will not ever provide any real evidence that they are right, If you show him and his cronies facts, they will ignore them, if you show them the math they will ignore that as well. If you link them articles they will skim for keywords to support their case, while failing to comprehend the article refutes his position. I think it must be because the article sailed far above roadkill's head.

Roadkill must use straw arguments and hyperbole, along with the forum version of 'shouting down others' in order to try and make his case.

He will not provide math or facts because he has none, what he provided before was an epic reading comprehension failure on his part. He knows better than to link to articles that refute his position, as He seems unable to comprehend even the most basics of science and math It seems he has decided to not continue to gamble on citing references that he doesn't understand. Once burnt twice shy so the saying goes.

I guess Roadkill feel's I am the only person who has this position, he is wrong but that is no surprise.

As Roadkill has said, everything is working fine with Elo, the MatchMaker is working fine basing itself off of PGI's Elo scores, The vast majority of matches are well balanced, it is only the rare mismatch that occurs.

Roadkill has also told us that only 200 to 300 games are needed to correctly seat you in your Elo rating. He just KNOWS in his inner soul that this is the true number of games.

He will explain how wrong you are and how Everyone just KNOWS he is right!

Roadkill will tell you Elo is working as intended in multiplayer games, all those people working on other rating systems are idiots, all they can do is make it work 'faster'.

Roadkill will tell you many things, but he can never seem to actually back them up.

[Koniving]

Simple idea. Two separate queues. Solo queue, with other solo players populating it.
Premade queue, with groups of various sizes from 2 to 12 fill the ranks. Don't got enough people in your premade 'cause it only has 3 people, well we'll hook you up with this premade which has 7 or this one with 5 and a give you a couple of 2 man commando teams.

Bam. Done.

[Roland]

Roadkill, on 08 April 2014 - 10:34 AM, said:

There was another very, very long thread in which multiple people presented multiple research papers that all concluded the same thing - Elo works fine in a multiplayer environment. Abivard ignored all of them, even going so far at one point as to claim that he's smarter than a PhD from MIT.

No, that's not really what those papers proved.

What they demonstrated was that in a particular environment, Elo converged upon a rating which reasonably predicted win percentages, eventually.

That's not even remotely the same thing as you are claiming. You are misinterpreting the data.

I've already explained exactly where your misinterpretation lies, but you don't seem to understand the criticism to the degree necessary to address it. You don't have mathematical proof of the statements you are making. You're just handwaving and using pseudo-math to try and make your position sound legitimate.

Quote

I've implemented Elo before in a multiplayer environment. I've run simulations to determine whether or not it works. It does.

This statement is meaningless, as there each game introduces new variables which will affect the convergence rate. Your experience in other "multiplayer environments" isn't applicable to this one.

You have no idea what convergence rate you're going to get in MWO with Elo. You would need to actually look at real games, take the results, and then compare the Elo predictions to the actual game results to establish the convergence rate.

[ImperialKnight]

ELO is bad in MWO and results in weird situations. You CANNOT just take W/L into consideration, It needs to take a lot more variables into consideration to determine how good the player is.

My light mech class is in what people call ELO hell. My RVN has 0.5 W/L, but >3 K/D. I constantly do top damage and top kills on my losing team where half the team fails to break even 100 dmg. But I will never move up the ladder. So my W/L keeps going down while my K/D keeps going up. It may be fun to keep picking off new players who don't know what they're doing but it gets boring after a while.

[Grits N Gravy]

It's not that Elo itself is not a good ranking system, it's just implemented so poorly in MWO that it's effectively worthless. An Elo ranking system is a mathematical way of saying how likely you are to beat an opponent. IE if your Elo score is 100 points higher than your opponent you will win that match up 64% of the time. Using team average, the wide spreads and a host of other factors,( covered in numerous other threads) reduces the accuracy of Elo predictions and therefore effectiveness of matching to nil.


In effect, the MWO matchmaking system only prevents the absolute worst players from facing the absolute best. Other than that there is no real functioning skill based matchmaking.

[Mawai]

It is standard procedure to blame the match maker ... and maybe the matchmaker is the problem some of the time ... but ...

1) I get better results in terms of W/L even with the most minimal of comments or suggestions made at the beginning of the match. If you are in games where no one says anything remotely useful then don't be surprised you lose. If someone on the other team is providing some cohesion their odds of winning went up a lot.

2) You can usually tell within the first minute of the match whether you are likely to have a chance of winning or not.
- if your team splits into 2 groups you will likely lose unless the other team did the same thing
- if 2 or 3 mechs run off on their own (even lights ... unless just scouting) ... they will probably die and again you will lose
- if your team gets bottled up at a ridge, choke point or building corner taking turns peeking to fire and being hit by several opponents each time then again you will likely lose
- if a group figures that it doesn't need the rest of the team to win ... it will often lose

There are so many losing behaviours which I have seen even good players fall into ...

So ... if you see one of these happening ... try making a suggestion in chat instead of blaming the matchmaker after the fact.


Finally, the 12:0, 12:1, 12:2 results are common because once one side gets an advantage they can easily convert the momentum into a steamroll. Six mechs on the opposing team can easily roll up a line of opposition taking them out one at a time so quickly that unless the defenders are paying attention ... which they often aren't without VOIP ... the opponents will wipe them out.

Look at the games you have played ... look at what happens ... look at the common elements ... then do something to counteract them and your W/L goes up. Most W or L are not the fault of the matchmaker - they are the fault of the players.

(Note: the matchmaker does setup uneven matches ... tonnage, maybe ELO ... but there is no convincing evidence that the quality of the match isn't due to the tactical and strategic decisions of the players rather than the matchmaker ... though I agree that actual new players should have a separate queue for the first 25 games or so ... )

[Dan Nashe]

If matchmaker doesn't measure skill, then why the complaint? It's no worse than a random matchmaker.

Further query: how is a ladder going to work better than matchmaker? Won't that still have the same "23 other people" issue without being weighted at all like Elo in terms of movement being dependent on relative Elo? I don't understand how a ladder system would work. Elo I understand.

[Grits N Gravy]

Mawai, on 08 April 2014 - 03:54 PM, said:

(Note: the matchmaker does setup uneven matches ... tonnage, maybe ELO ... but there is no convincing evidence that the quality of the match isn't due to the tactical and strategic decisions of the players rather than the matchmaker ... though I agree that actual new players should have a separate queue for the first 25 games or so ... )

There is also no evidence that quality of matches isn't anything but the result of the law of large numbers. As it stands now the difference in team average Elo scores is 175 points. Therefore at any given drop you have a 50% chance of being on the team that is almost guaranteed to lose or being on the team that is guaranteed to win. So it is easy to tell which team you are on right away, the real question is why did u end up on that team.

The whole concept of a matchmaker is to provide a gauge for how likely someone is to engage in either losing or winning behavior. People often can't unbiasedly asses their own skill or behavior. It's also true that players behavior often doesn't change; no matter what advice is provided to them, a.k.a who reads tool tips. Thus the concept of a matchmaker is to identify and place players in a environment where players are just as likely to engaging in the same "losing behavior". In a perfect world they would have a 50/50 chance of winning every game.
Which is not the same as having a 50% win ratio.

It's actually relatively easy to track the effectiveness of a system. All you have to do is look at prediction error, and a few other metrics.

Source for average difference in Elo scores is 175: http://mwomercs.com/...ate-jan-212014/

[AccessTime]

I don't believe in Elo at all, to me it's just a solution in search of a problem.

[Hellen Wheels]

Fools.

There is no "Elo" in this game.

There is not "MM" in this game.

Dropping into a game now is, for all practical purposes, entirely random.

Enjoy.

[Ordellus]

knightsljx, on 08 April 2014 - 03:21 PM, said:

ELO is bad in MWO and results in weird situations. You CANNOT just take W/L into consideration, It needs to take a lot more variables into consideration to determine how good the player is.

My light mech class is in what people call ELO hell. My RVN has 0.5 W/L, but >3 K/D. I constantly do top damage and top kills on my losing team where half the team fails to break even 100 dmg. But I will never move up the ladder. So my W/L keeps going down while my K/D keeps going up. It may be fun to keep picking off new players who don't know what they're doing but it gets boring after a while.

Maybe stop using your light mech to clean up after the other 11 people do all the work.

Hellen Wheels, on 08 April 2014 - 06:31 PM, said:

Fools.

There is no "Elo" in this game.

There is not "MM" in this game.

Dropping into a game now is, for all practical purposes, entirely random.

Enjoy.

Link?

[Roadkill]

Roland, on 08 April 2014 - 03:11 PM, said:

No, that's not really what those papers proved.

What they demonstrated was that in a particular environment, Elo converged upon a rating which reasonably predicted win percentages, eventually.

That's not even remotely the same thing as you are claiming. You are misinterpreting the data.

Huh? That's exactly what I'm claiming. I'm claiming that Elo works just fine in a multiplayer environment. Abivard claims that it doesn't. That's the point. Those papers did, in fact, back up my claim that Elo works just fine in a multiplayer environment, yet Abivard refuses to acknowledge that fact.

Quote

I've already explained exactly where your misinterpretation lies, but you don't seem to understand the criticism to the degree necessary to address it. You don't have mathematical proof of the statements you are making. You're just handwaving and using pseudo-math to try and make your position sound legitimate.

I understand that you're trying to derail the central point here by diverting down a side issue that isn't relevant. Furthermore, I understand that you also have no mathematical proof of your claim in that side issue, either. In my case I know that the proof exists but don't care to look it up again. In your case I know it doesn't exist because I know that what you're claiming isn't true. (Technically your phrasing is very carefully true, but irrelevant. Yes it is possible that it might take millions of games, but ridiculous.)

So shall we just ignore the side issue, then?

[Roadkill]

Abivard, on 08 April 2014 - 03:01 PM, said:

As Roadkill has said, everything is working fine with Elo, the MatchMaker is working fine basing itself off of PGI's Elo scores, The vast majority of matches are well balanced, it is only the rare mismatch that occurs.

As usual, you are wrong.

I've never said that the matchmaker is working fine.

Elo works just fine in this environment. How the matchmaker uses Elo ratings to create games may or may not be working fine depending on your point of view. If you want to argue that matches are AFU, that's an entirely different issue and I'm not going to argue one way or the other.

But your continued, unsubstantiated claim that Elo doesn't work in a multiplayer environment is FALSE. You conflate Elo with the matchmaker. They are not the same thing.

And you have never offered any kind of proof, link or otherwise, that Elo doesn't work in a multiplayer environment. In fact every link you have offered proves that Elo does work in a multiplayer environment, but that some other system might work better.

Our difference is that the people who agree with me have also posted proof that Elo works in a multiplayer environment. Neither you nor any of the people who have agreed with you have offered any proof to the contrary. Want to know why? Because it doesn't exist.

But go ahead, keep repeating your falsehood. Keep putting words in my mouth and then claiming that those words I never said aren't true. It just makes you look even sillier every time you do it.

[stjobe]

It's not that hard to type "multiplayer elo" into google and at least skim the first few hits.