202 replies to this topic

[Grits N Gravy]

Flaming oblivion, on 10 April 2014 - 09:56 AM, said:

Your not telling me the 3 guys at the end of every match that have done 400-1k damage are on the same skill level as the 3 guys that did 0-150 (on the winning side). wins cant be the decider of who's fit to face who

It's a major problem with the approach used in MWO. Team based games by their nature require orders of magnitude more games played to determine an accurate ranking of an individual. Using the team average Elo scores to compute changes in individual Elo scores, makes the situation worse. As I said earlier MWO matchmaking only functions to separate the absolute worst players from the best, and it doesn't do that all the time either.

Wins are used because they often the best predictor of future outcomes. Though it's true that matchmaking needs to be much more sophisticated to deal with a team based environment. I doubt we will get anything beyond 3/3/3/3 as that is clearly seen as the band-aid fix by the development team.

[Boris The Spider]

Flaming oblivion, on 10 April 2014 - 09:56 AM, said:

You cant be saying the matchmaking system pits people against each other based off of wins, As I've said in other threads i'm new but even I know this would be a bad idea just looking at the score sheet at the end of the match , Your not telling me the 3 guys at the end of every match that have done 400-1k damage are on the same skill level as the 3 guys that did 0-150 (on the winning side). wins cant be the decider of who's fit to face who. It may work if battles were smaller because you'd have to do your bit to compete but not in this madness that is 12v12.

Hypothetically, who did more to ensure the teams victory, the guy who threw his mech into the breach at precisely the right moment to break the enemy line, crippling himself in the process, or the guy who skirted round the edge of the battle to meticulously pick apart a disconnected assault mech? Elo is the measure of your ability to win games, or your teamwork, not just your ability to rack up points.

If you are bad and keep throwing your mechs away, your going lose Elo. If your bad and rack up loads of points at the expense of your team, your going to lose Elo. If your good and know when to sacrifice your mech, or if your good and can rack up loads of points supporting your team, your Elo will rise.

[MischiefSC]

Bhael Fire, on 09 April 2014 - 08:12 PM, said:

That's what I've been saying too; Elo is wrong for a game like MWO.



Ah man, I could have saved you a LOT of wasted time typing if I could have stopped you after this sentence. I think I see the source of your confusion about how a matchmaker should work in a game like MWO:

Winning ≠ Good Match

And yes, of course my W/L ratio is high; that's the problem. It's inordinately high (around 1.3 for weeks now and getting higher) because I keep getting matched with players above my pay grade. The game thinks it needs to "challenge" me, when all it's doing is tossing me in meat grinder, after meat grinder; I often die regardless if my team wins or not.

That's the issue with Elo in a game like MWO. As I explained earlier, it's too easy to fall into Elo "funnels" (as I call them) because the game equates "winning" with player skill...which is NOT even remotely in the same ballpark when we're talking about a game that puts 11 random people on your "team."

So, it's entirely possible to get caught on a higher Elo "plateau" where, because of your Elo, you are forced to fight other players that are better than you or at the very least just as good. So...while on this plateau you get batted around match after match as the really talented players and groups dance the 50/50 dance....until you eventually get caught in downward funnel to a plateau closer to your personal skill level.

So to answer your question, what do I want from the matchmaker? Well it has nothing to do with "winning" because winning is meaningless in this game. It means nothing at all (at least not in it's current form).

What do I consider to be a good match? There are a lot of things that I think contribute to a good match, the least of which is dying match, after match...after match without any kills....for days....weeks on end.

That's not "fun" and even if you "win" it still makes for a frustrating gaming experience when it happens as frequently as it does in MWO.

That makes sense, and is one of the things I complained about before. Fortunately the new tier system for matching helps fix that.

Currently you absolutely can, and do, get all manner of issues derived from trying to match high/low Elo to a target or even just pugging with a high enough Elo that you end up dropping with a significant portion of premades. It can make it a lot harder to really get your teeth into a match. While there's only 14% of the players dropping in premade teams the benefits that offers is going to stack them significantly into the higher Elo tiers.

Fortunately the same patch that's bringing 3/3/3/3 is going to break Elo into 3 bands. It sounds like you're probably over 1500 Elo, especially with a 1.3 win/loss. What this means is that you'll drop in matches with teams ranging from 1500 to 2800, giving a wider but flatter distribution of Elo scores. It will also mean that you won't drop with newbies or people who struggle to even keep a 1.0. Everyone on both teams is going to have an Elo score from 1500 to 2800 and you'll be in matches that just try to keep everyone as close to the same as possible score-wise in that 'band' of Elo score and still match 3/3/3/3.

So currently you might have a team of 4 people at 2,000 poptarting it up in their Victors, 4 people at 1100 with their Dragons all packed full of MPLs to go with their Gauss, and 4 people at 1400 in another 4man of spiders. That's not going to be possible. Instead you'll have 12 people, all as close to 1600 as possible but nobody under 1500, no more than 1 premade team and exactly 3 of each weight class.

The result will be actually widening the Elo on each team while keeping a skill-level minimum and maximum instead of matching high/low to a target. It will also ensure that both your team and the other team have an even mix of mechs; this means that you'll have viable targets for any sort of build be it a light-hunter or an LRM boat or a sniper.

In the end though win/loss is the only viable means of ranking people for matchmaking. It's the only variable you can't 'game' - you either do things that drive a win or you don't. I absolutely agree that winning doesn't equal fun - the matchmaker isn't about making the game 'fun'. That's done by balancing other things like mechs, weapons, gameplay and the like. The matchmaker is about building teams with approximately equal odds of winning. You can't build a matchmaker to make games 'fun'. Not everyone finds the same things fun. Generally though people have more fun when they have at least even odds of winning.

That matches at higher Elo levels tend to be brutal crushing one-sided events.... there isn't a fix for that, other than playing on off-peak times. that I can tell you for certain. That and, well, just trying to get better so you have more of an impact. That's not a criticism or saying the problem is 'L2P'. The other people on each team who are driving the crushing win for their side, they're probably having fun. They're certainly very engaged. If you're in their stomping ground that speaks well of your skills. The problem is....

Someones got to be there. I hate to say it but, again, it's not an Elo or matchmaker issue but one related to game balance.

[Abivard]

Mischief, Why do you start your reply's with straw arguments?

I do not believe that it is by accident, it is far to deliberate and occurs far to often in your posts.

First you present the straw argument that I ' claim the fundamental basis of analytics and statistics is false' which is such a ludicrous position for anyone to take it is tantamount to a declaration that if a person claims this, nothing else they say can be correct!

I have and still maintain that a basic and unrefined, unmodified basic Elo equation is unsuitable to this game.
The amount of games that would be necessary for it to arrive at a reasonably valid rating would require a very LARGE number of games.

While MM and Elo are two different systems, they DO have an interaction on each other. One must not conflate the two but at the same time one can not ignore the interaction of the two systems, as each DRIVES the other.

MM attempts to use Elo to balance the teams, it does this by Averaging the teams Elo, that is probably it's first bad move.

Your Individual Elo is determined by the results of the Match, did your team win or lose.

But with 24 players in a match, 12 a side... averaging Elo's introduces widely different player compositions at any given Elo average point. In other words, a team can be 3/4's scrubs and 1/4 elite vs 12 right on the team average players, or really any of many outrageous combinations.

What works fine on the whiteboard doesn't always work well in the Field. theoretically with a large enough player base and a large enough sample size it may one day start providing workable results, but that one day may be years or even decades in coming.

The reality is MWO does not have a sufficient player pool to use Elo as a basis for Matchmaking.

As Elo ratings in this game are used solely as the basis for MM to work, and are solely derived from Matches played under the MM systems, welp, The one drives the other and it is NOT working as intended!

This is where we come to the shit in shit out part, no matter how elegant and refined the process is, when you feed it bad stuff, bad stuff comes out.

Just because I know I how much you dislike homily's:

The proof is in the pudding.

[Abivard]

BTW

When the drop button is pushed, 86% of the time it is a solo player.
The other 14% it is a group. 12 mans are not counted.

This does not mean 86% of the players are solo and only 14% of players group!

Basically what it is saying is that on average every team has at least a 3 man premade on it, or two premades on average in every drop! discounting 12 mans where all 24 people in the drop are in premades.

Of course these results were 'Cooked' by PGI as they appear to have been taken at the point in time where UI 2.0 had made friends lists virtually unusable, and many team players out of frustration were dropping solo or simply not playing MWO.

For whatever reason PGI seems to want to downplay teams and premades so they are presenting the stats in such a way as they did.

the real percentage of players dropping solo vs players dropping in groups was rather close to 55/45. As PGI refused to really say how many 12 mans, thou they hinted at one point that 1% of the time the drop button was clicked, it was in a 12 man.
Just in case someone did not know this, only the party leader presses the drop button, and the drop button is what PGI was counting, not players.

Edited by Abivard, 10 April 2014 - 11:52 PM.

[Joseph Mallan]

Roadkill, on 09 April 2014 - 07:47 PM, said:

Bhael Fire, your complaint is valid but you're leveling it at the wrong system. You even responded to the post that should have caught your attention.

No matter how accurate your skill ranking is; no matter how perfect the system is that's generating those skill rankings; no matter how fast your skill ranking converges on your true skill... none of that matters if the matchmaker basically throws it all out the window and selects players from a 1400-point ratings range.

That's not a problem with Elo. That's a problem with the matchmaker, which is what I've been saying all along.

I guess I don't see this as a problem them RoadKill. Rookies work with vets all the time, in all facets of life, even in recreation. I am someone who is happy there is a better and a worse player than me in every match. Experience and skill are both necessary for a good team.

[Kaldor]

o0Marduk0o, on 05 April 2014 - 03:43 AM, said:

Carry harder...

We have had this issue in the past and my back is sore... Run 4 man, whatever builds we want, the other 8 pilots on the team "derpherd to the pugzapper" and so your 4 man is left to fight the entire other team. So what do you do? You either run super heavy, assaults, or run pure meta builds. Either way it sucks.

Roadkill, on 05 April 2014 - 04:51 AM, said:

Elo works just fine in multiplayer games, and has been repeatedly proven to do so. It just takes longer for your rating to stabilize.

The problem with MWO is that the matchmaker uses too wide of a rating spread when creating matches, which is necessary to create matches in a reasonable amount of time because the player base is so small.

Exactly.... The player base is too small, so ELO never really has a chance to work properly.

Funkadelic Mayhem, on 06 April 2014 - 12:15 AM, said:

Paul Inouye:

"The Match Maker uses a scoring system to determine if your team is more likely to win or lose"


THIS is the issue.

Even if it was taken out of context, its still funny

Abivard, on 06 April 2014 - 05:24 PM, said:

This would work ok sorta maybe IF the player base was HUGE, I mean 100's of THOUSANDS of players looking for matches at the same time. But this is not the case with MWO. So it will grab solo players with Hi Elo, match them with one or MORE players of relatively low Elo to average them out to be the same as an equal amount of individuals whose true Elo is the Average for that Drop.

Another spot on post. We simply do not have enough people playing to use ELO properly.

New players should NOT be in the mid ELO levels in this game. They need to start at the bottom. All having them at mid ELO does is frustrate the players that are legitimately there.


Solution: Go back to pure weight/class matching with up to groups of 8...

[MischiefSC]

Abivard, on 10 April 2014 - 11:31 PM, said:

Same 'ol baseless stuff

Show your work please. Show me the mathematical principle that identifies how you can not determine the statistical impact of 1 person out of 12 in randomly determined 12 v 12 teams.

That you don't realize that what you're arguing is that statistical analysis doesn't actually work just proves the point further.

I've linked you, repeatedly and in numerous threads, to the documentation behind not just Elo and Glicko but exactly how and where the Elo equation is used in TrueSkill, which extracts not just the players impact on probability of a win for their team but does so across almost any game or challenge environment.

So, again. Please show proof that solving for the variable of a single individuals performance out of 12 people on a team vs another team of 12 is not possible.

Elo infers performance from historical data involving other tracked individuals. Do you understand why it uses a logical and not a normal distribution? Do you understand what the k-factor is and why its weighting is important?

The only difference in distilling Elo down to the player and not the team level is the corresponding increase in the number of samples required to get reasonably accurate readings. That's the Law of large numbers and central limit theorem. There is no mathematical difference though in predicting a players performance out of win/loss data then there is, say, an organized sports teams, so long as the individual player has a large enough sample of matches from which to infer his specific performance out of a variable group of teammates. It's actually easier for pugging because then you don't have to individually model all his teammate and extra their performance from the team and account for any variance created by specific individual synergy.

Anyway. I'll save my time and yours and offer my mathematical prediction right now -

You don't have any facts, data, formula or otherwise real analysis from which to offer your opinion other than that it matches your opinion and personal anecdotal experience. That's fine, you don't need to. That you don't like the matchmaker is peachy. That you don't like Elo is peachy. That you think your opinion is in any way, shape or form relevant to the their viability (good or bad for that matter) is just sorta funny.

You're wrong. That's okay, people are wrong sometimes. That you're aggressively wrong.... well, at least it makes you a good foil for showing everyone else who isn't unable to separate their opinion from mathematical fact exactly how and why Elo works in MW:O. For that I thank you.

[Charons Little Helper]

Kaldor, on 11 April 2014 - 05:01 AM, said:

New players should NOT be in the mid ELO levels in this game. They need to start at the bottom. All having them at mid ELO does is frustrate the players that are legitimately there.

By the way Elo works - they must start at mid level. Whatever number you give to new players will eventually be the middle.

New players are given an Elo nerf during their cadet period.

And at least it isn't as bad as back when new players were both new to this game and piloting stock mechs.

[Abivard]

MischiefSC, on 11 April 2014 - 12:57 PM, said:

Blah,blah,blah.

That is what you offer? more straw arguments? More of the 'everything is working fine, nothing to see here, move along folks' comments?

You have not shown anything, you merely try to intimidate others to your point of views.

You have yet to respond to any points I have raised outside of that they should be ignored.

You ask others to 'show their work' but you never show yours.

You ask others to trust your views, but you don't want them to see sources. You are being deliberately obtuse or truly lack reading comprehension when you twist what is written to fit your own ends.

The proof of it being misapplied and highly inefficient is in the results everyone sees in MWO.

Things are not working fine and dandy in MWO, keep on in your denial of the reality, after all, it should work, your paper work proves it! To bad no one is allowed to SEE your paperwork.

[MischiefSC]

Abivard, on 11 April 2014 - 01:26 PM, said:

That is what you offer? more straw arguments? More of the 'everything is working fine, nothing to see here, move along folks' comments?

You have not shown anything, you merely try to intimidate others to your point of views.

You have yet to respond to any points I have raised outside of that they should be ignored.

You ask others to 'show their work' but you never show yours.

You ask others to trust your views, but you don't want them to see sources. You are being deliberately obtuse or truly lack reading comprehension when you twist what is written to fit your own ends.

The proof of it being misapplied and highly inefficient is in the results everyone sees in MWO.

Things are not working fine and dandy in MWO, keep on in your denial of the reality, after all, it should work, your paper work proves it! To bad no one is allowed to SEE your paperwork.

As predicted, no reference to any mathematical proof of Elo being unable to function in this environment.

Show my work? Okay. Here you go:

Law of large numbers
Central limit theorem
Variance and standard deviation of a random variable

Just to help you understand that one here's standard deviation

How to identify and solve for margin of error

That's a good start.

Just because you seem to keep forgetting, I'm going to link you again to how Elo is used in TrueSkill.

Here you go

Elo in MW:O is not designed to stack rank you. It's not even required to be that accurate - there certainly are more complex systems or additional qualifiers that could and would refine your Elo ranking down more precisely.

They have absolutely no use in MW:O as it stands right now however because all the Matchmaker needs is a rough approximation of skill; currently you're matched in a high/low to target system, going forward you're just going to be in one of 3 buckets - 0-1000, 1001-1500, 1501-2800. Even a 10% margin of error is perfectly acceptable and within a useful tolerance.

Elo is designed to measure performance. As performance varies in circumstance when measuring something like skill in people it can't therefor measure absolutely but must do so by inference. The most accurate means of measure by inference is via historical data from similar circumstance. The biggest challenges to that are self-selection of allies and opponents (hence why premades skew Elo) and the fact that peoples performance is not static - ergo there is no absolute score you can set a persons performance at. Your own behavior will skew your relative rating, thus any rating must be inferred.

I realize you're not going to read those links. If you actually cared you could have educated yourself via Google in an afternoon and understood why the idea that Elo simply can't work in MW:O is something that anyone who does understand statistics would laugh at.

That's why I'm still posting here; it's not really to you. It's not a lucky guess to say that you don't understand statistics or how/why Elo would work because if you did you wouldn't hold the position you do. You don't want to understand, you want to blame something you don't understand for some bad experience you're having. What this does do however is let me point out, for those people who do care and would like to understand, exact what and how.

What's funny is that Bhael Fire already hit this one on the head - winning /= fun. The issue isn't that Elo doesn't work for MW:O. It works exactly as designed and does exactly as it should. They should and hopefully will continue to use it as the basis for whatever the matchmaker evolves into. As the player population grows and more value can be found in seating peoples scores more precisely hopefully it'll drill down to performance via chassis, loadout and teammates in a drop vs other teams composition and map. Currently that's irrelevant; it's got issue enough just filling matches roughly with what's available.

The issues that people are blaming on the matchmaker have nothing to do with the matchmaker but instead other balance issues. The importance of the deathball, how heat works, tonnage mismatches and huge disparities of viability between chassis. Pinpoint damage, poor SRM performance, lack of objectives aside from deathmatch. Lack of lobbies, it's a long list. Fortunately much of it seems to be getting chipped away at.

I work in a field where psychology and statistics cross, identifying peoples behavior to specific interpersonal encounters, distilling from that what motivated what response and predicting future behavior and adjusting business metrics to respond to (and alter/control) that behavior. I know that the biggest issues in MW:O are more community (or lack of it) driven; some of that's being fixed by the changes in the matchmaker. 3/3/3/3 will create enough predictability to allow players more control over the viability of their build in any given game. If you always know there will be 3 lights, a light-hunter will always be useful for example. Matching Elo to tiers ensures a more stable population dispersion within each tier, significantly reducing the impact of particularly high/low performers to skew a matches outcome.

What's missing and in being missing actually creates most the issues you're complaining about is a lobby system. It's part of why people gravitate to the forums; there's no way to actually establish a connection with your fellow players. You rotate through random groups and unless you join a guild (which has plenty of barriers on its own, as well as plenty of limitations) you have no means of connecting to your own performance as part of a larger social whole. If I could, for example, go PUG in matches that are largely populated with members of Murphys Law or even just chat with them in the lobby between games I would establish a value on their opinion of how I play and an expectation that they would see me play often enough in the future that such an opinion had weight. In a lobby or more integrated social environment that web or inter-significance would tie me emotionally to my performance in the view of how other people would see it. Currently I've got no reason to care, every game is completely selfish. Hopefully a lobby comes with/after CW.

Anyway. A whole other topic and it's got a lot of facets. MW:O has tons of issues. The use of Elo isn't one of them. The new matchmaker is a HUGE step forward. That you don't understand how or why Elo works for MW:O is a personal issue and one you choose to create, given that it's an easy one to solve. I'm not going to waste two hours of my friday night walking you through an equation to solve for the statistical impact of 1 in 12 in a 12v12 win/loss environment with 23 variable players, identify the margin of error and adjust required sample size accordingly. Do it yourself - but you won't. Instead you'll pretend that it's all too hard and that you being completely ignorant of statistics and how they work still qualifies you to judge the viability of a statistical equation in a given analysis because of your anecdotal experience and uninformed personal opinion.

Shine on, you crazy diamond.

Edited by MischiefSC, 11 April 2014 - 08:39 PM.

[Abivard]

You write a very lucid and well thought out post, I applaud and agree with just about everything you say....

However I still see you wish to make a straw argument out of my position. parts of your post seem to indicate it is much more a personal dislike of me more than what I am really saying that is coloring you towards what I have said.

I have tried framing it different ways to no use it seems.

First off no one outside PGI knows what the equation they are using actually is, it was said to be an Elo like application. If you know the equation as well as the K factor used, etc.. please do share.

If PGI is using The basic Elo equation with no modification, it will not be efficient at producing usable results in a desirable time period. This is due to the HUGE amount of the various influences beyond just the players skill, which is much more than simply 24 total people in a match.

The only Variable of which PGI makes a nod to is weight class distinction. But even a single mech variant can be built to vastly different 'builds' let alone all the mechs and variants across a weight class.

Let us break down how Elo effects us on a practicable level.

I have a rating, this rating is compared to an opponents rating, if mine is higher I am expected to win, the greater the difference the greater the chance I should win. Elo assumes everything else is on an equal basis and only the players skill affects the outcome.

The point where Elo assumes everything is equal...... But we Know that is not the case with MWO, the players are never individually matched and placed on an even footing where only the skill of the player decides the outcome.

The game mode, the mech chosen, the load out of that mech none of this is accounted for, now aggravate this situation by randomly assigning 11 persons to that players team, all of whom will also be in random mechs in random builds at random points of mech skills earned. We can safely conclude that PGI does not have people of the caliber of Arpad Elo working for them so they cut and pasted the Elo formula without any attempt to account for all the various Influences present that determine whether a player wins or loses.

It simply ASSUMES only the one person's skill matters.

We all know this is NOT the case in MWO, it is the team's efforts that win or lose the match. Your skill is often buried under the other 11 people on your team and the 12 on the other team.

In conclusion I make no claims that Elo itself does not work. I claim it does not apply to MWO as it is.

Attempting to assign an Elo rating to players as a means to matchmaking balance in MWO is unwise, doubly so as the player base is to small,so it averages Elo ratings when Matching which skews the results even more. PGI knows it, or they wouldn't be slapping band **** like 3/3/3/3 and buckets.

A better system than a simple Elo rating is needed to base matchmaking off in MWO!

Elo is not up to the job by itself!

Address these two points not the straw argument that you are saying I claim Elo itself is false.

[MischiefSC]

Abivard, on 12 April 2014 - 01:13 AM, said:

You write a very lucid and well thought out post, I applaud and agree with just about everything you say....

However I still see you wish to make a straw argument out of my position. parts of your post seem to indicate it is much more a personal dislike of me more than what I am really saying that is coloring you towards what I have said.

I have tried framing it different ways to no use it seems.

First off no one outside PGI knows what the equation they are using actually is, it was said to be an Elo like application. If you know the equation as well as the K factor used, etc.. please do share.

If PGI is using The basic Elo equation with no modification, it will not be efficient at producing usable results in a desirable time period. This is due to the HUGE amount of the various influences beyond just the players skill, which is much more than simply 24 total people in a match.

The only Variable of which PGI makes a nod to is weight class distinction. But even a single mech variant can be built to vastly different 'builds' let alone all the mechs and variants across a weight class.

Let us break down how Elo effects us on a practicable level.

I have a rating, this rating is compared to an opponents rating, if mine is higher I am expected to win, the greater the difference the greater the chance I should win. Elo assumes everything else is on an equal basis and only the players skill affects the outcome.

The point where Elo assumes everything is equal...... But we Know that is not the case with MWO, the players are never individually matched and placed on an even footing where only the skill of the player decides the outcome.

The game mode, the mech chosen, the load out of that mech none of this is accounted for, now aggravate this situation by randomly assigning 11 persons to that players team, all of whom will also be in random mechs in random builds at random points of mech skills earned. We can safely conclude that PGI does not have people of the caliber of Arpad Elo working for them so they cut and pasted the Elo formula without any attempt to account for all the various Influences present that determine whether a player wins or loses.

It simply ASSUMES only the one person's skill matters.

We all know this is NOT the case in MWO, it is the team's efforts that win or lose the match. Your skill is often buried under the other 11 people on your team and the 12 on the other team.

In conclusion I make no claims that Elo itself does not work. I claim it does not apply to MWO as it is.

Attempting to assign an Elo rating to players as a means to matchmaking balance in MWO is unwise, doubly so as the player base is to small,so it averages Elo ratings when Matching which skews the results even more. PGI knows it, or they wouldn't be slapping band **** like 3/3/3/3 and buckets.

A better system than a simple Elo rating is needed to base matchmaking off in MWO!

Elo is not up to the job by itself!

Address these two points not the straw argument that you are saying I claim Elo itself is false.

Here's the thing -

It's not a strawman argument. That you think it is a strawman argument is part of the problem - I would be the first one to say that to expect you to have to have a working knowledge of statistics, modeling and analytics to be able to criticize something involving statistics is unfair. However, at a certain point in debating a topic you need to be able to express exactly what you think is failing. You've done so in an 'I feel that' or 'my opinion is' on what you think, from an uneducated perspective (I don't mean that as being ignorant or in a negative context, simply to indicate that you're not fully aware of what the principles behind the topic are) but the reality is that the formula works just perfectly for this situation.

We know exactly what the formula is -

This is the link to exactly how it's calculated.

It's possible that they've adjusted the k-factor. We also don't know if they're using a standard, logical or bayesian (gaussian) distribution for the results. What is that important? Sampling distribution of difference between means.

What's all that stuff mean?

The principles upon which Elo as a means to infer a players probability of driving a win when playing with and against players of a specific skill range are rock solid, long proven and not really debatable. You don't have to understand those reasons; I certainly don't understand more than a fraction of a percentage of all the real principles and mechanics about how my computer works to be able to use it. Yeah, generally I know a ton of stuff, but the actual mechanics of a microprocessor? Hell no.

That doesn't mean it doesn't work. It doesn't mean that if X:Rebirth crashes all the time my CPU was poorly designed; my CPU may be perfect and the issue is that X:Rebirth was coded like shit.

It also doesn't mean that the crashing isn't a problem - it just means that I can't adequately say what (if any) issues may be attributed to my CPU vs how the game is coded. I can say I know enough about coding and what people who do have the knowledge and skill set to tell the difference say. Or I can just say 'it sucks, anyone who buys X:Rebirth deserves what they get and no way in hell am I giving Egosoft my money ever again'.

You are absolutely welcome to say that 'Elo doesn't work for the matchmaker for MW:O'. Just understand that when you say that you're wrong. More accurately you can say 'I don't enjoy the matches in MW:O!' The cause of that is not using Elo in the MatchMaker. The cause of that is a number of balance issues. The use of Elo in the matchmaker is why we're not still in the environment we were in before where being in a 4man meant you won 90% of your matches and literally about half your team wandered off and died stupidly every match. With the tier system coming up it's going to mean that you'll drop in matches with people all roughly in your own skill range - though that will vary based on their familiarity with chassis and loadout. They are still going to make stupid choices and die at the wrong time, they'll still run off alone sometimes or just make poor choices.

My issue is you attempting to insist that Elo doesn't work and can't work for MW:O. I am familiar with statistics, modeling and analytics. Saying Elo doesn't work for that is literally, and I do mean literally, like saying gasoline doesn't work to fuel cars because your car keeps breaking down. No, gasoline as a hydrocarbon fuel works great - it's got some issues, yeah, but that it functions well as a fuel isn't a reasonable thing to debate. If your car is breaking down then something else is going on.

Every additional factor you add to refine the Elo equation introduces its own issues and potential problems. It's also not going to change the MM results to any meaningful degree. This is also what I seem to be really struggling to help you understand -

The matchmaker can only match you with the people hitting 'launch' within the same 2 or 3 minute timeframe as you. In fact functionally it's closer to a range of 90 seconds; the total fill time for a match is about 3 minutes.

A perfectly accurate Elo score doesn't change who is available to match with. High population density is what gives you a more homogenous (as in balanced) Elo spread per match, NOT a different criteria for finding that score.

Also you're trying to take anecdotal examples of extremes and say they disprove Elo as viable. That's not the case - they are a product of the process and Elo absolutely takes them into account and does so effectively. It's why it has a k-factor and why the matchmaker predicts your win/loss odds in a given match and adjusts your score change (if any) accordingly.

A persons skill varies dramatically in any given event. Nobody plays at 100% all the time, thus the term 'inference' and not 'absolute' to describe the results. While you can get extremely accurate inferred results for someones average performance and over a big enough set of samples predict their success or failure very accurately that average will be the product of significant small-sample variance. This means that a skill of 50% will include plenty of 0% and 100% examples, even if the majority fall between 40% and 60%.

A matchmaker that gives you perfectly balanced matches every game is a poor system for gauging player performance. You have to play in matches against people within a range of your own skill in matches you are expected to lose or expected to win to allow for mistakes to be caught in peoples score and that score adjusted accordingly.

So, again, Elo works for MW:O. When the population gets denser then a more granular and detailed version would be warranted. For right now the most it would do would be to possibly alter the makeup of teams for players likely within 100 points of either 1,000 or 1500 Elo and only on some matches - a statistically negligible group who performance, good or bad, should end up moving them out of any margin for error anyway.

Your skill matters. I could chart your skills impact in teams of 100 vs 100 - it'd just take a lot more matches and a larger player base (to prevent too much repetition of the same players on each team). That you don't think it does is not a problem of the equation, it's a problem with your perception.

You want to blame bad games on something blame it on balance issues and. The matchmaker is doing exactly what it's supposed to do.

[Kaldor]

Charons Little Helper, on 11 April 2014 - 01:05 PM, said:

By the way Elo works - they must start at mid level. Whatever number you give to new players will eventually be the middle.

New players are given an Elo nerf during their cadet period.

And at least it isn't as bad as back when new players were both new to this game and piloting stock mechs.

No, they do not need to start them in the middle. Drop them down 500 points or so, so they really are in the starter ranks. If they are good enough, they will get to mid level and above quickly.

New players are given an ELO boost during cadet. Then after the first 25 matches, they get nerfed a bit, but not enough IMO.

Ever wonder why stock mechs suck so much? Bad heat system which pretty much makes SHS worthless. Say it with me: "heat coffin"

[Axeman1]

No sympathy for premades, just laughter.

[Master Maniac]

Do you guys even have the slightest inkling as to how a protracted battle is typically fought and won?

The common "wisdom" here is that if the match does not arbitrarily end in a "close" victory or an "even stretch," it is therefore inherently the fault of some grave lack of "balancing."

This is complete trash. Almost every one of those 12 to X matches that get so much QQ are the DIRECT result of the losing team running out of steam in a battle of attrition. It's like a house of cards: when the base starts to waver, it's the beginning of the end. The line breaks, and units get crushed under a focused offensive. One side tends to lose a lot of units and very quickly.

That does not automatically mean one team is so very simply much better than the other. That is such a hamfisted, idiotic and unthinking observation. The fact that EVERY game doesn't end in a "perfectly close" score is a testament that this game is working exactly the way it should. Otherwise you'd just be playing a formulaic numbers game without even knowing it.

[Nathan Foxbane]

Master Maniac, on 13 April 2014 - 12:49 PM, said:

Do you guys even have the slightest inkling as to how a protracted battle is typically fought and won?

The common "wisdom" here is that if the match does not arbitrarily end in a "close" victory or an "even stretch," it is therefore inherently the fault of some grave lack of "balancing."

This is complete trash. Almost every one of those 12 to X matches that get so much QQ are the DIRECT result of the losing team running out of steam in a battle of attrition. It's like a house of cards: when the base starts to waver, it's the beginning of the end. The line breaks, and units get crushed under a focused offensive. One side tends to lose a lot of units and very quickly.

That does not automatically mean one team is so very simply much better than the other. That is such a hamfisted, idiotic and unthinking observation. The fact that EVERY game doesn't end in a "perfectly close" score is a testament that this game is working exactly the way it should. Otherwise you'd just be playing a formulaic numbers game without even knowing it.

This is phenomena is occasionally referred to as Combat Loss Grouping (CLG) when it applies to damage by some of the more lore savvy players on these forums and it is a major cause of stomps.

For those players wondering what CLG is, it is a point at which accumulated damage across multiple units over the course of an engagement becomes critical and those units are destroyed or otherwise neutralized in rapid succession. Tactics and strategy can influence how quickly or even at what stage in the game the CLG is reached. It is why a heavy push can suddenly turn into a route as the attackers take too much damage to sustain the assault and become easy prey for a counter assault. Conversely, it is why a successful push completely flattens the target team as damage output not only destroys focused targets, but secondary targets accumulate too much damage near the end to mount a successful defense or counter attack.

CLG does not refer to destroyed units as by definition they are already lost. The lance lost probing the other team's flank is not part of the CLG as the lost lance has already past its own CLG. The two lances holding the line are subject to CLG as they accumulate damage. However the loss of the probing lance does affect the CLG of the remaining two lances as they are now the sole focus of the other team. Multiple CLG's can occur for both teams in any give battle which is why 'Mech often hit the kill board in small groups rather than the whole team collapsing one after the other. A unit lost is not necessarily a bad thing for a team either. If said unit does enough damage or provides critical information before destruction that contributes to forcing the opposing team's CLG. A well done sacrifice, even if it results in no kills, is a great example in a 'Mech stalling and damaging several 'Mechs to their CLG point at the cost of the 'Mech. This action forces a group of opposing 'Mechs to their CLG while stalling the team's own CLG.

[Master Maniac]

Hellen Wheels, on 13 April 2014 - 04:33 PM, said:

I'll tell you what it means: NOTHING.

You admitted yourself that you have no idea what any of these purported "clues" PGI has provided mean anything at all. Oh, well, they might have changed the "K" factor, you say? To what? You don't know.

You state quite rightly, that "We also don't know if they're using a ... [insert your latest speculation]"....

Fact is: YOU DON'T KNOW AND EVERYTHING YOU POSIT IS BASED ON PURE SPECULATION.

So admit it: their "system" is nothing but smoke and mirrors. Just admit it is random, and you'll save yourself all that supposed brainpower you've applied to absolutely nothing but random speculation.

See also: "vaporware"

Now, do you feel better?

[Murzao]

Carry harder. If you are 50/50 you aren't as good as you think you are. Hell I'm 48-32 since stat reset in 100% Centurions and 75% of that is pugging. That's not even me trying either I'm just testing builds sometimes....like:

Just yesterday I was in a 3LL/2SRM2 Cent Caustic Valley this highlander meta 2PPC/AC20 poptart starts targeting me so I lay into his CT wih the 3LLs, he shoots off my shield arm and leg and by then his CT internal is cherry red. What's he do? Goes and hides behind an Atlas.

Oh and I won a match a few hours ago 12-3 when we were outtonned by 150 and had a DC on top of that, love the newb heavies shooting at the Raven flock when they let my Wang AC20 em in the back.

Most 'meta' players are only good at 1 thing...hiding behind others. In fact half the strat of said meta players is to cower behind teammates....because they are terrible at 1v1s.

Edited by Murzao, 13 April 2014 - 06:41 PM.

[MischiefSC]

Hellen Wheels, on 13 April 2014 - 04:33 PM, said:

I'll tell you what it means: NOTHING.

You admitted yourself that you have no idea what any of these purported "clues" PGI has provided mean anything at all. Oh, well, they might have changed the "K" factor, you say? To what? You don't know.

You state quite rightly, that "We also don't know if they're using a ... [insert your latest speculation]"....

Fact is: YOU DON'T KNOW AND EVERYTHING YOU POSIT IS BASED ON PURE SPECULATION.

So admit it: their "system" is nothing but smoke and mirrors. Just admit it is random, and you'll save yourself all that supposed brainpower you've applied to absolutely nothing but random speculation.

See also: "vaporware"

.....

What?

So you think Elo isn't real? They tell us exactly what the equation is. The only things we don't know is changes they haven't announced (which is called 'buying trouble'. That's where you assume things are different because you want to create reasons for it not to work) and what distribution curve they used.

The only difference knowing the distribution curve would make is identifying how tightly Elo scores are grouped. Doesn't change much in how it works overall, just how 'split up' the different inherent groupings of players Elo scores will be.

I gave you a bunch of links. Click on them. We know exactly what equation they are using - they told us. It works, for the same reason that all math works. If you take 1, and you add another 1, you get 2. This isn't something that only works sometimes. 1+1=2.

Here's the Elo equation:



That, at least, is if they're using a logistic curve. Again though the only difference between logistic, normal or gaussian distribution is what the curve the scores are placed on looks like; how steep it is. Given that we're moving to 3 tiers of Elo that's almost irrelevant as well. You're either under or over 1500, that's about it.

I do know. The same way that anyone who knows math knows how math works.

If math is all 'smoke and mirrors' to you then, well, I'm sure the matchmaker (and much of how the modern world works) must feel very arbitrary and punitive to you. That's a problem well outside the scope of what can be address on a forum. It has nothing to do with the absolute mathematical fact that Elo works as the basis of a matchmaker for MW:O.

Best of luck to you though. That's a tough row to hoe.