87 replies to this topic

[Flak Kannon]

As I approach 8,000 lifetime 'DROPS', 99% of that in the solo que, I am getting more and more curious about how ELO is functioning in MWO.

I can't say I read everything ever posted, but I try to stay current.

Tell me what you have read about how the ELO system works for us.

As a solo player, and maybe currently rightly so, I a paired with many, many new players.. quite often actually.

That's fine. Some I take under my wing and really try to help. Other are hopeless... or perhaps unable to understand English. Very well.. understood.

But why as I play more and more, should my experience degrade as my ELO raises and my teammates get worse to complete a range.

That is how it feels. I don't know if that's the case, but recently, it has seemed really skewed poorly in my favor. I'd rather have a HIGHER OVERALL AVERAGE ELO GROUP, than 3 or 4 ringers on each team surronded by8 or 9 low ELO, or new players.

Yes, I can join groups. I understand that

As it stand now my W/L is +150, and my K/DR is highly positive.

What else affects ones ELO.

I'd love a little transparency on this.

Perhaps also a ELO overhaul on the RANGE..??


Please have lower ELO players, or first 100 drop players have their OWN bracket.

Lets talk a little on ELO.

Discuss:

[Nothing Whatsoever]

Here's these:

http://mwomercs.com/...65#entry1626065

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

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

[Flak Kannon]

Thanks for those links.

I am still a bit in the dark.

From all that I just read, it feels like ELO is 'mostly' derived from ones previous matches and the WIN / LOSS ratio of those matches against the other ELO team. If they were higher, and I win, I go up in ELO, if the team average was lower and we win I do not raise in ELO? Does that seem to sum up what those links said?

Where does your own personal contribution to a team come into play.

I play a match, I do 600 damage, I score 4 kills, and we lose against a lower overall ELO team and my ELO will go down? Yet I personally did well. .. ?? What? Is it all predicated on the team performance?

So basically it means nothing to my personal ELO if a player that just started yesterday, and has a 1300 or so ELO makes the kill shot on me and get the kill?


I really wish their was more transparency on this, because if it's all about your W/L ration, and mine keep climbing into the POSITIVE, I am handicapping myself??? What benefit is it to play and win over time?

It seem a little odd. Why can't I join higher level GROUPS as opposed to being asked to carry more lower level teams?


It feels like I would be best served to try my best to get my Win Loss ratio back down to 1.0 or below, allowing me to have a better average spread of players instead of 3 good ones and 9 terribad ones..

I am just confused.

[Adiuvo]

Performance in match doesn't directly matter. Only winning or losing does and nothing else is part of the Elo calculation.

[Bhael Fire]

Flak Kannon, on 24 September 2014 - 03:13 PM, said:

Where does your own personal contribution to a team come into play.

It doesn't...well, not directly.

The "theory" is that if you are a good player you will help carry your team to success...which is true to some degree, but in no way completely reflects your individual skill as a player.

Your individual performance is basically rated on how well your team does on average.

[DAYLEET]

Elo cant work in a chaotic environment with 23 other players, it makes no sense. At the end of the day a win is the only thing that matters and the rest might have helped and might not. Hopefuly elo also takes into consideration game mode.

[Sandpit]

Elo is based on W\L not individual effort. Elo isn't nearly as effective when you throw it into the mix that relies on a group's performance.

New pplayers have their Elo deflated during cadet bonus. Groups have elo inflated.

Having a high w/l is not always indicative of individual elo. I cam drop and do 10 damage and still have elo go up if my team performs well.
Vice versa
I can drop and get 7 kills and do 1100 damage and have elo drop if my team loses.

Team elo means you could jave a 2000 elo and a teammate jas say 600. That would average your elo to 1300. Putting that 600 player into a match against 1300+ opponents is what contributes to stomps and what you're asking about

Elo is working, it's just not meant to average out like that

[Flak Kannon]

Thanks Sandpit.

Pretty much what I thought.

As I have a +150 wins over losses, and I'm not saying I'm the greatest player, but pretty dam influential on the winning side, hence my ELO must be on the high side.

And my ELO CAN NOT go up if my sides loses right? So any win percentage over 1.0 mean at the very least an above average ELO if average is 1300.. I am starting to understand more.

Hence I am getting alot of LOW ELO players to average the team.

I am not saying I'm great. ...
But at +150 my team really, really feels composed of many, many new and poor players that lose quite often, it would seem..

My game experience is going down as I win more. That just feel contradictory...

I'd rather have alot of really good players on my team and against me on the other team, than the massive swings in ability I am seeing on my team as I spectate, with many players not even remotely able to cope with a 12v12 game like this.

There needs to be some retooling. Maybe 3or 5 levels of brackets that your range puts you in, not this MASSIVE swing in player ability.

What I do understand now is that if I am playing with people that are still in their Cadet stage, maybe that puts a CAP on how how of an ELO I have, right? All new players start out at 1300, and the range they wanted to keep was +, - 1200 I seem to remember reading once recently..so at most I am at 2500 or so..

So if I got up to 2700 I would never see another new player I guess?????2700-1200= 1500 range..??

Things need to change in the solo que I feel.






I hope they can come up with a better system

Sandpit, on 24 September 2014 - 03:56 PM, said:

Elo is based on W\L not individual effort. Elo isn't nearly as effective when you throw it into the mix that relies on a group's performance.

New pplayers have their Elo deflated during cadet bonus. Groups have elo inflated.

Having a high w/l is not always indicative of individual elo. I cam drop and do 10 damage and still have elo go up if my team performs well.
Vice versa
I can drop and get 7 kills and do 1100 damage and have elo drop if my team loses.

Team elo means you could jave a 2000 elo and a teammate jas say 600. That would average your elo to 1300. Putting that 600 player into a match against 1300+ opponents is what contributes to stomps and what you're asking about

Elo is working, it's just not meant to average out like that

Edited by Flak Kannon, 24 September 2014 - 04:54 PM.

[Mawai]

Winning and losing are the only thing that affect your Elo skill.

Your WIN/LOSS ratio is IRRELEVANT and MEANINGLESS as far as your Elo skill goes.

Here is how it works.

1) The matchmaker takes a seed player to start the match. Their Elo defines the initial range for other players to be added to the match.

2) The matchmaker adds people to the teams alternating between them selecting players to match weight class and Elo requirements.

3) When the teams are full the matchmaker takes the combined Elo scores for both teams and uses that to decide which team is favoured to win and by how much.

4) Players play the match. The matchmaker itself does not care who wins. People have not been placed on teams to encourage a loss after a winning streak or ANY of the other odd theories I have heard.

5) The matchmaker uses the result of the match to determine if the Elo should be adjusted and by how much.

- If the team that was predicted to win .,.. wins the match then the Elo values do not change for ANYONE

- if the team that was predicted to win ... loses the match then the Elo values for all players change. The maximum change is 20 points and how many of the 20 points are applied is determined depending on how much of an upset the match was.
In this case, the Elo of each player on the losing team is reduced by the calculated amount and the Elo of each player on the winning team is increased by that amount.


Your Elo can only go up on a win and it can only go down on a loss ... BUT only in those matches which did not agree with the matchmaker predictions AND the amount by which the Elo changes will be DIFFERENT in each case depending on how much of an upset the match was.

NOW ... look at the curve in the third link from the second post in this thread.

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

It used to be that new players were seeded near the middle of the distribution. I believe that they may now apply a handicap for the first few games ... maybe coinciding with the cadet bonus? longer? Part of the issue is that the game has no way to tell real new players from alt accounts. In any case, if you note how the distribution is peaked in the middle of the range and coincides with new players ... even above average players are within a couple of hundred points of the beginning players ... this is why you will likely see new players on most teams from time to time. (When you add in the fact that the matchmaker has to relax Elo constraints from time to time to form teams faster .. then it is almost guaranteed that there will be some new players on the teams).

Edited by Mawai, 24 September 2014 - 05:19 PM.

[Adiuvo]

Flak... if your W/L is 4150/4000 that's a 1.04 W/L ratio. Elo is basically working perfectly for you, you're winning about half your matches indicating that the skill level of your team and the enemies is consistently fairly even.

Edited by Adiuvo, 24 September 2014 - 05:20 PM.

[Willard Phule]

Oh, for a straight line like this every day...

Here's what we know about Elo.....

1 - You start out with a random low number that's SUPPOSED to be lower than what is in the "General Populace" at any given point. It's a lie, by the way. You think you're safe during your first 25 matches...you aren't. Depending on how many players are on at any given time, you might be thrust into a normal match and get your lunch handed to you. If you have a problem with that, take it up with Russ or Paul on Twitter....they don't pay attention to the forums.

2 - You are given an artificially inflated Elo score ( as if it matters ) as soon as you finish those first 25 matches. It's like 1700-1800...but hardly matters because the Matchmaker is given a 1400 point spread to put people together for a team. That means there is "no Elo" as far as the Solo queue is concerned. Suck it up, buttercup, you're fodder and will be given no chance to learn. It's better for you, and PGI, if you just drop $100.00 into the game then uninstall. Trust me, you don't want to get hooked on this crap.

3 - Your Elo only moves (either up or down) if you do the opposite of what the Matchmaker predicts you're going to do. Ok, I'm getting ahead of myself here. There is a Matchmaker. It's most likely an inflatable sex doll at the PGI office, but whatever. When it puts two teams together, it predicts which team is going to win and which one is going to lose. This is VERY IMPORTANT. If you were put on the team predicted to win but you actually lost, your Elo goes down. Same thing if the opposite happens. The MAXIMUM movement is 50 points....again, with a 1400 point spread, 50 points is pretty much useless.

4 - The Matchmaker uses an "average Elo score" to put teams together with. This is also very important. The Matchmaker was re-written around the Clan release to make sure that each team has an equal "average" Elo score. That means that if Team A has 12 people that all have an Elo score of 1500, then their average is 1500. If Team B is composed of 4 people with an Elo score of 2500, 4 people with 1500, and 4 people with 500 Elos...they have an average of 1500 as well. The discrepancy is obvious to everyone that plays the game more than once a week. Not so much for the PGI developers....they tend to have lower than average Elo scores, as evidenced in the latest "kill a Dev" challenge. Most people never saw them...because most people have a higher Elo than they do. It is what it is, but keep in mind that everyone in this game is judged on a baseline set by sub-standard players.

4 - I can't speak for the Group Queue but in the Solo Queue, your Elo amounts to nothing. You stand as much chance ending up on a team full of competent players as you do with new players at any given time. Nothing adjusts those odds. If you have issue with that, take it up with Russ or Paul on Twitter. None of them pay any real attention to the forums here.

[Kvaneal]

Matchmaker seems to prioritize the 3/3/3/3 weight preference over pilot skill right now. Discussion on these forums indicate that most of us believe that skill > mech. They need to loosen up the 3/3/3/3 requirements and let the matchmaker focus on keeping player skill more closely matched.

[MischiefSC]

I am going to (perhaps unwisely) present a very, VERY rough means of calculating your Elo.

All that matters is wins more than losses. Nothing else is relevant.

It is different for each weight class.

Not every win will change your Elo. Only wins where it predicted you would fail. Conversely you only lose Elo when it predicted you would win.

So:

Add up all the win/loss results for every mech you have in a given weight class. Technically these results will include a bit of your 'archived' results but there's no accurate way to sort those so just use 'current'.

However many more 'wins' you have than losses, multiply by 7. So if you have 100 more 'wins' in heavy mechs than 'losses', that would give you a total of 700.

Add that number to 1300.

So that would, in the example above, give you an incredibly crudely estimated Elo of 2,000 in heavy mechs.

There is no way of being certain of this number because we don't know exactly how often your 'wins' were in matches you were predicted to win or what the relative Elo was.

If anything this calculation will probably give you a number on the high-side, but I suspect everyone is happiest thinking their Elo is higher than it really is

[Davegt27]

They have something better than ELO it’s called a computer

This is my very first online game to me it’s blatantly obvious

I remember the computer chess games back in the early 1980’s it would go ahead something like 26 moves working out every possible combination

IMHO that’s what you see a very balanced (overall) game

And they are getting better at the balancing of the game as time goes on, they can even balance the weapons mid game

I suspect the early players did not have a lot of balance so there number might be a little skewed but give it time

The moral of the story is be careful what you ask for you just might get it

Edited by Davegt27, 24 September 2014 - 06:33 PM.

[Sandpit]

That's another thing to consider. Individual Elo based on team results mean that each match can mean one player has no change while another has a drastic change depending on a win or loss.

Essentially this is how Elo works in general

Player A = 1000 Elo rating
Player B = 1000 Elo rating

When these two players compete, their Elo may not change at all or a very minimal change regardless of outcome because they're "evenly" matched

Player A = 1200 Elo
Player B = 900 Elo

If player A wins, their Elo will not change. If player B wins, B's Elo will increase while A's Elo will decrease.

Now when you throw in 11 other people's Elo and average that against another team of 12, you can get some pretty drastic variations of individual Elo even if the team's average Elo is evenly matched against the opponents'.

There's a few factors that go into this in regards to MWO. It's not JUST Elo that's considered. There's also map, game type, and release valves as well. One way to help mitigate the Elo variations is to cancel match finding before the release valves are hit. I usually hit cancel at about the minute mark. That helps keep the release valves from tossing me into a match with huge Elo discrepancies. If you select one game type, you're almost forcing the release valves to spread out since you can only be placed in a match with that one game type. Map selection would cause even further discrepancies.

One of the things that would help (at least in my opinion) is starting new players at the bottom end of the Elo scale and having them play their way up. It would help mitigate roflstomps and decrease chances of new players playing above their Elo ratings. It also allows them to find their "normal" Elo rating gradually as they'll eventually hit the plateau and "even" out so they'll also experience less losing streaks (as will veteran players having to "carry" new players).

Now you also have to realize that you have an Elo for each weight class as well. Even though you might be an "ace" in mediums, you might be more of a "scrub" in assaults. That also means that you may be an "ace" in a particular chassis, but not so much in the other chassis in that weight class.

Personally I think Elo in general was a bad way to go in a team based game. It's an individual skill rating based on a team's wins or losses as opposed to individual effort within that match. If Elo is going to be used, the stats need to be based on individual performance.

[MrMadguy]

Learn to use Google. Search "Forced 50% winrate MMR".

[MischiefSC]

Elo isn't a forced 50% win rate. If your win rate is only 50% then you're not improving. Many will find themselves in to 55-60% win rate range.

Someone above gave some incorrect information so let's clarify that -

In your first 25 matches your Elo is 1,000.

After that you go to the default of 1300.

Most matches vary from team to team by less than 50 pts in Elo.

Most matches have a variance of high to low of a few hundred points (not 1400). Is it possible to go to 1400? Sure, but nobody on this thread (save possibly Adiuvo) would ever conceivably be in such a situation because almost everyone on this thread is probably sub-1800.

The real problem with Elo is that it doesn't work like 'scaling enemies' in Skyrim and the like - where you are level 20 so all enemies are level 15, so you win all the time.

Reality is that players are never as good as they think they are. Ever.

Elo works just fine for this sort of environment. That's the biggest problem with it for a lot of people.

[pulupulu]

Don't worry about your own elo, if you get too good, MM will be kind enough to bring it down by giving you multiple disconnected team mates repeatedly; then you can go back playing with noob team mates.

[Glythe]

MischiefSC, on 24 September 2014 - 07:39 PM, said:

Elo works just fine for this sort of environment. That's the biggest problem with it for a lot of people.

Elo is flawed in a game like this because it is designed for one player against another.That's fine if 2 players go away from the entire rest of their team and duel with no assistance. But that never happens.


How can you possibly decide Elo with multiple players leading to your death?

Imagine a new player in a trial nova, who uses controlled firing bursts until he reaches heat threshold. Then he alphas and intentionally kills himself leaving you mortally wounded. A high ranked Elo players kills you. How is that one scored?

There are an endless series of questionable scenarios.


There is no benchmark challenge with zero variables. There isn't an obstacle course to destroy mechs, there isn't an obstacle course to avoid laser/ppc/lrm fire. Any assigned Elo rating is quite honestly arbitrary and worthless because you can't be sure who beat you.

Elo was one of the features that did not help this game. It was introduced right after the sting of ECM had lessened quite a bit.

Elo makes the game less fun for skilled players because as you get better you are more likely to get members on your team who might as well not be there.

Elo makes the game more fun for bad players because they never go against people who will decimate them.

Having Elo in the game is like playing any racing sim with rubber band AI. Run an enemy into the wall and he falls about half a lap behind your car. And then about 10 seconds later he's right behind you again. If you're driving the fastest car in the game (think something like F-zero where the cars have different stats) there's no way in hell that should be possible. But the AI cheats and does it anyway.

MischiefSC, on 24 September 2014 - 07:39 PM, said:

The real problem with Elo is that it doesn't work like 'scaling enemies' in Skyrim and the like - where you are level 20 so all enemies are level 15, so you win all the time.

Elo achieves a "running in place" scenario where you never feel you get any better no matter how well you perform. That's a terrible mentality for a player if you want to keep people playing/paying. You're level 20 so the enemies on a good match are 19,20 and 21. But if you have to wait too long it will pull from 17-23 (take the idea as an example rather than concrete numbers).

In the end Elo gives less fair matches in terms of "randomness". Most people would be better off playing more games with a completely random match maker. Their math for Elo as far as I am concerned is basically voodoo. They don't tell you what your score is and they don't tell you what the score is for anyone else in the game (even after it is over). There's no way to know if your Elo went up or down. For all you know there's a computer program that randomly assigns you a score.

Edited by Glythe, 24 September 2014 - 08:19 PM.

[Sandpit]

MischiefSC, on 24 September 2014 - 07:39 PM, said:

Elo isn't a forced 50% win rate. If your win rate is only 50% then you're not improving. Many will find themselves in to 55-60% win rate range.

Someone above gave some incorrect information so let's clarify that -

In your first 25 matches your Elo is 1,000.

After that you go to the default of 1300.

Most matches vary from team to team by less than 50 pts in Elo.

Most matches have a variance of high to low of a few hundred points (not 1400). Is it possible to go to 1400? Sure, but nobody on this thread (save possibly Adiuvo) would ever conceivably be in such a situation because almost everyone on this thread is probably sub-1800.

The real problem with Elo is that it doesn't work like 'scaling enemies' in Skyrim and the like - where you are level 20 so all enemies are level 15, so you win all the time.

Reality is that players are never as good as they think they are. Ever.

Elo works just fine for this sort of environment. That's the biggest problem with it for a lot of people.

I agree for the most part. My only issue with Elo is the wsy it's implemented here. You can personally have a fantastic game but still drop in Elo because your team loses.

I think it's also notable that you see far less complaints about Elo from players who primarily drop in large (er) groups.

Solo players have to rely on 11 random people for their Elo ratings. They also tend to not realize that "average" Elo means just that. Average.

Elo isn't "forcing" anything in regards to w/l. It simply puts your team against another team with a similar team El9.