277 replies to this topic

[PhoenixFire55]

Hauser, on 06 July 2013 - 03:30 AM, said:

You can use the raw data sets for that. Mechs per team per match were recorded. But you're better of if you send a mail to hammer for the raw data sets in a spreadsheet.

I might do just that but not before I gather my own data so I have smth to compare.

Hauser, on 06 July 2013 - 03:30 AM, said:

I didn't wanna say this, but the ones where I'm not playing with you tend to go allot better. The feeling is probably mutual.

Its mostly crap for me no matter who I play with, so not really.

Hauser, on 06 July 2013 - 03:30 AM, said:

I hope it soon will be possible to setup private matches and just play for fun against friends.

Hell yeah.

[Sug]

Hauser, on 05 July 2013 - 06:09 AM, said:

Again anecdotes are not data. And to make it worse you're only describing parts of both teams. Last time we did a systematic check of what people were dropping in the weights weren't that far off and the extra tonnage did not increase the chance to win.

http://mwomercs.com/...19#entry2220919

And lastly, as much as you see complaints on the forums about the matchmaker, the weights are roughly even (the maximum was 135 tonnage difference), and the win rate of the heavier team is roughly 50%, which seems to actually suggest it's doing something right.

Not sure that means anything. Weight will only matter in games where all the players are equally skilled. Which is not what the current matchmaker is trying to do. No data is really going to tell you anything when pugs are arbitrarily put on low Elo teams to give them a 50/50 win loss ratio.


In my hunchback i'm worth at least 3 trial Atlases.

[Hauser]

Sug, on 06 July 2013 - 09:38 AM, said:

Not sure that means anything. Weight will only matter in games where all the players are equally skilled. Which is not what the current matchmaker is trying to do. No data is really going to tell you anything when pugs are arbitrarily put on low Elo teams to give them a 50/50 win loss ratio.

In my hunchback i'm worth at least 3 trial Atlases.

The effectiveness of a player is a combination of his skill and equipment. That is very true. I was countering the specific assertion that games were unbalanced in tonnage and that the extra tonnage was a major advantage. The statistics gathered show neither is true.

I'm not sure what you mean by your last sentence. The purpose of the match maker is to provide people with a fair match by matching them against people of equal skill and balance out the tonnage between teams. As such green players should be playing green players, experienced players against experienced players. As a result of being matched against equals, people should trend towards a 1.00 win loss ratio. The ratio is a result of good matchmaking, not the actual goal.

[PhoenixFire55]

Hauser, on 06 July 2013 - 10:01 AM, said:

The ratio is a result of good matchmaking, not the actual goal.

It is an actual goal, because game needs money and that means keeping most people happy including terribads, and thats why you need to give them 50% win ratio as well.

[Sug]

Hauser, on 06 July 2013 - 10:01 AM, said:

I'm not sure what you mean by your last sentence. The purpose of the match maker is to provide people with a fair match by matching them against people of equal skill and balance out the tonnage between teams. As such green players should be playing green players, experienced players against experienced players. As a result of being matched against equals, people should trend towards a 1.00 win loss ratio. The ratio is a result of good matchmaking, not the actual goal.

Hard to define "green" when every new player starts with an Elo of 1300.

Can't quote cause it's locked but:


"CLARIFICATION:


At the start of a match, all player's on ONE team have their Elo ratings totaled and divided by 8 (max players). You may realized that this is simply the AVERAGE of a team's total Elo. Team 1's average and Team 2's average are then used to calculate the probability of win (as per the formulas above). If Team 1 beats Team 2, then the appropriate math as above is applied to each player using the probability score calculated by the team averages. "


http://mwomercs.com/...79-matchmaking/



Maybe be out of date by now (Phase 3 of Matchmaking) correct me if i'm wrong please.

It takes the average of the team.

Which team will most likely win? Purely pugging.

Team 1

Player1 = 2800
Player2 = 1100
Player3 = 1000
Player4 = 1100


Team 2

Player1 = 1500
Player2 = 1600
Player3 = 1400
Player4 = 1500


Team 3

Player1 = 2100
Player2 = 1300
Player3 = 1250
Player4 = 1350


All three teams have an average Elo score of 1500. The highest score you can have as a player is 2800. All players start with 1300.

Team 1 is made of a Player with the highest Elo possible, balanced out by low Elo players.
Team 2 is made of 4 players that consistently win.
Team 3 is made of an above average player and 3 players that are either new players or below average.

Team 2 probably has the advantage as these are all solid players.

Unfortunately the Matchmaker currently sees these teams as even and balanced.

Just some examples. I know the Mm probably works "ok" "most" of the time but taking the average score is not a great way to balance pugs.

I know stats are almost meaningless but after a year of playing, only pugging, my w/L ratio is 47/53 and my KD is 1.76. I consider myself an average or above average player.

And yet I'm still matched with people in trial mechs that are asking how to group weapons and how to move forward (really). This makes me think my Elo isn't far away from the 1300 that all new players start with.

And worse, typically new players are in the Assault class trial mechs which can further imbalance matches because weight is not a consideration for matching pugs.

The graph under the spoiler in this post http://mwomercs.com/...25#entry2265225

shows that most players are grouped around the 1100 - 1400 Elo ranking.

So if most players are around 1100-1400, and everyone starts at 1300, and we're matched according to team Elo averages....

How can the better pug players advance in rank? If everyone is ranked average and the scores are averaged and the teams are averaged then yes, we are being artificially held at a 50% (AVERAGE) win ratio so everyone (new players) is happy.

See sig.

Edit: yeah I know. "join a group"

Edited by Sug, 06 July 2013 - 12:03 PM.

[Hauser]

PhoenixFire55, on 06 July 2013 - 10:34 AM, said:

It is an actual goal, because game needs money and that means keeping most people happy including terribads, and thats why you need to give them 50% win ratio as well.

Nobody is given the 50% win ratio. You're only given the match makers best attempt at two teams of comparable skill. The expected outcome is a 50% win ratio. These two are not the same.

[PhoenixFire55]

Hauser, on 06 July 2013 - 11:39 AM, said:

Nobody is given the 50% win ratio. You're only given the match makers best attempt at two teams of comparable skill. The expected outcome is a 50% win ratio. These two are not the same.

You know perfectly well what I mean.

[Sug]

Also, not sure if this means anything, but when I, as a pug, join a match, my name is the first one that pops up. Sometimes I will be alone for almost 15 seconds before my team mates slowly filter in.

On a brand new account, my whole team appears with me at the same time.

It makes me think that as a "new" player my Elo allows me to fit on almost any team but my real account requires some math to make the team balanced.

Regarding http://mwomercs.com/...25#entry2265225,

I'd want to graph to look like a wave. Clusters of low skill, average skill, and above average skill players, all playing against each other within their skill level, with thin lines of people moving slowly up or down to the next group.

Don't artificially inflate the Elo of low skill players by averaging out the teams.

Right now as a pug I can almost accurately predict whether i'll win a match or not by flipping a coin at my desk.

Edited by Sug, 06 July 2013 - 11:58 AM.

[Hauser]

Sug, on 06 July 2013 - 11:25 AM, said:

Hard to define "green" when every new player starts with an Elo of 1300.

As of 21 may new players start with a 1100 Elo rating.

http://mwomercs.com/...23-21-may-2013/

Quote

Match Making Update

- All new players starting the game will have 2 Elo scores assigned to them.
- New Player Elo = 1100 (N-Elo)
- Standard Player Elo = 1300 (S-Elo)
- Any new player (has 25 games or less in their history) will be using their N-Elo score to be matched with other players.
- The outcome of the game will affect both their N-Elo score and their S-Elo score.
- When the player has completed their 25th game, the N-Elo score is eliminated and the player from that point on will use their S-Elo score for match making.
- This change makes the initial experiences of a new player much easier and prevents the player from playing players with numerous matches under their belts.

Sug, on 06 July 2013 - 11:25 AM, said:

Can't quote cause it's locked but:

"CLARIFICATION:


At the start of a match, all player's on ONE team have their Elo ratings totaled and divided by 8 (max players). You may realized that this is simply the AVERAGE of a team's total Elo. Team 1's average and Team 2's average are then used to calculate the probability of win (as per the formulas above). If Team 1 beats Team 2, then the appropriate math as above is applied to each player using the probability score calculated by the team averages. "


http://mwomercs.com/...79-matchmaking/

The part you quoted describes how the teams elo rating is determined after two teams have been assembled. It does not describe how the teams are assembled. And this is the thing, the match maker does not try to average out the Elo on both teams. Rather it looks for players near a certain Elo value. When it can't find people near the value it wants it becomes less picky.

Now this is not to say that the examples you mention can't happen. When it happens it is not because the match maker decides you are too good and need to be brought down a peg to balance the team. Rather they happen because there aren't enough players dropping at that exact moment to get a match going. So it's either that or no match at all.

Sug, on 06 July 2013 - 11:25 AM, said:

And worse, typically new players are in the Assault class trial mechs which can further imbalance matches because weight is not a consideration for matching pugs.

The match maker takes both Elo and weight into account when setting up matches. It takes weight into account in the same fashion as Elo. It looks at what is in the queue that is nearest to what it needs to even out the teams weight. When it can't find anything it gets less picky about what it needs.

Sug, on 06 July 2013 - 11:25 AM, said:

I know stats are almost meaningless but after a year of playing, only pugging, my w/L ratio is 47/53 and my KD is 1.76. I consider myself an average or above average player.

The KD is nice but you've only played some 100 games since stat tracking. I don't know if you've played much more before stat tracking was implemented, or how those 100 games are distributed over your mechs that but that's barely enough to grind out your elite efficiencies and upgrade your mech(s). If you are indeed better then average you might be held back by your gear.

Sug, on 06 July 2013 - 11:25 AM, said:

How can the better pug players advance in rank? If everyone is ranked average and the score are averaged and the teams are averaged then yes, we are being artificially held at a 50% win ratio so everyone (new players) are happy.

They will advance simply by being better. The way you can get better is by recognizing your mistakes. Your team is full of equal players, and they make mistakes. They make mistakes that can cost you the match. But so do you. You're at the same Elo afterall. If you stop making mistakes then there will be 7 guys on your team that make mistakes against 8 on the other. You might not win every game but on the long run you should go up.

But a more important question is this. Do you want to improve your rank? And why? It isn't visible anywhere and the game won't become easier when it goes up. If you're having fun right now, forget about Elo, forget it exists, for all you care it's a tool to get you into a somewhat balanced match. If you join a unit, join one for sake of playing with other people, not because you want to win.

[Hauser]

PhoenixFire55, on 06 July 2013 - 11:42 AM, said:

You know perfectly well what I mean.

Yet you fail to articulate it; you don't like a fair fight and resent PGI for taking your stomping grounds away.

[Sug]

Hauser, on 06 July 2013 - 12:55 PM, said:

As of 21 may new players start with a 1100 Elo rating.

http://mwomercs.com/...23-21-may-2013/

Good to know ty.

Hauser, on 06 July 2013 - 12:55 PM, said:

The match maker takes both Elo and weight into account when setting up matches. It takes weight into account in the same fashion as Elo. It looks at what is in the queue that is nearest to what it needs to even out the teams weight. When it can't find anything it gets less picky about what it needs.

The last thing I read said that weight is only taken into account for groups of 2-4. It doesn't apply to 8 mans or pugs. My info might be out of date.

Hauser, on 06 July 2013 - 12:55 PM, said:

The KD is nice but you've only played some 100 games since stat tracking. I don't know if you've played much more before stat tracking was implemented, or how those 100 games are distributed over your mechs that but that's barely enough to grind out your elite efficiencies and upgrade your mech(s). If you are indeed better then average you might be held back by your gear.

Where did the 100 come from? Just wondering if there's a counter somewhere. In my stats i have 3012 games played, since the last wipe, and in the "Game Mode" section of my Stats is says i've got about 1,000 matches, almost balanced between Assault and Conquest.

(also seeing that i've been a member since November 2011 killed me a bit inside.)

Hauser, on 06 July 2013 - 12:55 PM, said:

They will advance simply by being better. The way you can get better is by recognizing your mistakes. Your team is full of equal players, and they make mistakes. They make mistakes that can cost you the match. But so do you. You're at the same Elo afterall. If you stop making mistakes then there will be 7 guys on your team that make mistakes against 8 on the other. You might not win every game but on the long run you should go up.

Its too long a run. Your Elo only goes up significantly if the matchmaker decides that, based on your teams average Elo, you have low odds of winning. A victory against even odds rewards you very little. And a loss also would drop you very little.

So if you get almost no reward/penalty from wining/losing against an equal team, and the match maker tries to give you an equal team......

yeah.

Bingo bango. no one moves very far.

Hauser, on 06 July 2013 - 12:55 PM, said:

But a more important question is this. Do you want to improve your rank? And why? It isn't visible anywhere and the game won't become easier when it goes up. If you're having fun right now, forget about Elo, forget it exists, for all you care it's a tool to get you into a somewhat balanced match. If you join a unit, join one for sake of playing with other people, not because you want to win.

While I do appreciate a good game, even a loss (some of my favorite games have been close losses) what I don't appreciate are the losses or wins that are 0 - 8 blow outs. That's not a good game.

Right now I want to advance in rank so I'll stop getting team mates in trial mechs that don't know how to target or how to group weapons. Right now if I lose a game it's typically because my team lost the game, not that the other team won.

Basically though, I play to win. Sorry but 50% is a failing grade in almost every school. I know I'll never have a 90% win ratio but there's no way i'm going to accept 50% when CW and real game play starts. (and I join a unit)

Why would I play a game I lose half the time?

MWO: The game that combines the excitement of coin flipping with the thrill of sitting there.

I'm not going to take that and I know most other serious players won't either.

Hey that might be a new sig.

Edit: just read those may 21st patch notes. I think the "secret" new player Elo ranking is a good start but after 25 games they get thrown to the wolves.

Yeah, no one on their 26th game should be considered an experience player. Try a couple hundred games at least PGI.

Edited by Sug, 06 July 2013 - 01:35 PM.

[Sug]

Also. These graphs they gave us to show Elo distribution....



....why do they think the blue one is better?

First off, A) don't change the scale of two graphs you're trying to compare....

http://www.amazon.co...f/dp/0393310728 is still a good read.

B.) The first graph shows the distribution after matchmaker Elo adjustments, with what they're happy with. All that shows is yes, there will be a wider distribution around the mean(?) with more data samples if the data is not allowed to deviate too far from the mean. Woot.

The second graph shows before the matchmaker adjustments.. What I see is, a cluster of low to average Elo players, all playing against each other, then a slow line of people slowly working their way up the ranks to the smaller cluster of high Elo players.

That's how it should be!

Edited by Sug, 06 July 2013 - 01:50 PM.

[Hauser]

Sug, on 06 July 2013 - 01:23 PM, said:

The last thing I read said that weight is only taken into account for groups of 2-4. It doesn't apply to 8 mans or pugs. My info might be out of date.

No you're mistaken.

The game has two queues. Eight man and regular. The eight man queue doesn't care about anything afaik. There aren't enough people playing so most groups determine before hand what they use to drop against each other.

In the regular queue the match maker looks for a target elo and tries to balance weight. Groups of up to 4 players can drop in the regular queue and are treated as having the average elo of their group.

Sug, on 06 July 2013 - 01:23 PM, said:

Where did the 100 come from? Just wondering if there's a counter somewhere. In my stats i have 3012 games played, since the last wipe, and in the "Game Mode" section of my Stats is says i've got about 1,000 matches, almost balanced between Assault and Conquest.

You said your w/l was 47/53. I assumed you copied if off your profile.

Sug, on 06 July 2013 - 01:23 PM, said:

Right now I want to advance in rank so I'll stop getting team mates still earning their cadet bonus. Right now if I lose a game it's typically because my team lost the game, not that the other team won.

You might find this interesting:



The video talks about Elo hell which is a term used to describe having such a low Elo that you're only playing with bad players on your team and thus can't win. It is essentially what you are complaining about. The tone is kinda harsh but he's trying to get the point across that you and not your team is the problem. The other team is just as bad.

Sug, on 06 July 2013 - 01:23 PM, said:

Basically though, I play to win. Sorry but 50% is a failing grade in almost every school. I know I'll never have a 90% win ratio but there's no way i'm going to accept 50% when CW and real game play starts. (and I join a unit)

Why would I play a game I lose half the time?

That's just a matter of perspective. Under an Elo system you win loss ratio doesn't show you how well you do against a random player, it doesn't show your skill compared to the overall population. Rather it shows you how well you'd compare against your peers. In every system that uses Elo you win loss ratio will eventually trend towards 50%. Only at the edges you might have a higher or lower ratio because there are no peers to match you with.

For a real world comparison don't look at the grade, but rather your position in the grade curve. Being in the 50th percentile (right in the middle) means nothing. You can be in the 50th percentile of CalTech or the University of Wyomigen.

So right now, you can't play to win. You can't even tell if you're winning. But life isn't too different.

Edited by Hauser, 06 July 2013 - 02:59 PM.

[Hauser]

Sug, on 06 July 2013 - 01:48 PM, said:

....why do they think the blue one is better?

First off, A) don't change the scale of two graphs you're trying to compare....

http://www.amazon.co...f/dp/0393310728 is still a good read.

B.) The first graph shows the distribution after matchmaker Elo adjustments, with what they're happy with. All that shows is yes, there will be a wider distribution around the mean(?) with more data samples if the data is not allowed to deviate too far from the mean. Woot.

The second graph shows before the matchmaker adjustments.. What I see is, a cluster of low to average Elo players, all playing against each other, then a slow line of people slowly working their way up the ranks to the smaller cluster of high Elo players.

That's how it should be!

I'm afraid not.

There was a bug where a player with higher elo winning from a player with lower elo would get the reward the lower player would get when defeating the higher player and vice versa. What you're seeing is the player population with a w/l of 1.00 or greater rapidly running away into the elo cap.

The blue graph shows what appears to be a normal distribution. For random natural occurrences such as the skill of mwo players the normal distribution is somewhat the expected outcome. These graphs were made some time in the past, I expect that the graph will have a slightly longer tail on the right hand side like this graph from League of Legends.

[PhoenixFire55]

Hauser, on 06 July 2013 - 01:19 PM, said:

Yet you fail to articulate it; you don't like a fair fight and resent PGI for taking your stomping grounds away.

I haven't had a fair fight ever since elo kicked in. And you can keep your stomping bullshyt to yourself. 90% games I played back there and play now are solo.

[PhoenixFire55]

Sug, on 06 July 2013 - 01:23 PM, said:

Its too long a run. Your Elo only goes up significantly if the matchmaker decides that, based on your teams average Elo, you have low odds of winning. A victory against even odds rewards you very little. And a loss also would drop you very little.

So if you get almost no reward/penalty from wining/losing against an equal team, and the match maker tries to give you an equal team......

yeah.

Bingo bango. no one moves very far.

Thats what I said right away, but Hauser and others told us 'its gonna average itself up'. Well after 4000 games I still see trial mechs who can't shoot straight in my games.

Sug, on 06 July 2013 - 01:23 PM, said:

Basically though, I play to win. Sorry but 50% is a failing grade in almost every school. I know I'll never have a 90% win ratio but there's no way i'm going to accept 50% when CW and real game play starts. (and I join a unit)

Why would I play a game I lose half the time?

Because crybabies who'd otherwise lose 90% of their games think elo matchmaker is 'fair'. You work your butt off and do 800 dmg in your zone while they can noob-kill 8 mechs standing still in their games and do same 800 dmg and get same exp and c-bills. Seems really fair to me.

Sug, on 06 July 2013 - 01:23 PM, said:

Edit: just read those may 21st patch notes. I think the "secret" new player Elo ranking is a good start but after 25 games they get thrown to the wolves.

Yeah, no one on their 26th game should be considered an experience player. Try a couple hundred games at least PGI.

Not the case. I've asked people in trials I see how many games they've played. Typical answer is 2-3.

[PhoenixFire55]

Hauser, on 06 July 2013 - 02:56 PM, said:

The blue graph shows what appears to be a normal distribution. For random natural occurrences such as the skill of mwo players the normal distribution is somewhat the expected outcome.

The blue graph shows one thing. Elo scores. Elo is based off win/loss ratio and thats what it shows. It shows that 95% people have a win loss ratio between 46%-54%. It has NOTHING to do with skill, because now matter how good/bad you are you can always have same good/bad player on the other team to negate your effect just as you can have a really horrible/excellent team that cancells/covers for you.

The only 5% of people outside the main elo zone are two types...
1) people who ONLY play in really good groups of 4 and thus win like 80-90% games by killing everything themselves
2) people who start doing smth like TK'ing right from the start of the game, so that their team loses ~90% of games just because its 2-3 mechs short to begin with.

Those graphs are just another PGI fail tho. No values on Y-axis means you can't tell how many players actually do have high/low elo scores. The 'height' of those elo edges can be 1 player for all we know.

[Hauser]

You know, I can't wait for Elo rankings to become public. I could get payed to play MWO and boost peoples rating out of Elo hell.

[Znail]

The short story is that the game is currently at it's worst when it comes to getting even matches. It's far below how it was back before Elo. This may sound strange, but it's not so strange if you think about it. The stage 4 Elo pretty much eliminates the diffrences between player Elo by bunching up everyone in the middle, so skill has little impact on the match maker now and thus the Elo doesn't help at all. But we still get the downside of the Elo by having one team randomly getting more or less weight, wich depending on the map gives advantage to one side. Then we have the new maps that are huge and really punnish anyone with short range weaponry, thus adding a further random factor that can greatly favor one team. The end result is that getting an even match is so rare that I can't even remember when it happened last. It makes the game rather unfun with alternating stomps with getting stomped.

The Stage 3 was pretty good actually and the one stage that was actually ahead of the pure random match maker.

Edited by Znail, 07 July 2013 - 11:24 AM.

[Enigmos]

WardenWolf, on 19 April 2013 - 02:02 PM, said:

My biggest concern left with matchmaking is the experience of new players. The light blue on the Elo chart is people with less than 50 games - aka new players. They are right there in the middle of the distribution, which means you are dropping new players in with 'average' veteran players... which means about half of those new players (those destined for lower Elo eventually) will struggle and may bail on the game if they have bad experiences early-on (matched against better players + stuck in trial mechs).

I would still suggest having the newer players come in at a lower Elo area, rather than right in the middle. I'd be interested in seeing how that might help the new player experience, and what it might do to the distribution graph as well.

I believe WardenWolf's observation is being bourn out just now with what appears to be a fairly massive influx of new players. If we do not drop in 4-mans we are being massacred over and over again by those who do team up in four mans simply because with complete neophytes as our teammates there is absolutely no sense of tactics, no coordination no skill in aiming... it is a tragedy. It's like 'What's a flank?'