156 replies to this topic

[Mystere]

Mudhutwarrior, on 12 August 2016 - 09:04 AM, said:

What the tryhards always leave out in their arguments of 8x10 full color glossy's and fact and figures is this game is based on sales.

In sales one fact rises above all the din and that is perception is reality to the customer. No amount of facts, figures, shaming or put downs will alter the perception. In fact it will only reinforce that.

Hmm. That kind of reasoning sounds just like this guy:



I hope he wins.

[Dread Render]

Lostdragon, on 12 August 2016 - 09:52 AM, said:

The reality is players have different skill levels. If the MM was really unfair and horrible as people want to paint it them the mean and median W/L would not be 1, it would be significantly lower. In this case the customers saying the MM is terrible and doesn't work are demonstrably wrong. I am sure it would help PGI's sales is people could win all the time but that is just not realistic.

If I worked for PGI I would definitely consider the stats of the QP matches to be acceptable and as a reasonable player I do as well. I am definitely not a white knight, but the constant griping about the MM is just a distraction from the real issues the game has and should be ignored

Yes I agree the median W/L should be 1, I don't think it is, at least not for me and the many others who agree with me.
That is what we are talking about.
But if you agree that it should be one then by default you agree that You personally can not have more than about 1/12th effect on it in any one game.
Therefor if you have a W/L much greater or lower than 1 there is likely a problem with the MM.
Are we Learning Yet?
So here we are... back to square one where some people (with a good W/L ratio) want to believe it is because they are Soooo Gooood and are just superior players and every one else knows better.
Enjoy your rose colored glasses.

--- edit---
"griping about the MM is just a distraction from the real issues"
What Real issues. MM has the most effect on every player every game.
It should be the 1# issue.

Edited by Dread Render, 12 August 2016 - 10:07 AM.

[Ghogiel]

Johnny Z, on 12 August 2016 - 08:54 AM, said:

I call out aim botters in matches to. Guess what? I don't see them in game again. Know why? Because they were botting.

Although it is rare I can be sure enough to call them out in a match. But it has happened. If I am really sure I ask my entire team to send a complaint.

You've never got anyone banned for botting in this game.

[Mister Blastman]

Dread Render, on 12 August 2016 - 10:03 AM, said:

Yes I agree the median W/L should be 1, I don't think it is, at least not for me and the many others who agree with me.
That is what we are talking about.
But if you agree that it should be one then by default you agree that You personally can not have more than about 1/12th effect on it in any one game.
Therefor if you have a W/L much greater or lower than 1 there is likely a problem with the MM.
Are we Learning Yet?
So here we are... back to square one where some people (with a good W/L ratio) want to believe it is because they are Soooo Gooood and are just superior players and every one else knows better.
Enjoy your rose colored glasses.

--- edit---
"griping about the MM is just a distraction from the real issues"
What Real issues. MM has the most effect on every player every game.
It should be the 1# issue.

Uh, he already proved to you that the median win/loss ratio is 1 by combing through the data of 34,000 players of season 1's leaderboard.

So what exactly is your point?

Players achieve a greater than 1:1 win/loss ratio in two ways:

1. They play in the Group Queue. These players are pretty easy to spot as their win/loss is often higher than 2:1 while their match score is lower.

2. They are damn good solo players. These can be identified with a positive win/loss ratio, a high average match score and a good KDR. (KDK3 use aside--though, some great players use KDK3 and the result is demolishing)

Good players do exist. Anyone can be one, if they want to work at it.

Edited by Mister Blastman, 12 August 2016 - 10:15 AM.

[Lostdragon]

Dread Render, on 12 August 2016 - 10:03 AM, said:

Yes I agree the median W/L should be 1, I don't think it is, at least not for me and the many others who agree with me.
That is what we are talking about.
But if you agree that it should be one then by default you agree that You personally can not have more than about 1/12th effect on it in any one game.
Therefor if you have a W/L much greater or lower than 1 there is likely a problem with the MM.
Are we Learning Yet?
So here we are... back to square one where some people (with a good W/L ratio) want to believe it is because they are Soooo Gooood and are just superior players and every one else knows better.
Enjoy your rose colored glasses.

I literally just showed you the mean and median w/l is 1.0 then you go on to say you believe it is not and lecture me about stats and game impact. I am astounded.

The people who have high w/l also have good KDR and match score, as shown in the thread Blastman linked. Guess what that means? The better you are the more likely you are to have a good w/l, that is not conjecture or epeen stroking, that is a statistically proven reality.

There has never been and bnever will be a PVP game where everyone has a 1:1 w/l ratio. Some people are good and win more. Some people are bad and lose more. If the average and median are around 1.0, though, that means the MM is working.

The argument that every person has equal impact on the game is just stupid. If you believe that is true then I challenge you to a 2v2 and I get Schopenhauer on my side and you get the player with the worst w/l. Players can and do carry matches and have the W/L ratio to prove it. I carry matches sometimes, but the best guys do it a lot and have the best W/L because of that.

[Mawai]

Mister Blastman, on 12 August 2016 - 06:25 AM, said:

That's a pretty bad win/loss ratio on your main account. If you are below 1.0 win/loss then you need to be trying significantly harder to help your team. The way the matchmaker is structured, it aims to put all players at 1/1 win/loss. To go below 1/1 means you're doing things that hurt your team.

Wow. The matchmaker does NOT NOT NOT aim to put all players at a 1/1 win/loss. Not at all. Not even the slightest. It is one of those stupid myths that float around ... the Matchmaker DOES NOT CARE what your results are.

However, IF the matchmaker is working correctly ... i.e. you get perfectly balanced matches every time which is the IDEAL and GOAL of the matchmaker ... correct? IF it works correctly THEN your chance of being on the winning side in any specific match are EXACTLY 50% BECAUSE it is a balanced match.

Understand? If your win/loss is close to 1:1 this is because you are being matched against players of equivalent skill and as a result you end up with a 1:1 W/L BECAUSE the matchmaker is working ... NOT because the matchmaker cares about your win/loss at ALL.

So ... what is happening to the OP? His main toon is in tier 1 with a high PSR. His alt is in tier 2 with a lower PSR. When the matchmaker creates a balanced match ... his tier 1 toon is grouped with other folks rated tier 1 with a PSR near his while the tier 2 is grouped with lower rated folks.

If he was being placed into balanced matches that reflect his skill then the win/loss on the tier 1 toon would be around 1/1 while the tier 2 toon would tend to have a win/loss > 1 as the toon moves up the ranks.

However, what may be broken is that the OPs PSR does not reflect his skill at the game ... it tends to reflect his experience over a large number of matches (as long as performance is "reasonable" in terms of match score). The matchmaker uses the PSR to form matches among equally "skilled" or at least "rated" players. In theory, if your PSR is too high then you will lose more than win and your PSR drops ... at least it works that way in an Elo ranked game ... however PSR is not conservative and has an upward bias ... you have to do really badly to drop your PSR significantly ... just a bit bad and it will stay steady ... average and it will rise.

From the sounds of it, the OPs PSR on his main toon has moved into the top of tier 1 where there are a variety of players including the best ones. As a result, he is matched at that skill level but his W/L can't keep up because the PSR value is exaggerated. At the moment, the tier 2 toon gets better matches since it is more often matched with players of a comparable skill level.

This "problem" will only continue to get worse as PSR gets more and more broken over time. This issue was pointed out when PSR was first introduced ... but PGI doesn't really care about problems that will take a year to 18 months to develop ... unfortunately for them the problem has become noticeable and will only get worse until they correct the issues with PSR. Unfortunately, I suspect that this will mean worse and worse matches over time for those in tier 1 as the actual skill level of tier 1 rated players becomes wider.

[Dread Render]

Lostdragon, on 12 August 2016 - 08:50 AM, said:

I downloaded this data and did a quick couple of formulas. For people with at least 10 games (32049 players) 50.57% are at 1.0 or better. Only 5.79% are at 0.5 W/L or lower. About 22% are lower than .75 W/L.

The average W/L is 1.09 and median is 1.0.

So looks like all the crying about the MM can just be dismissed as it is indeed only whining based on skewed perceptions.

"22% are lower than .75 W/L."
to me that also means that about 22% are over 1.25 W/L
that's 44% are to high or to low.
I don't think that is good enough.
I think there are areas in the MM where people seem to get stuck in a rut so to speak and find them self in games one after the other where you either wipe out the enemy or they wipe out your team.
its a flaw in the MM
I think the flaw is significant enough to require attention.

[Paigan]

Poro.

[Lostdragon]

Dread Render, on 12 August 2016 - 10:36 AM, said:

"22% are lower than .75 W/L."
to me that also means that about 22% are over 1.25 W/L
that's 44% are to high or to low.
I don't think that is good enough.
I think there are areas in the MM where people seem to get stuck in a rut so to speak and find them self in games one after the other where you either wipe out the enemy or they wipe out your team.
its a flaw in the MM
I think the flaw is significant enough to require attention.

Take some stat classes and learn about distributions then we will rehash this. What you want is unrealistic.

[Mister Blastman]

Mawai, on 12 August 2016 - 10:30 AM, said:

Wow. The matchmaker does NOT NOT NOT aim to put all players at a 1/1 win/loss. Not at all. Not even the slightest. It is one of those stupid myths that float around ... the Matchmaker DOES NOT CARE what your results are.

If it does not then why over 34,000 players analyzed is the median win/loss 1:1 and the average 1.09:1?

The rest of your post is pretty good and makes solid points, btw.

[Novakaine]

2 Shooters.
2 Average Players.
2 Pro Players.
3 Utter iredeemiable gits
1 Drunken sod.
The average team complement.

[El Bandito]

Dread Render, on 12 August 2016 - 10:36 AM, said:

"22% are lower than .75 W/L."
to me that also means that about 22% are over 1.25 W/L
that's 44% are to high or to low.
I don't think that is good enough.
I think there are areas in the MM where people seem to get stuck in a rut so to speak and find them self in games one after the other where you either wipe out the enemy or they wipe out your team.
its a flaw in the MM
I think the flaw is significant enough to require attention.

Bads and goods exist. Why should everyone have 1:1 WLR? The MM does not enforce 1:1 WLR in the first place. Those who put no effort in playing well will be stuck with low WLR for a long time, like those Bronze Elo League players. On the other hand, those who do their best in their matches will get higher than 1.0 WLR over time. Just how things work. Age of course, matters as well.

Edited by El Bandito, 12 August 2016 - 10:43 AM.

[Dread Render]

El Bandito, on 12 August 2016 - 10:42 AM, said:

Bads and goods exist. Why should everyone have 1:1 WLR? The MM does not enforce 1:1 WLR in the first place. Those who put no effort in playing well will be stuck with low WLR for a long time, like those Bronze Elo League players. On the other hand, those who do their best in their matches will get higher than 1.0 WLR over time. Just how things work. Age of course, matters as well.

The point of the MM is to create even Teams.
That means if it works the average W/L for every person should be around 1 to 1
if your W/L deviates much off that, it indicates a flaw in the MM.
... and I had a stats in college.
I'm not the one misunderstanding things here.

You may be a great player... that means your Skill Rating should be High.. That's Great!!! Cool!! I can dig it.
But that is not what the OP is talking about. The MM is the problem.

Edited by Dread Render, 12 August 2016 - 10:54 AM.

[El Bandito]

Dread Render, on 12 August 2016 - 10:51 AM, said:

The point of the MM is to create even Teams.
That means if it works the average W/L for every person should be around 1 to 1
if your W/L deviates much off that, it indicates a flaw in the MM.
... and I had a stats in college.
I'm not the one misunderstanding things here.

You may be a great player... that means your Skill Rating should be High.. That's Great!!! Cool!! I can dig it.
But that is not what the OP is talking about. The MM is the problem.

More specifically, PSR system needs some work, such as getting rid of upwards bias. Too many undeserving players are diluting the higher tiers.

[Mister Blastman]

Dread Render, on 12 August 2016 - 10:51 AM, said:

The point of the MM is to create even Teams.
That means if it works the average W/L for every person should be around 1 to 1
if your W/L deviates much off that, it indicates a flaw in the MM.
... and I had a stats in college.
I'm not the one misunderstanding things here.

If you had stats in college then you'd understand that an average is comprised of a set of numbers in a range that has a minimum and maximum value...

And right now the average win/loss ratio is 1.09 (most likely due to group queue disparity), the median is 1:1

-and-

You can get 1.09 with a set of numbers ranging from 0.2 all the way up to 2.2 with a undeclared mix between those two poles.

Does that make sense?

This means that there WILL be some players with an average win/loss ratio of 1.3, 1.4, 1.5, 1.6 and even 2.0 and this is not because the matchmaker is broken, it is because they are good at the game.

The statistics support this.

You will never have a matchmaker that fits everyone in at 1:1 and if you do somehow manage to create one, you will see a large exodus of the player population because frankly good players don't like to lose half the time and bad players like to think they should win more than half the time.

El Bandito, on 12 August 2016 - 10:56 AM, said:

More specifically, PSR system needs some work, such as getting rid of upwards bias. Too many undeserving players are diluting the higher tiers.

This for sure. If the system was weighted on a moving bell curve it would help weed out the bads and get rid of the "progress bar."

[Mister Blastman]

PORONOPAPOS, on 12 August 2016 - 11:05 AM, said:

in my alt account I played 123 matches with Cheatah and around 150 with Battlemaster.

In main account played 74 matches in assault (including those of the event) and 14 with lights, also had some matches with heavies and meds.

Being T1 from last early December, I basically see same names all this time, although I feel their is some "rotation" of the players. I see some very often for a period , then some other names for some days and so on....As already said my only problem is that after late June more and more of the defeats are stomp ones, only in my main account

Perhaps you have met your "maximum bench" threshold.

You do well in Tier 2, but are doing poorly in Tier 1. Now I realize you have been Tier 1 since December but consider this, PSR is nothing more than a "progress bar" and during that time between December and June, many players have "progressed up" from lower tiers to higher tiers, thus altering the composition you face as tier 1. Now the same is true from tier 2, but, the burden placed on you in tier 1 is greater.

Therefore, perhaps tier 2 is your sweet spot and tier 1 is too much for your abilities to handle? This is possible for sure and highlights a problem with the matchmaker in that it is too generous in allowing folks to move up but equally makes it too hard for them to move back down--thus you have a net positive pressure in the tiering system which carries people and subsequently keeps them higher.

[Deathlike]

PORONOPAPOS, on 12 August 2016 - 11:05 AM, said:

in my alt account I played 123 matches with Cheatah and around 150 with Battlemaster.

In main account played 74 matches in assault (including those of the event) and 14 with lights, also had some matches with heavies and meds.

Being T1 from last early December, I basically see same names all this time, although I feel their is some "rotation" of the players. I see some very often for a period , then some other names for some days and so on....As already said my only problem is that after late June more and more of the defeats are stomp ones, only in my main account

Depending on when you play... if you are seeing the same names... it's usually a population issue at that moment in time. That's not really fixable other than actually having more people play the game.

Also, the list tends to change if you're not running the same class all the time or running a "less popular" weight class at a given moment.

[Mawai]

Mister Blastman, on 12 August 2016 - 10:39 AM, said:

If it does not then why over 34,000 players analyzed is the median win/loss 1:1 and the average 1.09:1?

The rest of your post is pretty good and makes solid points, btw.

Why are the numbers 1:1 on average? ... because on average the matchmaker is working.

That is the point I was making ... 1:1 is NOT forced by the matchmaker ... 1:1 RESULTS when the matchmaker is more or less working.

HOWEVER ... you can still get 1:1 average results with terrible match quality ...

Consider the following ...

- PSR is perfect ... the matchmaker can create good well balanced matches where everyone is comparable skill and mech weight classes are matched ... your odds of winning are 50% and over enough matches your W/L -> 1:1 (Great ... this is the result we want).

- PSR is terrible ... it is SO bad that it is no better than picking players of random skill levels ... the matchmaker puts players with equal PSR against each other but PSR doesn't mean anything. So ... do you know what your odds of winning a match are on average? It is STILL 50% if the skills of all players are drawn from a random distribution. You get good and bad players on each side. You will win some and lose some. On average 50% resulting in a W/L of 1.0. However, the big difference is that the average match quality in this case will be VERY POOR. There will be more stomps since it doesn't take much to swing the match balance and the wide range of skills means that some folks will die quickly or just won't do the right things to win.

Unfortunately, you CAN'T use the W/L ratio to decide between these two extremes. Both Random player ranking and a perfect PSR will give an average 50% chance to win matches and result in an overall average 1:1 W/L. .. note that the MM does NOT care about your individual results .. if you play better your W/L will be greater than 1 and you will be matched against higher ranked opponents because you are better until you reach a point where you W/L is 1 BECAUSE you are playing against equally skilled opponents (that is the basic idea).

To resolve this you need more data. I'd suggest looking at death/loss ratio and the death/total games played since most games devolve to skirmish type game modes. If you define a good match as one where there are few survivors then you can get an idea of the average game result in terms of deaths/match. It should be between 12 and 24 .. the closer to 24 the more balanced the matches in theory.

Lostdragon: How do you download the leaderboard data?

[Mister Blastman]

Mawai, on 12 August 2016 - 11:12 AM, said:

Unfortunately, you CAN'T use the W/L ratio to decide between these two extremes. Both Random player ranking and a perfect PSR will give an average 50% chance to win matches and result in an overall average 1:1 W/L. .. note that the MM does NOT care about your individual results .. if you play better your W/L will be greater than 1 and you will be matched against higher ranked opponents because you are better until you reach a point where you W/L is 1 BECAUSE you are playing against equally skilled opponents (that is the basic idea).

I'm skeptical the matchmaker ignores win/loss as prior to PSR that is all it was based on. PGI has proven to us time and time again when they make coding changes, the bits sometimes come out like they were spun in a salad shaker.

[Dread Render]

Mister Blastman, on 12 August 2016 - 10:58 AM, said:

This means that there WILL be some players with an average win/loss ratio of 1.3, 1.4, 1.5, 1.6 and even 2.0 and this is not because the matchmaker is broken, it is because they are good at the game.

I think the part that you are not getting is the W/L ratio is just a gauge of how well the MM is working...
NOT HOW GOOD YOU ARE. Your skill rating is for that, not the W/L ration...

Good players should be matched against other good players. which means their W/L ratio should be about 1to1
Bad players should be matched against other bad players. which means their W/L ratio should be about 1to1

again... the W/L of every player is really just a gauge of how well the MM is working for YOU.