Jump to content

Pugs Are Probably Exaggerating


262 replies to this topic

#61 abloobloo

    Member

  • PipPip
  • 21 posts

Posted 23 January 2013 - 09:46 PM

View PostSnowhawk, on 23 January 2013 - 09:44 PM, said:

Well... I think there is a little problem in your statistic because:




You asked the people from this forum..... And that makes a huge difference. Please keep in mind that gamers who are often reading in a forum are "sophisticated players". They evolve much faster then the real casual gamers who not even know the damage values of the weapons.


Maybe, but then again, you could also speculate that people who spend less time on the forums are spending more time improving their game.

#62 Thirdstar

    Member

  • PipPipPipPipPipPipPipPipPip
  • 2,728 posts
  • LocationIndia

Posted 23 January 2013 - 09:49 PM

View Postabloobloo, on 23 January 2013 - 09:46 PM, said:

Maybe, but then again, you could also speculate that people who spend less time on the forums are spending more time improving their game.


Based on what? Unicorns and rainbows? Did the leprechauns help you come to this speculation.

I speculate that all premade teams secretly eat babies to improve their PUBstomping prowess.

Edited by Thirdstar, 23 January 2013 - 09:50 PM.


#63 Wolf Ender

    Member

  • PipPipPipPipPipPip
  • 495 posts
  • Google+: Link
  • LocationSacramento, California

Posted 23 January 2013 - 09:50 PM

View PostRifter, on 23 January 2013 - 08:03 PM, said:

Sorry to break this to you but you are going to need 2000+ samples to get reliable stats.


not exactly.

presidential polls are often done on sample sizes as low as 500 and that's considered scientifically accurate +- 3-5% for a universe of more than 130 million likely voters.

the size isn't so much important as making sure all demographics are accurately and proportionally represented in that sample. if he's only getting 2000 people who voluntarily submit their scores from this forum only, without any kind of controls for other factors of those gamers... it doesn't matter if he gets 30 or 3,000, its still not going to be REALLY trustworthy in terms of results.

I think there are a lot of different factors you would have to take into account in order to figure this out really, and it might just be too difficult to do.. i mean just based on the top of my head you would have to control and make sure you're getting the right distribution of stats from people in different time zones represented, different ping levels, what mech class they piloted during that time, what types of configs and play styles are represented.

then I think you have to question whether or not it's a good idea to poll only puggers. even though you're getting stats from people who ONLY pug (if you believe the players) you don't have any stats about who they were playing against. it's a factor that we really just don't know

#64 Mechwarrior Buddah

    Member

  • PipPipPipPipPipPipPipPipPipPipPip
  • Legendary Founder
  • Legendary Founder
  • 13,459 posts
  • LocationUSA

Posted 23 January 2013 - 09:52 PM

View PostGhogiel, on 23 January 2013 - 07:58 PM, said:

out of people who post and read the forum.


Yeah 31 of the bottom 5% of all the people in the game

#65 Thirdstar

    Member

  • PipPipPipPipPipPipPipPipPip
  • 2,728 posts
  • LocationIndia

Posted 23 January 2013 - 09:53 PM

View PostWolf Ender, on 23 January 2013 - 09:50 PM, said:


not exactly.

presidential polls are often done on sample sizes as low as 500 and that's considered scientifically accurate +- 3-5% for a universe of more than 130 million likely voters.


You make excellent points but I will point out that many many pollsters were horribly wrong in the 2012 Presidential Elections.

#66 abloobloo

    Member

  • PipPip
  • 21 posts

Posted 23 January 2013 - 09:56 PM

View PostThirdstar, on 23 January 2013 - 09:49 PM, said:


Based on what? Unicorns and rainbows? Did the leprechauns help you come to this speculation.

I speculate that all premade teams secretly eat babies to improve their PUBstomping prowess.


The very same unicorns and rainbows that snowhawk based his assumptions on. That was the point I was trying to make.

#67 Wired

    Member

  • PipPipPipPipPipPipPip
  • Knight Errant
  • Knight Errant
  • 822 posts

Posted 23 January 2013 - 09:57 PM

In the topic: OP posting science and facts, and a handful of people who should just be ignored trash him for it.

/thread.

#68 Thirdstar

    Member

  • PipPipPipPipPipPipPipPipPip
  • 2,728 posts
  • LocationIndia

Posted 23 January 2013 - 09:59 PM

View PostWired, on 23 January 2013 - 09:57 PM, said:

In the topic: OP posting science and facts, and a handful of people who should just be ignored trash him for it.

/thread.


31 data samples. Science and Facts.

In the news today. Wired has no idea what Science and Facts actually look like.

#69 Applecrow

    Member

  • PipPipPipPipPipPip
  • Bridesmaid
  • Bridesmaid
  • 370 posts

Posted 23 January 2013 - 10:00 PM

First rule of MWO pugging: If the blue arrows aren't working together, it's because the red arrows are a sync-dropped cheese build premade just there to stomp-n-farm.

Teamwork is OP and needs to be nerfed used more often.

#70 abloobloo

    Member

  • PipPip
  • 21 posts

Posted 23 January 2013 - 10:00 PM

View PostWolf Ender, on 23 January 2013 - 09:50 PM, said:


not exactly.

presidential polls are often done on sample sizes as low as 500 and that's considered scientifically accurate +- 3-5% for a universe of more than 130 million likely voters.

the size isn't so much important as making sure all demographics are accurately and proportionally represented in that sample. if he's only getting 2000 people who voluntarily submit their scores from this forum only, without any kind of controls for other factors of those gamers... it doesn't matter if he gets 30 or 3,000, its still not going to be REALLY trustworthy in terms of results.

I think there are a lot of different factors you would have to take into account in order to figure this out really, and it might just be too difficult to do.. i mean just based on the top of my head you would have to control and make sure you're getting the right distribution of stats from people in different time zones represented, different ping levels, what mech class they piloted during that time, what types of configs and play styles are represented.

then I think you have to question whether or not it's a good idea to poll only puggers. even though you're getting stats from people who ONLY pug (if you believe the players) you don't have any stats about who they were playing against. it's a factor that we really just don't know


I wanted to know what comes out when you throw pugs into the meat grinder that is the random matchmaking system. Sure, there are various factors that affect individual performance, and that's reflected in the data. Not every pug has a win/loss ratio 0.92; some score above and some below, and I would hazard a guess and say that some of those factors you mention play a part in the variation in individual scores.

#71 Mechwarrior Buddah

    Member

  • PipPipPipPipPipPipPipPipPipPipPip
  • Legendary Founder
  • Legendary Founder
  • 13,459 posts
  • LocationUSA

Posted 23 January 2013 - 10:02 PM

View PostWired, on 23 January 2013 - 09:57 PM, said:

In the topic: OP posting science and facts, and a handful of people who should just be ignored trash him for it.

/thread.


"Science and facts" of what % of all the players in the game?
1%?

#72 Adrienne Vorton

    Member

  • PipPipPipPipPipPipPipPipPip
  • 3,535 posts
  • LocationBerlin/ Germany

Posted 23 January 2013 - 10:06 PM

the problem with the MC guys is indeed that the stats on the website only show the CURRENT MC as well as CURRENT CBILLS... so someone who has MC on his account, hasn´t just spend them... that means that people who buy but don´t use MC are in advantage?

#73 Critical Fumble

    Member

  • PipPipPipPipPipPipPip
  • Knight Errant
  • 810 posts

Posted 23 January 2013 - 10:09 PM

Your conclusion is flawed, not by bad data, but because you gathered data that is largely irrelevant to the conclusion. You gathered PUG statistics to say that premade teams don't have an undue influence on the game, when in reality as a PUG you could be on either side of an uneven premade match - IE: a premade team can cause you to win as well as lose. A premade team could still very well be the determining factor. If you were trying to show that a premade teams don't imbalance matches, you need a different kind of data. Probably by match, with people going solo, in a simple premade with average mechs, and in a organized premade with a good set of mechs. If that data showed a minimal difference in success, then you might be able to prove a point.

I have no idea what you were thinking when you wrote: "However, I was disappointed to see no strong relationship between player performance and the amount of MC in their accounts." But I will say that a relationship between putting money into the game and success is expected in a F2P title, its F2P marketing 101, really. If a person does well, they gain interest, and interest encourages spending money. Either that or the game has strong P2W aspects, but I have yet to see a bobblehead win a match, so I doubt the second option matters here.

#74 Thirdstar

    Member

  • PipPipPipPipPipPipPipPipPip
  • 2,728 posts
  • LocationIndia

Posted 23 January 2013 - 10:09 PM

View PostAdrienne Vorton, on 23 January 2013 - 10:06 PM, said:

the problem with the MC guys is indeed that the stats on the website only show the CURRENT MC as well as CURRENT CBILLS... so someone who has MC on his account, hasn´t just spend them... that means that people who buy but don´t use MC are in advantage?


Yes. Clearly the conclusion is that you should buy MC and then not spend it. Then your K/D is sure to soar to unheard of heights.

#75 Grand Ayatollah Kerensky

    Member

  • PipPipPipPipPipPipPip
  • 749 posts

Posted 23 January 2013 - 10:11 PM

View PostAdrienne Vorton, on 23 January 2013 - 10:06 PM, said:

the problem with the MC guys is indeed that the stats on the website only show the CURRENT MC as well as CURRENT CBILLS... so someone who has MC on his account, hasn´t just spend them... that means that people who buy but don´t use MC are in advantage?


I think that the difference comes from the fact that a lot of people who responded had irregular amounts of MC indicating that they had already spent a good chunk of it. It could be that the higher kdr among people with MC in their accounts is due to the fact that most of them were able to skip the trial mech meat grinder.

Read the excel file.

Aslo, neat study, op.

Edited by Narcisoldier, 23 January 2013 - 10:30 PM.


#76 abloobloo

    Member

  • PipPip
  • 21 posts

Posted 23 January 2013 - 10:23 PM

View PostCritical Fumble, on 23 January 2013 - 10:09 PM, said:

Your conclusion is flawed, not by bad data, but because you gathered data that is largely irrelevant to the conclusion. You gathered PUG statistics to say that premade teams don't have an undue influence on the game, when in reality as a PUG you could be on either side of an uneven premade match - IE: a premade team can cause you to win as well as lose. A premade team could still very well be the determining factor. If you were trying to show that a premade teams don't imbalance matches, you need a different kind of data. Probably by match, with people going solo, in a simple premade with average mechs, and in a organized premade with a good set of mechs. If that data showed a minimal difference in success, then you might be able to prove a point.


I wasn't asking a question about individual performance, I was asking a question about group performance. What happens when you add pugs, premades, and other factors to the current matchmaking system? Do pugs as a group lose out big time? The data suggests no.

#77 Thirdstar

    Member

  • PipPipPipPipPipPipPipPipPip
  • 2,728 posts
  • LocationIndia

Posted 23 January 2013 - 10:31 PM

View Postabloobloo, on 23 January 2013 - 10:23 PM, said:


I wasn't asking a question about individual performance, I was asking a question about group performance. What happens when you add pugs, premades, and other factors to the current matchmaking system? Do pugs as a group lose out big time? The data suggests no.


So you are attempting to counter the assertion that individual experiences of PUGs is effected by the presence of premades in the matchmaker by stating that PUGs as a group don't lose out 'big time'.

Your question is flawed and has nothing to do with the issues that the community has brought up.

#78 aniviron

    Member

  • PipPipPipPipPipPipPipPipPip
  • 3,752 posts

Posted 23 January 2013 - 10:42 PM

It's a good day for this thread. Patch day always brings out all the premades- solo pugging, as usual, I have lost ten games in a row now, usually with 4-6 teammates under 100 damage, and an announcement from the other side that I should visit clanname.com

#79 Grand Ayatollah Kerensky

    Member

  • PipPipPipPipPipPipPip
  • 749 posts

Posted 23 January 2013 - 10:53 PM

.

Edited by Narcisoldier, 23 January 2013 - 10:53 PM.


#80 Ken Fury

    Member

  • PipPipPipPipPipPipPipPip
  • 1,016 posts
  • LocationGermany

Posted 23 January 2013 - 10:53 PM

Well since he asked to get stats from people who NEVER played in a premade this should make the gathered data actually quite worthwile. And 31 Posters is a rather high % of active posters.

I'd wonder if there are some relation between postcount and W/L, K/D...





11 user(s) are reading this topic

0 members, 11 guests, 0 anonymous users