Jump to content

Hammer Needs Your Help To Make 3050 Weekend Data Collection Weekend!


123 replies to this topic

#41 Jace Lancer

    Member

  • PipPipPip
  • Bridesmaid
  • Bridesmaid
  • 64 posts

Posted 05 April 2013 - 02:35 PM

View PostRG Notch, on 05 April 2013 - 11:08 AM, said:

The hilarious part is you can never get a truly random sample so it's entirely meaningless.

huh? there's and entire science behind it

#42 Wintersdark

    Member

  • PipPipPipPipPipPipPipPipPipPipPip
  • 13,375 posts
  • Google+: Link
  • Twitter: Link
  • LocationCalgary, AB

Posted 05 April 2013 - 03:16 PM

View PostCaptain Stiffy, on 05 April 2013 - 11:20 AM, said:

This is so ********. NO DON'T MOD ME YET I'm going to explain that.

PGI _HAS_THIS_DATA_ and I can't stress that enough.

Why don't they show your ELO score to you? Why? Think about that for a minute. What possible reason could they have to hide your own rating from you?

Especially since they show you your exact stats in the forum!

THE ONLY REASON THIS CAN BE IS BECAUSE IT IS BROKEN AND THEY KNOW IT!

You don't need random data collection to prove this (because you can do it with simple logic) - and by the way, to be even remotely statistically accurate, you would need at least one to two hundred results from every single player in the game to create a real trend from which you could model the rise and fall of ELO based on who is grouped with who.


It's not about Elo. You're right, you'd need waaaaay more results to chart Elo ratings, because particularly depending on the time your playing at the "band" of Elo ratings in a match can vary dramatically.

That said...

Why does PGI not share Elo data? It has nothing to do with it being broken or not. Even if it were 100% awesomely perfectly accurate, they'd do very, very well to not release the scores. Showing the scores breaks the system, as it becomes very much more "gameable". Also, it just leads to forum e-peen measurements, which is just never very good for the community as a whole.

Then you'd see all kinds of "Look at this screenshot of my Elo rating, it means I'm right. Show yours, so we can tell if you're right or not. Fail to show and obviously you're just hiding an embarassing score" and things like that.


Seriously though, dude, just relax. Hammer here is just looking to do some sampling to gather data, and he's a smart enough guy to know how accurate the gathered data will be and the potential pitfalls. It's not rocket science.

#43 Wintersdark

    Member

  • PipPipPipPipPipPipPipPipPipPipPip
  • 13,375 posts
  • Google+: Link
  • Twitter: Link
  • LocationCalgary, AB

Posted 05 April 2013 - 03:21 PM

View PostJace Lancer, on 05 April 2013 - 02:35 PM, said:

huh? there's and entire science behind it

What he means is this: You ask people at random on the forum for data. Some of these people respond.

This sample is not random.

You're taking results from people
1) Who are of the Forum Dweller clan, these are people who by and large are more polarised about their beliefs and their impression of the state of the game particularly given how the majority of non-Forum Dwellers will be entirely unaware of Huge Pressing Problems With The Game simply because they've never read about them on the forums or bothered to think about it.
and
2) Are willing to put in the added effort to respond. This breaks into two primary subsets: Those who have a point to make, and those who just want to help out.

So, yeah, these are not random results, as you're polling a specific subset of players.

As to polling overall, it's like asking for opinions on Public Education only in an upperclass neighborhood with a great school. You may be asking people at random, but your sample is from a limited demographic so the results cannot be extrapolated to the whole country.

With all that said, it is what it is, and Hammer is dealing with the cards he's dealt. Attempting to get a better dataset is pretty pointless, so he's going to do the best with what he can get. Works for me.

#44 RG Notch

    Member

  • PipPipPipPipPipPipPipPipPip
  • The 1 Percent
  • 2,987 posts
  • LocationNYC

Posted 05 April 2013 - 03:33 PM

View PostJace Lancer, on 05 April 2013 - 02:35 PM, said:

huh? there's and entire science behind it


View PostJace Lancer, on 05 April 2013 - 02:35 PM, said:

huh? there's and entire science behind it

LOL that's not a real science, unless you consider social "sciences" which cannot be tested scientifically science. You might as well claim astrology is a science if you're going that way.
Even if you regard the "science" behind polling to be valid, this polling isn't even scientific by those guidelines. Anyways as a wise man said there are lies, damn lies and statistics.

#45 hammerreborn

    Member

  • PipPipPipPipPipPipPipPipPip
  • Legendary Founder
  • Legendary Founder
  • 3,063 posts
  • LocationAlexandria, VA

Posted 05 April 2013 - 03:35 PM

Added another rule:

Not sure if it needed to be said but I prefer 8v8 games with no DCs.

#46 Hauser

    Member

  • PipPipPipPipPipPipPip
  • Legendary Founder
  • Legendary Founder
  • 976 posts

Posted 05 April 2013 - 03:46 PM

Hammer did you post this too reddit too?

http://www.reddit.com/r/mwo/

And maybe try NGNG as well.

#47 FunkyFritter

    Member

  • PipPipPipPipPipPip
  • 459 posts

Posted 05 April 2013 - 04:05 PM

http://i.imgur.com/6vKOwjE.png
http://i.imgur.com/e3LrnPJ.png
http://i.imgur.com/G7nVlFD.png
http://i.imgur.com/BTTHJT7.png
http://i.imgur.com/tnwqFOH.png
http://i.imgur.com/9JXBQdO.png
http://i.imgur.com/jdYEyig.png
http://i.imgur.com/SJo1eWM.png
http://i.imgur.com/1iddGOs.png
http://i.imgur.com/naHCUAC.png

#48 hammerreborn

    Member

  • PipPipPipPipPipPipPipPipPip
  • Legendary Founder
  • Legendary Founder
  • 3,063 posts
  • LocationAlexandria, VA

Posted 05 April 2013 - 04:14 PM

One last rule I think we'll be good to go:

Assault only please! I forgot people actually play conquest. Sorry Rigatoni and FunkyFritter! Though your data did make me remember that since I last did this 2 new mechs and 2 heroes have been released. Although if I get enough conquest submissions maybe I'll do two seperate data lists.

I want to stick to Assault because people play vastly different mechs in each mode which can make the results wonky.

View PostHauser, on 05 April 2013 - 03:46 PM, said:

Hammer did you post this too reddit too?

http://www.reddit.com/r/mwo/

And maybe try NGNG as well.


I don't do either of those things, though if you want to post something about it feel free and I'd appreciate it.

Edited by hammerreborn, 05 April 2013 - 04:26 PM.


#49 Kobold

    Member

  • PipPipPipPipPipPipPipPipPip
  • Elite Founder
  • 2,930 posts
  • LocationChicago, IL

Posted 05 April 2013 - 04:32 PM

Maybe it was already stated and I missed it: What are you hoping to find?

#50 hammerreborn

    Member

  • PipPipPipPipPipPipPipPipPip
  • Legendary Founder
  • Legendary Founder
  • 3,063 posts
  • LocationAlexandria, VA

Posted 05 April 2013 - 04:37 PM

View PostKobold, on 05 April 2013 - 04:32 PM, said:

Maybe it was already stated and I missed it: What are you hoping to find?


Mech diversity primarily, amoung other useless statistics. Also how many people I can get to submit 10 screenshots for data collection and possible future attempts

For instance, from both peoples submissions and my own data so far, there are a lot of people making use of this hero weekend.

Maybe depending on how this goes I'll have a weekday data collection that won't be skewed by tourney/weekend bonuses.

Edited by hammerreborn, 05 April 2013 - 04:39 PM.


#51 Ph30nix

    Member

  • PipPipPipPipPipPipPipPipPip
  • 2,444 posts

Posted 05 April 2013 - 05:11 PM

View PostRG Notch, on 05 April 2013 - 11:08 AM, said:

The hilarious part is you can never get a truly random sample so it's entirely meaningless.

yep and yet companies and politicians always still try and use "statistical data" to prove what they want it to prove. Also the fact you can manipulate just about any set of numbers in some way that it will tell the story you want.

#52 Rigatoni

    Member

  • Pip
  • Knight Errant
  • Knight Errant
  • 15 posts

Posted 05 April 2013 - 05:23 PM

View Posthammerreborn, on 05 April 2013 - 04:14 PM, said:

One last rule I think we'll be good to go:

Assault only please! I forgot people actually play conquest. Sorry Rigatoni and FunkyFritter! Though your data did make me remember that since I last did this 2 new mechs and 2 heroes have been released. Although if I get enough conquest submissions maybe I'll do two seperate data lists.

I want to stick to Assault because people play vastly different mechs in each mode which can make the results wonky.



ahhh. i usually just play solo random. although i do have a friend that i play with off and on. if you want i will still post the conquest, but i will make sure it is segregated.

#53 hammerreborn

    Member

  • PipPipPipPipPipPipPipPipPip
  • Legendary Founder
  • Legendary Founder
  • 3,063 posts
  • LocationAlexandria, VA

Posted 05 April 2013 - 05:41 PM

View PostRigatoni, on 05 April 2013 - 05:23 PM, said:



ahhh. i usually just play solo random. although i do have a friend that i play with off and on. if you want i will still post the conquest, but i will make sure it is segregated.


I won't stop you if you do both =P

#54 Krzysztof z Bagien

    Member

  • PipPipPipPipPipPipPip
  • Giant Helper
  • 710 posts
  • LocationUć, Poland

Posted 05 April 2013 - 06:10 PM

View PostWintersdark, on 05 April 2013 - 03:16 PM, said:

Also, it just leads to forum e-peen measurements, which is just never very good for the community as a whole.

Then you'd see all kinds of "Look at this screenshot of my Elo rating, it means I'm right. Show yours, so we can tell if you're right or not. Fail to show and obviously you're just hiding an embarassing score" and things like that.


Replace 'Elo rating' with 'K/D ratio', 'Win/Loss ratio', 'Number of matches played', 'Number of posts' etc... It's good we can't see those things. Oh, wait...
Seriously, with all due respect, what you said being a reason not to show Elo rating just doesn't make any sense (I know the devs use this logic too, but it doesn't make it less stupid).

BTW - I can't see a way Elo system could work properly in MWO. If I understand it right it only takes into account whether you win or lose a match (or am I missing something?). It was designed to be used in 1vs1 games, so that would be OK if it rated a team as a whole (teams could be treated as one collective player), but using it to rate individual players in a team based game is simply wrong. If my performance in a match was so significant that it basically decides whether my team wins or not it COULD work, but then again that would mean other players' performance is irrelevant, so they can't be rated.
Something more like Microsoft's TrueSkill should be used here (it's patented though, so PGI would have to obtain a license for it).

#55 hammerreborn

    Member

  • PipPipPipPipPipPipPipPipPip
  • Legendary Founder
  • Legendary Founder
  • 3,063 posts
  • LocationAlexandria, VA

Posted 05 April 2013 - 06:37 PM

So just a glimpse of how this is probably going to look when all is said and done:

Obviously this is only 10 matches so isn't really evident of anything, but with maybe 10 more of these (or hopefully a lot more!)

Posted Image

Interesting notes: Tonnage was relatively balanced. When you look at the mechs as actual mechs, you think that you're being massively outtonned (such as the game where the enemy had 3 D-DCs), but when you put pencil to paper it actually wasn't that off.

#56 Rigatoni

    Member

  • Pip
  • Knight Errant
  • Knight Errant
  • 15 posts

Posted 05 April 2013 - 07:21 PM

10 more A. matchs for you.

#1
Spoiler

#2
Spoiler

#3
Spoiler

#4
Spoiler

#5
Spoiler

#6
Spoiler

#7
Spoiler

#8
Spoiler

#9
Spoiler

#10
Spoiler


#57 hammerreborn

    Member

  • PipPipPipPipPipPipPipPipPip
  • Legendary Founder
  • Legendary Founder
  • 3,063 posts
  • LocationAlexandria, VA

Posted 05 April 2013 - 07:41 PM

And here's you results:
Interestingly, you have a lot more light mechs seen than I did. But that's why I need MOAR data

Spoiler

Edited by hammerreborn, 05 April 2013 - 07:55 PM.


#58 jay35

    Member

  • PipPipPipPipPipPipPipPip
  • The Hammer
  • The Hammer
  • 1,597 posts

Posted 05 April 2013 - 08:52 PM

Here dude, banged this out for you over the past couple hours. Hope it helps.

Spoiler

Keep in mind, evvvveryone is running Hero and Founders mechs tonight, so it's likely a lot of ECM mechs are in the hangar while people grab cash with the extra bonuses. For example, I've never had a team with THREE Founders Jenners on it before in the entire time I've played the game.

Edited by jay35, 05 April 2013 - 08:53 PM.


#59 hammerreborn

    Member

  • PipPipPipPipPipPipPipPipPip
  • Legendary Founder
  • Legendary Founder
  • 3,063 posts
  • LocationAlexandria, VA

Posted 05 April 2013 - 08:57 PM

Can you run one more to replace that 7v8 game when you get the chance?

Thank you both for the help.

Edited by hammerreborn, 05 April 2013 - 08:58 PM.


#60 jay35

    Member

  • PipPipPipPipPipPipPipPip
  • The Hammer
  • The Hammer
  • 1,597 posts

Posted 05 April 2013 - 09:14 PM

Spoiler






6 user(s) are reading this topic

0 members, 6 guests, 0 anonymous users