Jump to content


Looking For Random Battle Player Skill Logger.


  • Please log in to reply
8 replies to this topic

Coffee_ #1 Posted Dec 01 2017 - 01:59

    Staff sergeant

  • Players
  • 42894 battles
  • 276
  • Member since:
    06-17-2011

After so many months of MM putting all the better players on one side, I'd like to try an experiment.

I'd like some one that knows how, to write a mod that appends to a file the averaged wn8 of each team to a file..

I don't care about names or tanks.. just WN8 for each team
It can be really simple... Entries could just be something like this... 

RED: ##### GREEN : ##### : DATE/TIME

RED: ##### GREEN : ##### : DATE/TIME

RED: ##### GREEN : ##### : DATE/TIME

RED: ##### GREEN : ##### : DATE/TIME

##### being the averaged wn8 of the team.. Just add it up and divide by the player count... With the new large maps.. count won't always be 15.

I am tired of hearing that MM does NOT favor by skill when I am seeing it OVER and OVER, match after match.
This will prove either it does or does not!
If we get a dozen players to use the mod for a month and rack up a good data sample, the truth will be there for all to see.

After all the data is gathered in text files, I can write a tool to crunch the data and find the average difference between the two teams.

 

 



Kamahl1234 #2 Posted Dec 01 2017 - 02:07

    Major

  • Players
  • 18002 battles
  • 8,806
  • Member since:
    04-06-2012

You are aware that WN8 is a third party metric and not one WG created, right?

 

Their MM wouldn't use WN8 at all, even if it did do anything skill-based (which it does not). 



Buttknuckle #3 Posted Dec 01 2017 - 02:42

    Captain

  • Players
  • 41027 battles
  • 1,681
  • [GOONZ] GOONZ
  • Member since:
    03-19-2013

I hope you find what you are looking for OP. Its one thing to "feel" like something is happening and another to test to see if it is. The question is, once you have the data and the truth is there for all to see, will you believe it if it shows your hypothesis is wrong?

 

For the record, I believe you will learn that you are wrong. Over the years, lots of people have done similar testing and reporting of results on the forum. All of those that were presented in a logical, defensible, and reproducible manner have shown that what you are suggesting simply does not occur.  



GeorgePreddy #4 Posted Dec 01 2017 - 02:52

    Major

  • Players
  • 14334 battles
  • 6,703
  • [L_LEG] L_LEG
  • Member since:
    04-11-2013

If you get a REALLY LARGE sample, the average difference between teams using any skill metric will be close to zero since MM is random concerning any and all skill metrics, and being random, it will follow the Law of Large Numbers.

 

If you believe anything else it is just a tinfoil hat conspiracy theory unless you can prove it.

 

If you actually try to do what you suggest you will need a huge number of players tracking data to insure that the group averages out to average players.



ShuraBB #5 Posted Dec 01 2017 - 17:37

    Staff sergeant

  • -Players-
  • 6 battles
  • 278
  • Member since:
    10-20-2015

>I don't care about names or tanks.. just WN8 for each team

It is incorrect to use WN8 as it has nothing to do with WG at all, and moreover, does not reflect the objective level of the player.



Do_you_even_KV2 #6 Posted Dec 02 2017 - 05:31

    Private

  • -Players-
  • 577 battles
  • 5
  • Member since:
    01-17-2017

Best thing you can do is turn xvm off. It'll just make you angry! Every time I've used it, the majority of the time the teams are uneven. It nearly always slightly stacks one team wether by design or not. The rng alone can  make/break your game. When you are on a  bad rng streak you might as well not play. then you usually get a some good rng and you say, "HOW did that shot hit??" It's all random they say, it may well be over a certain sample size but the sample size itself is random!

Two players of exactly equal skill may play a 1000 games, one wins 570 and has worse teams and the other wins 800 and has better teams. 1000 games later that may be reversed.

In a game where there even exists a silly +/-25% rng you can't take anything seriously at all, those wn8 numbers usually only indicate how much premium ammo someone spams. As actually landing a shot has that silly 25% rng ribbish affecting it. You can be a super duper top 0.999% god player and +/-25% stupid mechanic can screw you over and let the 400 rating tomato 3 shot you.

 

The only displays of skill you can actually credit are tournaments where the ruleset for the game is sans 25% rng silliness.

 

It's pointless conducting studies on a game designed to mitigate skill so much in the first instance, you can't get any reliable data no matter how big the sample is-because there is always the possibility one person gets more or less bad than the next over that same sample size. Removing the 25% would bring it closer but you still can't quantify the competence of your team mates.



Coffee_ #7 Posted Dec 03 2017 - 05:04

    Staff sergeant

  • Players
  • 42894 battles
  • 276
  • Member since:
    06-17-2011

View PostGeorgePreddy, on Nov 30 2017 - 20:52, said:

 

If you believe anything else it is just a tinfoil hat conspiracy theory unless you can prove it.

 

And this my friend is exactly why I want to conduct this experiment.
If NOT WN8.. what other value(s) can one use?

 

EDIT:
I played 18 randoms yesterday... ID say by looking at win rate and WN8 maybe at best there was 5 matches that were some what even. The rest was a total blow out WIN or LOSS because of what appears to be a biasing of player skill. 
I know many disagree.. That's fine... but also... many agree that something is very fishy smelling. 

 



Coffee_ #8 Posted Dec 03 2017 - 05:23

    Staff sergeant

  • Players
  • 42894 battles
  • 276
  • Member since:
    06-17-2011

View PostButtknuckle, on Nov 30 2017 - 20:42, said:

I hope you find what you are looking for OP. Its one thing to "feel" like something is happening and another to test to see if it is. The question is, once you have the data and the truth is there for all to see, will you believe it if it shows your hypothesis is wrong?

 

For the record, I believe you will learn that you are wrong. Over the years, lots of people have done similar testing and reporting of results on the forum. All of those that were presented in a logical, defensible, and reproducible manner have shown that what you are suggesting simply does not occur.  

 

I first though of taking screen shots of every match... The issue with this is it can be easily tampered with but just capturing uneven matches. 
Also.. I want data from many sources.. Not just my battles. As I mentioned, Ideally.. getting data from a few dozen players over the course of a month would yield a good sample.. This sample can continue for as long as needed until everyone decides is a good enough data sample.
Compiling and charting the data would be pretty easy and verifiable. A good audit by a few people to make sure the numbers are crunched correctly would probably be best. I want a totally honest experiment ran. If it shows that the teams balance out to be near 50/50 than it puts an end to the biasing theory.  
 

Coffee_ #9 Posted Dec 03 2017 - 05:40

    Staff sergeant

  • Players
  • 42894 battles
  • 276
  • Member since:
    06-17-2011

I'm a .NET programmer... I can barely read Python scripts.. and Python is what this mod would need to be written in.
If you know me at all,, you have seen my Apps I wrote for WoTs. 
I have a model editor for the tanks. This lets you export the tanks to 3DS or Maya , edit them and then convert them back to the game's format.
I have written a Texture editing app for working with tank skinning.
I wrote a silly app that lets you change the names of the crewmen.. tanks.. and just about anything in the game.
I have a 3D Battle planning app to use in setting up strategies for clan wars. It loads nearly all the maps in full 3D almost exactly like the game.
You can find my work on GitHub... Get my apps HERE

 

As far as data mining.... 

I guess someone would need to modify XVMs app to gather the info as it already has the player network lookup code.. The only thing needed is a way to append the data until X number of matches have been reached... at which point, it should inform the player that the data sample is complete so it can be emailed to me or someone to hold for later crunching. What would be REALLY COOL is to have a way to send it automatically from the garage but... seeing how I do not know the game's network coding at all.. let alone Python to code it.. I don't even know if this can be done. However, some from of notifying the player really needs to happen so the data is passed on.  Also there is the issue of game updates.. The log files have to be preserved across updates.






1 user(s) are reading this topic

0 members, 0 guests, 0 anonymous users