Jump to content


New match making is poop

6.2 update 6.2 mm matchmaking

  • Please log in to reply
47 replies to this topic

Chris_Gunny #41 Posted 03 October 2019 - 10:33 AM

    Lance-corporal

  • Player
  • 36012 battles
  • 59
  • [DE1]
  • Member since:
    01-02-2015

stop using these match making engine/rules!!!!!!!!!!

i analyzed hundreds of games on two accounts. one is around 58% wr. the other over 60%. (only single player games were analyzed)

the assigned team mates on my side were in 90% of the games: 4-5 players with wr below 50% (and low exp), some even in the 30%. enemy team:   1-3 equally bad players. the rest above 50% and at least 2 players above 60% or 70%...

such matches could not be won by default. only way was when I (!!!) fully performed 100%. not one mistake! all was fine, etc. so these team didn’t carry anything in the game at all. 

 

it would be fair, if the teams are really randomly choosen OR the experience of the players are split equally on both teams. - at the moment it CANT be a random setup/mm. the figures show this clearly.


Edited by Chris_Gunny, 03 October 2019 - 10:35 AM.


Mjr_Leicht #42 Posted 03 October 2019 - 10:47 AM

    Senior Sergeant

  • Player
  • 15950 battles
  • 897
  • [BT_SV]
  • Member since:
    01-21-2019

View PostChris_Gunny, on 03 October 2019 - 10:33 AM, said:

stop using these match making engine/rules!!!!!!!!!!

i analyzed hundreds of games on two accounts. one is around 58% wr. the other over 60%. (only single player games were analyzed)

the assigned team mates on my side were in 90% of the games: 4-5 players with wr below 50% (and low exp), some even in the 30%. enemy team:   1-3 equally bad players. the rest above 50% and at least 2 players above 60% or 70%...

such matches could not be won by default. only way was when I (!!!) fully performed 100%. not one mistake! all was fine, etc. so these team didn’t carry anything in the game at all. 

 

it would be fair, if the teams are really randomly choosen OR the experience of the players are split equally on both teams. - at the moment it CANT be a random setup/mm. the figures show this clearly.


Interesting, I’ll run some analysis with the python scripts on my 2 accounts and see if I see the same thing...

How did you do this analysis?  Do you have the replays (or are they available on wotinspector.com) as it would be interesting for me to include those and compare & contrast...


560273467.png

WraithHawk #43 Posted 03 October 2019 - 10:59 AM

    Lance-corporal

  • Player
  • 4672 battles
  • 44
  • [CFG]
  • Member since:
    11-01-2015

View PostMjr_Leicht, on 03 October 2019 - 10:47 AM, said:


Interesting, I’ll run some analysis with the python scripts on my 2 accounts and see if I see the same thing...

How did you do this analysis?  Do you have the replays (or are they available on wotinspector.com) as it would be interesting for me to include those and compare & contrast...

Not everyone gets the same teams over a small number of games. 10k battles in each account would be a good start. And just for giggles, it would be interested to have one account that pays at least 1000 Guineas per year while the other one nothing. 



gallabru #44 Posted 03 October 2019 - 11:16 AM

    First Sergeant

  • Player
  • 18428 battles
  • 1,636
  • [LAGER]
  • Member since:
    09-01-2016

It is always funny to see that

1/ big imbalances are reported (like 90% of the games), but

2/ when it comes to statistics the need of tenth of thousands of games is requested to be convincing.

 

Either it is aimed at some players and 100 zillions tests of a second account is irrelevant, or the effect is big and there is no need of tousands of experiments.

Or there might be minute effects (which still could make sense from WG perspective) that require a lot of statistics to be relealed but they would go completely unnoticed by a normal player.


Edited by gallabru, 03 October 2019 - 11:18 AM.

My team stats in real time  : http://forum.wotblit...ats-repository/

Carlsberg clan family is recruiting. Contact us.


I_C4NT_W1N_ #45 Posted 03 October 2019 - 11:35 AM

    First Sergeant

  • Player
  • 50873 battles
  • 1,480
  • [8ASIL]
  • Member since:
    10-16-2015
It hasn't changed I don't think 

WraithHawk #46 Posted 03 October 2019 - 11:39 AM

    Lance-corporal

  • Player
  • 4672 battles
  • 44
  • [CFG]
  • Member since:
    11-01-2015

View Postgallabru, on 03 October 2019 - 11:16 AM, said:

It is always funny to see that

1/ big imbalances are reported (like 90% of the games), but

2/ when it comes to statistics the need of tenth of thousands of games is requested to be convincing.

 

Either it is aimed at some players and 100 zillions tests of a second account is irrelevant, or the effect is big and there is no need of tousands of experiments.

Or there might be minute effects (which still could make sense from WG perspective) that require a lot of statistics to be relealed but they would go completely unnoticed by a normal player.

1) and 2) are 2 different things. The first one doesn't require a high number of games because you can calculate the average win rate and then the odds for having a certain number of losses in a row. It might say the odds of a 10 game losing streak are once every 10000 games for a certain win rate and then it turns out that player had 3 such streaks in just 5000 games. The second point needs a lot more games because it does compare different tanks cause it's hard to find someone who played the same tanks in the same order on 2 accounts for the same amount of games.

The effects are indeed small and not all players notice it. They are probably around 10-15% of win rate when it comes to equally skilled players.



Chris_Gunny #47 Posted 03 October 2019 - 11:39 AM

    Lance-corporal

  • Player
  • 36012 battles
  • 59
  • [DE1]
  • Member since:
    01-02-2015

View PostMjr_Leicht, on 03 October 2019 - 11:47 AM, said:


Interesting, I’ll run some analysis with the python scripts on my 2 accounts and see if I see the same thing...

How did you do this analysis?  Do you have the replays (or are they available on wotinspector.com) as it would be interesting for me to include those and compare & contrast...


no, i did it manually. that means excel sheet and took the note of the match. the replays are not very interesting due to too high volume on data.

e.g. i took all players data at the end of the game and my place on the list and some judgement of my behavior/success.

 



Mjr_Eazy #48 Posted 03 October 2019 - 05:34 PM

    First Sergeant

  • Player
  • 42808 battles
  • 10,753
  • [BT_SV]
  • Member since:
    04-26-2017

View PostChris_Gunny, on 03 October 2019 - 11:39 AM, said:


no, i did it manually. that means excel sheet and took the note of the match. the replays are not very interesting due to too high volume on data.

e.g. i took all players data at the end of the game and my place on the list and some judgement of my behavior/success.

 


ok no worries, man that must have been painful, its what I used to do...

until...

jylpah released his python scripts, they make processing data in 100’s or even 1000’s of replays easy peasy...


547146227.png

 






Also tagged with 6.2, update 6.2, mm, matchmaking

1 user(s) are reading this topic

0 members, 1 guests, 0 anonymous users