Jump to content


Why I believe that Random battles MM is the best it can be and cannot be made fairer, better ect.


  • Please log in to reply
42 replies to this topic

jylpah #41 Posted 01 July 2020 - 05:33 PM

    First Sergeant

  • Player
  • 22935 battles
  • 5,242
  • [AFK]
  • Member since:
    08-05-2014

View Postgallabru, on 01 July 2020 - 07:16 PM, said:

Jylpah could you explain us what the 0-1... results are ?

 

A loss, the reds had one player left.

 

View Postgallabru, on 01 July 2020 - 07:16 PM, said:

If we discard draws, supremacy (strong approximation) and last second suicide (7-7), battle results should be N-7 (N=0:6). Then 'steamrolls' should account for 1/7 (counting red crushing just double all numbers) of the results, i.e. 14%. If we assume that supremacy accounts for half the battles and that steamrolls never happen in supremacy (strong approximation again) then 7-0 could account for 7% of the battle results.

At the other end of the statistical spectrum, we could assume that a there are more 4-7 than 14% of the games because there are many ways to get 4 deads. Then one N-7 result can happen 7!/N!(7-N)! times. For N=0, this is 1 again but the number of possible battle result amounts to 114. So, 0.8 % of steamrolls.

 

Simple(ton) maths for purely random results. But, obviously, we are hoping to have an impact on the result (good or bad) otherwise why play ?

 

You are interpreting the numbers other way than intended (the numbers denote the number of remaining players per team).

 

If I understood you correctly you are assuming that each (atomic) outcome (0-5, 0-3) are equally probable and then you are using combinatorics to calculate the probabilities, right (like calculating the probabilities for a sum of several dices)? Your steamroller estimate 0.8% is far from the real data so maybe the assumptions do not hold?


Edited by jylpah, 01 July 2020 - 05:42 PM.

Blitzanalysiz.com  In-depth analysis of WoT Blitz game statistics

Jylpah @EU  Jylpah @Asia  Blitz-tools  


Jukkis74 #42 Posted 01 July 2020 - 06:06 PM

    Senior Sergeant

  • Player
  • 35405 battles
  • 880
  • [AFK]
  • Member since:
    06-29-2014
Building a simulator that achieves close to observed results could be instructive. Will think about this a bit. On first thought, the simulator should reflect that 1) more guns left is better, 2) more dmg sinks left is better. My gut feeling is that ~10% steamrolls could be achieved with quite reasonable abstractions.

gallabru #43 Posted 01 July 2020 - 06:42 PM

    Senior Sergeant

  • Player
  • 18101 battles
  • 668
  • [LAGER]
  • Member since:
    09-01-2016
@Jylpah, thanks. So my numbers are just 7 minus yours.

I am proposing two boundaries : between 0.8% and 14% depending on whether we must consider all combinations yielding one result or that all total results are equiprobable. That can be weighted by the considered players' wr but if 7-0 is as annoying as 0-7 then it will be compensated.
Obviously, my assumption is also that I consider only random battles with an outcome (win ot lose) so that a team has 0 players left.

Finally, I would not trust my calculations as being more representative of the game than your analysis, hence my before last sentence. 

Sidenote : replays contain also the evolution of the game. It could be possible to correlate the battle results with the time at which the first player dies or anything else.

Edited by gallabru, 01 July 2020 - 06:43 PM.

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

Carlsberg clan family is recruiting. Contact us.





1 user(s) are reading this topic

0 members, 1 guests, 0 anonymous users