Jump to content


Blowouts. Numerical results


  • Please log in to reply
97 replies to this topic

LesterQuaestor #1 Posted Jan 27 2016 - 22:54

    Captain

  • -Players-
  • 11970 battles
  • 1,308
  • Member since:
    03-26-2015

 

  A lot of people complain about blowout battles. A lot of angst goes into "bad teams", "bad MM", "rigging", etc.

 

  But some people point out that blowouts should be more common than many people think they should. I agree with them. The underlying reason is that WoT is inherently unstable since there is no respawning. Once you begin to lose the chance of losing accelerates.

 

  I did some computations to show this. This is a very very simple probabilistic model (a "Monte Carlo" simulation). The chance of killing an enemy tank is directly proportional to the number of your tanks left. For this I have left out capping and draws.

 

  A blow out is defined as a result of between 15-0 and 15-4. I simulated 10,000 battles and the probability of each blowout score was determined.

 

 For this first example the teams are exactly balanced.

 

Win %    50/50

 

15-0      0.58%

15-1      1.81%

15-2      3.93%

15-3      6.61%

15-4      8.22%

 

Any Blowout:  21.15%

 

So it appears that in a 50:50 match you should expect a result worse than 15:5 every 5 battles. (This doesn't indicate which side you are on, so 15:4 and 4:15 are the same.)

 

Since you are interested, changing the balance does effect the number of blowouts, but not dramatically. I did four different balance ratios, but I will only show the greatest imbalance. You can interpolate the others.

 

This simulation shows 10000 battles between two teams that are significantly unbalanced.

win %     83/17

 

15-0       1.27%

15-1       3.97%

15-2       7.30%

15-3       10.20%

15-4       11.60%

 

Any Blowout:  34.34%

 

So for every 3 battles where the odds are as uneven as 87/13 there should be one blowout.

 

TLDR: MM is fine, and blowouts are about as common as would be expected from basic statistical probability.

 



_Fevs_ #2 Posted Jan 27 2016 - 22:58

    Captain

  • Players
  • 28537 battles
  • 1,533
  • Member since:
    03-31-2012

Stronk math.

 

However, it is not tinfoil-proof :trollface:



LesterQuaestor #3 Posted Jan 27 2016 - 23:01

    Captain

  • -Players-
  • 11970 battles
  • 1,308
  • Member since:
    03-26-2015

Block Quote

  it is not tinfoil-proof

 

  Nothing is. The only thing tinfoil protects you from is reason...

 

 



Yuri_Doujinshi #4 Posted Jan 27 2016 - 23:06

    Major

  • Players
  • 23056 battles
  • 2,971
  • [Y0UJ0] Y0UJ0
  • Member since:
    05-07-2012
This thread should be pinned.

sherman_elimanator #5 Posted Jan 27 2016 - 23:11

    First lieutenant

  • Players
  • 6639 battles
  • 823
  • Member since:
    04-29-2012

View PostLesterQuaestor, on Jan 27 2016 - 14:01, said:

 

  Nothing is. The only thing tinfoil protects you from is reason...

 

 

 

and logic.

View PostKirino_Desu, on Jan 27 2016 - 14:06, said:

This thread should be pinned.

^^^^^^^^^^^^^^^^^^^^^^^^^



killswitch95 #6 Posted Jan 27 2016 - 23:12

    Major

  • Players
  • 38774 battles
  • 8,804
  • [K0HAI] K0HAI
  • Member since:
    04-12-2012

But but, I have Salt that needs Intensifying due to MM

 

#ThejimmieshavebeenrustledREPEATthejimmieshavebeenrustled!



LesterQuaestor #7 Posted Jan 27 2016 - 23:14

    Captain

  • -Players-
  • 11970 battles
  • 1,308
  • Member since:
    03-26-2015

View PostKirino_Desu, on Jan 27 2016 - 23:06, said:

This thread should be pinned.

 

 

Gosh. I'm blushing... :medal:

MMHT #8 Posted Jan 27 2016 - 23:18

    Sergeant

  • Players
  • 13405 battles
  • 166
  • Member since:
    11-23-2013

So if I understand this right... 

View PostLesterQuaestor, on Jan 27 2016 - 21:54, said:

 

The chance of killing an enemy tank is directly proportional to the number of your tanks left.  

 

 

Means that the first iteration of computation is a 50/50 chance that a tank from each team will die.  After that there is a 15:14 or 14:15 chance of death, then leading to alternate possibilities of 15:13,14:14, and 13:15 on down the line until the chance of every possible outcome has been determined?

 

NOTE: Not trying to imply anything about MM or the validity of the argument, just want to understand the math.



LesterQuaestor #9 Posted Jan 27 2016 - 23:25

    Captain

  • -Players-
  • 11970 battles
  • 1,308
  • Member since:
    03-26-2015

Block Quote

  just want to understand the math.

 

  Not quite as you describe.

 

  For each iteration I select one tank from all tanks left alive (initially 30). It then has a 50% chance to kill an enemy (reduce the number of enemies by one). Then it goes to the next iteration. I simulated imbalance by giving each team a slightly different chance to kill.

 

 



power_Donky #10 Posted Jan 27 2016 - 23:29

    Major

  • Players
  • 25977 battles
  • 3,040
  • [HAVOK] HAVOK
  • Member since:
    11-06-2013
You have to remember that rng has a very large effect on the outcome of most of these balanced matches, it's not all team skill

t6c #11 Posted Jan 27 2016 - 23:29

    Staff sergeant

  • Players
  • 29678 battles
  • 403
  • Member since:
    11-24-2011
Excellent work. Needs to be pinned or put in a guide collection.

The Lanchester Equations show the same thing: casualties inflicted and the win chances vary with the force ratio squares. 15 tanks against 14 equals is
15^2/14^2, or 225/196= 15% advantge. 14 v 10 is about 96%.

Nice pic Lester. Always happy to see another ASL player. Can you imagine what would happen to the typical WoTer if WG changed the RNG so that there was a 2% chance that a Tier V could 1-shot a Tier X? WoT players know nothing of randomness compared to ASL players.


ratpak #12 Posted Jan 27 2016 - 23:31

    Major

  • Beta Testers
  • 26535 battles
  • 2,367
  • [DHO-X] DHO-X
  • Member since:
    11-24-2010
In a deterministic equation maybe this holds true, however this game and players are anything but that.  Way too many variables (human mainly) for this to "add up"

LesterQuaestor #13 Posted Jan 27 2016 - 23:35

    Captain

  • -Players-
  • 11970 battles
  • 1,308
  • Member since:
    03-26-2015

Block Quote

 You have to remember that rng has a very large effect on the outcome of most of these balanced matches, it's not all team skill

 

  100% of my result is based on rng. Well, Matlab's pseudo random number generator.

 

 



LesterQuaestor #14 Posted Jan 27 2016 - 23:38

    Captain

  • -Players-
  • 11970 battles
  • 1,308
  • Member since:
    03-26-2015

Block Quote

 The Lanchester Equations show the same thing

 

Not being a math, stat, or game theory person, I never heard of that. But I did consider using the square of the number of tanks left.

 

Block Quote

 Nice pic Lester. Always happy to see another ASL player. Can you imagine what would happen to the typical WoTer if WG changed the RNG so that there was a 2% chance that a Tier V could 1-shot a Tier X? WoT players know nothing of randomness compared to ASL players.

 

  Yes, WG uses hit point obviously as a concession to playability. And it largely works. I would personally prefer the ASL version. They could adopt the tables directly.

 



DopeRaider #15 Posted Jan 27 2016 - 23:48

    First lieutenant

  • Players
  • 14750 battles
  • 955
  • [S_U_X] S_U_X
  • Member since:
    06-12-2013
Excellent post!

LesterQuaestor #16 Posted Jan 28 2016 - 00:02

    Captain

  • -Players-
  • 11970 battles
  • 1,308
  • Member since:
    03-26-2015

Block Quote

 In a deterministic equation maybe this holds true, however this game and players are anything but that.  Way too many variables (human mainly) for this to "add up"

 

My model is not deterministic. But over 10000 battles, all the variables will cancel each other out. That's why I didn't simulate 10 battles.

 

 



t6c #17 Posted Jan 28 2016 - 00:10

    Staff sergeant

  • Players
  • 29678 battles
  • 403
  • Member since:
    11-24-2011
Have you tried the Excel add-ins? I use @Risk and ModelRisk. Also some R, but few people here understand it.

OstwindFlakpanzer #18 Posted Jan 28 2016 - 00:29

    Major

  • Players
  • 20919 battles
  • 2,966
  • [CUDAS] CUDAS
  • Member since:
    08-22-2012
this looks like some of that common core obama math.  blowouts should never happen.  fix your MM, WG!!

RyanGetzIaf #19 Posted Jan 28 2016 - 00:35

    Major

  • Beta Testers
  • 24360 battles
  • 3,800
  • [TYR] TYR
  • Member since:
    09-15-2010

View PostOstwindFlakpanzer, on Jan 27 2016 - 18:29, said:

this looks like some of that common core obama math.  blowouts should never happen.  fix your MM, WG!!

 

Blowout are NOT caused by MM, there is nothing they can "fix" to change that. 

ISNomads #20 Posted Jan 28 2016 - 00:44

    Major

  • Players
  • 39573 battles
  • 2,493
  • [WHAMO] WHAMO
  • Member since:
    11-30-2013

View PostLesterQuaestor, on Jan 27 2016 - 13:54, said:

 

 ...

  A blow out is defined as a result of between 15-0 and 15-4. I simulated 10,000 battles and the probability of each blowout score was determined.

ISNomads: So 10, 11, 12, 13, 14-0 is not a blowout?

 

 For this first example the teams are exactly balanced.

ISNomads: What does this even mean anyway? Skill, vehicle class, tier, exact vehicles, player skill level, spawn location, HP pool, armor, mounted equipment and consumables, ammunition type used, player mood? Are you using the flawed XVM formula? Are you doing a straight statistical analysis without regard to the many variables?

...

This simulation shows 10000 battles between two teams that are significantly unbalanced.

ISNomads: Ditto, what does unbalanced mean?

 

Does this fall under the saying "87% of statistics are made up on the spot"?

 

Even Slava has said that MM needs some work because teams are not balanced as they should be by the proper variables. Then you reach the conclusion that everything is fine based on this test?

 






1 user(s) are reading this topic

0 members, 0 guests, 0 anonymous users