Jump to content


The bell shaped curve (Gaussian distribution), otherwise known as why yelling about nerfs or buffs w...

Blanace Stats Damn stats Bacon

  • Please log in to reply
40 replies to this topic

Kebabsaurus Rex #21 Posted 16 March 2018 - 05:52 PM

    Captain

  • Players
  • 14364 battles
  • 1,326
  • [BNKR-]
  • Member since:
    02-07-2015

View PostRubbelito, on 16 March 2018 - 04:21 PM, said:

 

I basically agree with everything you've said except for this.

Skill based MM would make WR obsolete and I'm not sure that is a good way forward.

But as I said, everything else quite spot on. :honoring:

 

 

Happy to be challenged, TBH I am just trying to think of methods for improving some matchmaking i.e. less steamrolls

 

Could this be done, I dont know?

 

TBH I played my own part in many such games. One example was last night - top tier in the Defender with a platoon mate in a Fatherland on ravaged capital.

 

Only 2 tier 8s on red - we just drove forward blasting everything in our path.....they never had a chance.



Rubbelito #22 Posted 16 March 2018 - 06:09 PM

    Major

  • Players
  • 17718 battles
  • 6,032
  • [TNT]
  • Member since:
    02-04-2016

View PostKebabsaurus Rex, on 16 March 2018 - 06:52 PM, said:

 

Happy to be challenged, TBH I am just trying to think of methods for improving some matchmaking i.e. less steamrolls

 

Could this be done, I dont know?

 

TBH I played my own part in many such games. One example was last night - top tier in the Defender with a platoon mate in a Fatherland on ravaged capital.

 

Only 2 tier 8s on red - we just drove forward blasting everything in our path.....they never had a chance.

 

Yes, I assumed it was more of an advanced brainstorming than a ready made plan.

Well, first I'd like to make sure we separate balancing MM and introducing skill based MM.

They're not the same thing necessarily.

I'm all for more balance, as long as it doesn't involve skill.

What happens when you involve skill is that it becomes impossible to make progress (WR).

As soon as you get slightly better, a slightly better opposing team member will be picked as your match, making your own progress more or less invisible/redundant.

Personally, I probably want to see first what can be done with existing factors (tiers, classes, weight etc) to make the balance better, but this also comes down to game pop and queues.

Not sure I want to wait 5 min before every battle just to get a slightly more balanced team lineup.

 

But, it's not an easy task, nor an easy discussion for that matter.

 


[TNT] Scorpions

Rubbelito Youtube


Jherry Angel #23 Posted 16 March 2018 - 06:34 PM

    First lieutenant

  • Players
  • 27458 battles
  • 769
  • [BNKR-]
  • Member since:
    10-04-2014

Good post OP.

Factoid: 76.342% of statistics are made up by the person quoting them :hiding:


Edited by Jherry Angel, 16 March 2018 - 06:54 PM.


 "Whatever doesn't kill you, simply makes you ... stranger." 


Goddog #24 Posted 16 March 2018 - 06:45 PM

    First lieutenant

  • Players
  • 19142 battles
  • 542
  • [SDKFZ]
  • Member since:
    04-16-2014

View PostBrick S0lid, on 15 March 2018 - 01:22 PM, said:

All I ever wanted was for WGCB follow PCs lead where balance is concerned. Alpha nerf for tier X TDs, HESH pen nerf and waffle removal. All these things have happened on PC a long time ago and the tanks affected aren't underpowered because of it. 

 

No, the whole reason I am over here and not on PC is due to the terrible position WoT PC has been moving into. Stop that! no freaking stun mechanic 

Winner 3rd place in the  [UNOFFICIAL CHALLENGE #2] Tier V Sturmgeschutz III

 


NightOfDeaths #25 Posted 16 March 2018 - 06:49 PM

    Major

  • Players
  • 26585 battles
  • 15,909
  • [KMD]
  • Member since:
    07-16-2014

View PostJherry Angel, on 16 March 2018 - 12:34 PM, said:

76.345% of statistics are made up by the person quoting them :hiding:

 

lol

 Screw my elders!!!!

World of Tanks a game where it's 1 versus 29. 15 reds trying to kill you and 14 greens trying to support them. The yellows are your only hope. For signature and stats visit http://wotinfo.net/en/player-stats

2 & 3 MoEs

 


Brick S0lid #26 Posted 16 March 2018 - 08:18 PM

    Major

  • Players
  • 14262 battles
  • 4,403
  • Member since:
    06-12-2014

As the OP said himself data can be easily skewed by the player base. So keep that in mind if anyone is thinking that data regarding certain tanks proves their balanced. And why is it that the deathstar and waffle are played so much? You have look at the bigger picture which is hard since we don't always have all of the information. 

 

You also to consider the affect on gameplay some of these tanks have. Do they promote good gameplay or are they detrimental. Are they fun to play or fun to fight? I honestly believe this game could gain a greater semblance of balance if you let a small group of experienced unbiased players tweak a few stats. Don't need to look at a spreadsheet to tell you a few things that aren't right in this game.


Artillery punishes flankers, not campers.

 

3700+Wn8. 100 tanks with 3 MoE including 29 tier X's on my main account.

 


redshadowrider #27 Posted 16 March 2018 - 08:24 PM

    Major

  • Players
  • 31585 battles
  • 6,899
  • Member since:
    03-17-2014
I didn't like statistics the first time I had to take it.  Don't like it anymore now.  We all know RNG is based on the curve...that's all we need to know.  :teethhappy:
I'm just saying.....

x rocketfish x #28 Posted 17 March 2018 - 10:29 PM

    First lieutenant

  • Players
  • 13431 battles
  • 532
  • Member since:
    05-27-2015

View PostBrick S0lid, on 16 March 2018 - 01:18 PM, said:

As the OP said himself data can be easily skewed by the player base. So keep that in mind if anyone is thinking that data regarding certain tanks proves their balanced. And why is it that the deathstar and waffle are played so much? You have look at the bigger picture which is hard since we don't always have all of the information. 

 

You also to consider the affect on gameplay some of these tanks have. Do they promote good gameplay or are they detrimental. Are they fun to play or fun to fight? I honestly believe this game could gain a greater semblance of balance if you let a small group of experienced unbiased players tweak a few stats. Don't need to look at a spreadsheet to tell you a few things that aren't right in this game.

Very well said. I absolutely agree that unbiased players could balance this game in minutes. 

 

I think the most important stat they are hiding from us is whIch tanks are generating the most income from free xp conversion. 



Kebabsaurus Rex #29 Posted 19 March 2018 - 01:53 PM

    Captain

  • Players
  • 14364 battles
  • 1,326
  • [BNKR-]
  • Member since:
    02-07-2015

View PostBrick S0lid, on 16 March 2018 - 08:18 PM, said:

As the OP said himself data can be easily skewed by the player base. So keep that in mind if anyone is thinking that data regarding certain tanks proves their balanced. And why is it that the deathstar and waffle are played so much? You have look at the bigger picture which is hard since we don't always have all of the information. 

 

You also to consider the affect on gameplay some of these tanks have. Do they promote good gameplay or are they detrimental. Are they fun to play or fun to fight? I honestly believe this game could gain a greater semblance of balance if you let a small group of experienced unbiased players tweak a few stats. Don't need to look at a spreadsheet to tell you a few things that aren't right in this game.

 

I was going to include a discussion point about unicum/experienced players views.

 

Basically it was to say the view of these players, such as yourself, should be taken on board.

 

Overall I completely agree that just because a bunch of people pull the stats down on said tank does not mean it is broken - it is just skewed data. For example, 50% could be doing awesome and 50% could be terrible which gives overall balanced data - which is far from the case.



Perfecshionist #30 Posted 19 March 2018 - 03:00 PM

    Corporal

  • Players
  • 8796 battles
  • 71
  • Member since:
    03-28-2017

While the performance of players will fall along a Gaussian distribution, the stats of tanks do not follow this distribution, no will they unless WOT designers decide to do so intentionally. So your use of the Gaussian distribution to say nerfs and buss won't make a different is nonsense.

 

While it won't change win rates it will change the EXPERIENCE in the matches themselves.

 

If an individual tank is overpowered it is a problem for the in game experience. It does not skew win rates but players in matches that have those tanks enjoy the experience less whether they win or lose that particular match. 

 

So it is possible for WOT to shift the performance curve of players using a particular tank to better match the performance curve of players playing other tanks even if the performance curves of each individual tank follows a Gaussian distribution.



Rubbelito #31 Posted 19 March 2018 - 03:05 PM

    Major

  • Players
  • 17718 battles
  • 6,032
  • [TNT]
  • Member since:
    02-04-2016

View PostPerfecshionist, on 19 March 2018 - 04:00 PM, said:

While the performance of players will fall along a Gaussian distribution, the stats of tanks do not follow this distribution, no will they unless WOT designers decide to do so intentionally. So your use of the Gaussian distribution to say nerfs and buss won't make a different is nonsense.

 

Why?

 

Block Quote

If an individual tank is overpowered it is a problem for the in game experience. It does not skew win rates

 

Yes, it does.

 

 

 


[TNT] Scorpions

Rubbelito Youtube


Kebabsaurus Rex #32 Posted 19 March 2018 - 03:22 PM

    Captain

  • Players
  • 14364 battles
  • 1,326
  • [BNKR-]
  • Member since:
    02-07-2015

View PostPerfecshionist, on 19 March 2018 - 03:00 PM, said:

While the performance of players will fall along a Gaussian distribution, the stats of tanks do not follow this distribution, no will they unless WOT designers decide to do so intentionally. So your use of the Gaussian distribution to say nerfs and buss won't make a different is nonsense.

 

While it won't change win rates it will change the EXPERIENCE in the matches themselves.

 

If an individual tank is overpowered it is a problem for the in game experience. It does not skew win rates but players in matches that have those tanks enjoy the experience less whether they win or lose that particular match. 

 

So it is possible for WOT to shift the performance curve of players using a particular tank to better match the performance curve of players playing other tanks even if the performance curves of each individual tank follows a Gaussian distribution.

 

Sorry but I do not agree - if for example all tier 10 tanks do not fall within a certain tolerance of one another in terms of damage over time then this suggest some tanks are more OP than others: the exact situation WG states it prevents via the use of buffs and nerfs.

 

As stated above, although this may suggest balance it does not tell the full story. Data can easily be skewed by a majority of people performing badly while a small percentage perform well above average.

 

Hence why the deathstar, Waffle, and other tanks can be deemed "balanced". In summary, the more potatoes playing a certain tank will hide a small percentage doing extremely well, so objectively balanced.



Perfecshionist #33 Posted 19 March 2018 - 03:38 PM

    Corporal

  • Players
  • 8796 battles
  • 71
  • Member since:
    03-28-2017

View PostRubbelito, on 19 March 2018 - 03:05 PM, said:

 

Why?

 

 

Yes, it does.

 

 

 

 

Stats of the tanks do not follow the distribution because the stats of the tanks are not random. Whether a tank is overpowered compared to other tanks is based on the stats of the tank itself. Not the performance stats of the players playing the tank.  Designers determine the stats of the tanks, and thus they determine the distribution curve of the stats of tanks so they do not follow a Gaussian distribution. For example; they could, if they chose, make all the stats of all the tanks identical.  

 

Performance stats of players using each individual tank follow a random distribution along a Gaussian curve FOR EACH TANK. The shape of the performance curve for each tank is Gaussian but the DATA of performance of players playing each tank is NOT THE SAME AS OTHER TANKS.

 

For example:

 The number of players with these damage rates for tank A follow a Gaussian distribution; 1 player 500, 2 players 1000, 4 players 1500, 8 players 2000, 4 players 2500, 2 players 3000, 1 player 3500. 

 

Tank B might have the following;  1 player 1500, 2 players 1500, 4 players 2000, 8 players 2500, 4 players 3000, 2 players 3500, 1 player 4000. 

 

 

The shape of the curves for tank A and tank B is identical but the actual DATA that makes up the shape of the curve is not the same. Tank B is likely a better tank. The more players and the more DATA reinforcing the performance distribution of these two tanks the MORE CERTAIN that tank B is a better or "overpowered" tank.

 

Now obviously, there is more than damage that determines whether a tank if OP but if a tank has performance curves that show if performance better than other tanks consistently across measures of performance and the overall win rate for the tank is a standard deviation of more from the mean win rate it is clear the tank is more powerful in game than other tank of its tier and thus might need adjusting.

 

Adjusting that tank would change the experience of playing the game. There would be fewer of that tank making matches feel more varied because other tanks would be getting played more often. And players would feel less like the game is "unfair" when they face off with that tank. 

 

For instance the number of T-29s at tier VII is annoying. It effects the experience of playing other tier 7 tanks and it makes tier 7 matches feel less diverse with respect to the types of tanks you fight.

 

Tier X would feel more engaging if players were playing a wider variety of tier X tanks.

 

So the nerfs MATTER to the experience of playing the game.

 

 

The reason an overpowered tank does not skew overall win rates is because a specific model of tank is played by players of both sides. You are just as likely to face an overpowered tank as have one on your team. So overall win rates remain the same.

 

The win rates for players who consistently play overpowered tanks is going to be higher than a player of the same exact skill playing non overpowered tanks but the overall win rate for players in game remains the same. So when I said it does not skew win rates I meant it does not change overall in game win rate for all players.

 


Edited by Perfecshionist, 19 March 2018 - 03:42 PM.


Perfecshionist #34 Posted 19 March 2018 - 03:40 PM

    Corporal

  • Players
  • 8796 battles
  • 71
  • Member since:
    03-28-2017

View PostKebabsaurus Rex, on 19 March 2018 - 03:22 PM, said:

 

Sorry but I do not agree - if for example all tier 10 tanks do not fall within a certain tolerance of one another in terms of damage over time then this suggest some tanks are more OP than others: the exact situation WG states it prevents via the use of buffs and nerfs.

 

As stated above, although this may suggest balance it does not tell the full story. Data can easily be skewed by a majority of people performing badly while a small percentage perform well above average.

 

Hence why the deathstar, Waffle, and other tanks can be deemed "balanced". In summary, the more potatoes playing a certain tank will hide a small percentage doing extremely well, so objectively balanced.

 

I explained in the post above.

 

And choosing to disagree or agree is irrelevant. 

 

We are not talking about opinions. We are talking about Gaussian distributions. Opinions are irrelevant in statistics.



Rubbelito #35 Posted 19 March 2018 - 03:50 PM

    Major

  • Players
  • 17718 battles
  • 6,032
  • [TNT]
  • Member since:
    02-04-2016

 

You throw around a lot of fancy words. Almost as I believe you know what you're talking about..

But then you write this:

 

View PostPerfecshionist, on 19 March 2018 - 04:38 PM, said:

The reason an overpowered tank does not skew overall win rates is because a specific model of tank is played by players of both sides. You are just as likely to face an overpowered tank as have one on your team. So overall win rates remain the same.

 

 

 

..and I'm sorry, but now I know you don't know what you're talking about.

You're looking at a too small sample.

An OP tank will have a higher WR. Doesn't matter if "it can meet the same tank on the opposite side", over a large sample size it will still win more often.


[TNT] Scorpions

Rubbelito Youtube


SandM9 #36 Posted 19 March 2018 - 04:08 PM

    Major

  • Beta Tester
  • 27349 battles
  • 2,334
  • Member since:
    08-03-2013

View PostTemplarKnight75, on 15 March 2018 - 12:50 PM, said:

Didn't think anyone thought about how balancing and statistics actually worked. Just kinda bashed their heads into keyboards and screamed. Nice job although considering people didn't watch the video WG put out I unfortunately think people here won' change their mind with you logic and facts. 

I watched the video also.  Found it to be lacking.  A lot of verbage leading nowhere, but I think he explains it better..

View PostKebabsaurus Rex, on 16 March 2018 - 06:05 AM, said:

Summary of previous post

Thanks to Beta and and DSG_Angrybulldog for the video, I appreciate you posting this as I was not aware of it. I would highlight they don’t really go into their data too much and give pretty much no information on how they generate their heat maps – not that I expect them too as some of this is very likely to be confidential J In fact they present some redacted information to make this point.

 

So the summary of my previous post was to say: How does WG justify saying something is balanced?

Answer: By looking at lots of data points from thousands of games and then comparing that data with data from other tanks. BUT and here is the main point – balance is relative and can easily be skewed by human input and this is what I will discuss further.

 

So are WG lying when they say balanced?

This depends on your view point and primarily depends on what WG decide are the cut offs for saying something is balanced. It is not that the numbers do not lie – it is all about how data are interpreted and within what ranges you decide that balance falls within.

For example, if you decide that all the data should be comparable with a 1% tolerance then the game would suffer. Why? Because all tanks would be causing comparable damage, relatively, and all tanks regardless of role would be a bit boring.

So the tolerances for the game have to be variable otherwise all tanks would be causing similar damage and this would just suck.

However, if you decide your tolerances are ≥20% you are potentially introducing significant differences between tanks. This might be OK when comparing a tier 8 TD with a tier 10 TD BUT it throws up a lot of questions when comparing tier 10 TDs to tier 10 heavies, mediums, or lights.

 

So is too much balance a bad thing?

Overall I would say yes – if everything was the same the game would be boring.

 

Do I agree that everything should be balanced?

In a short answer, No.

Long answer: No. Why? because if you consider the game itself it simply cannot be balanced. Due to the tank tiers, different types of guns, different types of tanks, armour etc the game can never be totally balanced.

The only way it could be balanced would be for one tank at one tier – again this would be somewhat boring.

 

So is too little balance a bad thing?

Again my answer would be yes. Why? Because large differences lead to situations like the Deathstar, Waffle, and other tanks.

 

Do I think there is too little balance in WoT console?

Yes, to some degree and my reasons agree with what people post above – for example Brick Solid and Night of Deaths.

 

Why is lack of balance an issue, especially going forward?

  • Introduction of new tanks means that WG have to at least match those already in the game OR choose not to match them.
  • For example the new Chinese TDs do not seem to have comparable damage output to some existing tech tree tanks from other countries.
  • This therefore suggest that some of these existing tanks therefore may be OP but nothing is done about them…..
  • Which leads to the persistent questions about why these tanks are not nerfed?
  • I would also highlight here that the addition of premium tanks, primarily those at tier 8, has in my opinion also changed the dynamics of the game for players at tier 6 and 7 – not all in a positive way
  • However, I would also congratulate WG, to some extent, on nerfing tanks like the defender prior to it being introduced into the game

 

My main concern: If we get more tier 10s or tier 11 what will the stats be?

  • Leading on from the above, you have to ask if they introduce tier 11 what the hell would the stats of some of these tanks be?
  • Given the damage output of some tier 10s, many of the maps being small, and tier 10 mediums and heavies which can outspot tier 8 tanks you have to wonder what tier 11 could look like? Assuming +2/-2 matchmaking playing in a tier 9 could be more horrible that playing some tier 6 tanks at tier 8
  • Although this is a hypothetical point I am using it to illustrate one of the main issues that I feel affects the game – a lack of willingness of behalf of the devs to perform wholesale balancing when introducing new lines

 

What should have been done or what could be done?

  • My personal opinion, for what it is worth, is that WG should have rebalanced the whole game from top to bottom when they introduced tier 10 lights in 4.2
  • Rebalancing from the top down would have allowed tweaks to damage output, view range etc to be made from the top down
  • Changes could also have been made to the matchmaking system such as tanks at tier 4 and below seeing +1/-1
  • By not doing this I consider the devs are painting themselves into a corner. Why? Because if you do not remove something then you get into a situation of constantly adding something – be that more damage, more alpha, more pen, more view range etc.

 

Suggestions for what can be done in future?

  • Introduction of public test server prior to changes going live, similar to PC
  • Rebalance from tier 10 down, not from the bottom up – suggestions would include nerfing HESH pen on tier 10 TDs for starters
  • Skill-based matchmaking – currently matchmaking seems to prioritise speed of getting a game vs trying to balance teams
  • Altering the platoon system to stop fail platooning
  • Other suggestions welcome

 

 


Welcome to the forums. Where you can call people stupid, just don't call them passive-aggressive.

 

 


Brick S0lid #37 Posted 19 March 2018 - 05:56 PM

    Major

  • Players
  • 14262 battles
  • 4,403
  • Member since:
    06-12-2014

View PostKebabsaurus Rex, on 20 March 2018 - 01:53 AM, said:

 

I was going to include a discussion point about unicum/experienced players views.

 

Basically it was to say the view of these players, such as yourself, should be taken on board.

 

Overall I completely agree that just because a bunch of people pull the stats down on said tank does not mean it is broken - it is just skewed data. For example, 50% could be doing awesome and 50% could be terrible which gives overall balanced data - which is far from the case.

Yes. WGCB doesn't consider all the variables here, all they are looking at is whether a vehicle falls within the margins. 


Artillery punishes flankers, not campers.

 

3700+Wn8. 100 tanks with 3 MoE including 29 tier X's on my main account.

 


Perfecshionist #38 Posted 19 March 2018 - 06:08 PM

    Corporal

  • Players
  • 8796 battles
  • 71
  • Member since:
    03-28-2017

View PostRubbelito, on 19 March 2018 - 03:50 PM, said:

 

You throw around a lot of fancy words. Almost as I believe you know what you're talking about..

But then you write this:

 

 

 

..and I'm sorry, but now I know you don't know what you're talking about.

You're looking at a too small sample.

An OP tank will have a higher WR. Doesn't matter if "it can meet the same tank on the opposite side", over a large sample size it will still win more often.

 

I know what I am talking about. You  just failed at reading comprehension. I did not claim an OP will not have a higher win rate. I said an OP tank will not change the win rate for WOT overall. That is not the same thing. I specifically said an OP tank will have a higher win rate.

 

And I am not throwing around "fancy" words. I am using words according to their meaning. In fact, to make it easier to understand, I did not use words that have definitions specific to statistics that are not understood outside of statistics. 


Edited by Perfecshionist, 19 March 2018 - 06:57 PM.


Rubbelito #39 Posted 19 March 2018 - 07:47 PM

    Major

  • Players
  • 17718 battles
  • 6,032
  • [TNT]
  • Member since:
    02-04-2016

View PostPerfecshionist, on 19 March 2018 - 07:08 PM, said:

 

I know what I am talking about. You  just failed at reading comprehension. I did not claim an OP will not have a higher win rate. I said an OP tank will not change the win rate for WOT overall. That is not the same thing. I specifically said an OP tank will have a higher win rate.

 

And I am not throwing around "fancy" words. I am using words according to their meaning. In fact, to make it easier to understand, I did not use words that have definitions specific to statistics that are not understood outside of statistics. 

 

..or, maybe you failed in making a relevant point.

What does "overall wr" have to do with anything?

But glad you agree with me now.


[TNT] Scorpions

Rubbelito Youtube


TemplarKnight75 #40 Posted 20 March 2018 - 01:43 AM

    Captain

  • Players
  • 14784 battles
  • 1,076
  • [OJD]
  • Member since:
    12-05-2015

View Postredshadowrider, on 16 March 2018 - 03:24 PM, said:

I didn't like statistics the first time I had to take it.  Don't like it anymore now.  We all know RNG is based on the curve...that's all we need to know.  :teethhappy:

There are many things wrong with this sentence 







Also tagged with Blanace, Stats, Damn stats, Bacon

1 user(s) are reading this topic

0 members, 0 guests, 0 anonymous users