Jump to content
GehakteMolen

Popularity of class in relation to wn8

Recommended Posts

Hello,

 

While reading a little here, i constantly see people using data sets to obtain information about X or Y, these data set(s) are also used to calculate for example the baseline of wn8 (if i understood all ok).

 

But do we know if the used population comes (any) close to the ``real`` population?

 

wot-news has updated their recent stats again, so i could copy past all data for past 4 weeks for Ru, EU and NA server into excel and calculate popularity by class / nation, and the differences are quite big between servers. This means a data set for NA players will probably be really different from a EU or RU dataset.

Both images contain the share of vehicle per class, to on EU server 12,3% of all tanks as russian heavys and 1,59% are UK mediums.

 

The coloring on the first image is from most popular till least popular on server

Pop_1.png

 

 

Second image shows which class is on which server the most and least popular, EU server is almost always middle ground, while RU and NA have either the most of the least of a certain nation / class

Pop_2.png

 

Some example:

 

- On NA server RU heavys and US TDs are almost equal popular,

- On EU there almost 2x more USSR heavys,

- On RU there almost 3 times more RU heavies (15,94 vs 5.68)

 

Tier distribution ofc also matters, but suchs massive shitfs in played vehicles will have a significant influence on the outcome

of battles, aswell as kill / dmg / cap distribution, and those kinds of data are used to base ratings on...

 

While xxx dmg might equal to YY % of win(chance) on NA, it can / will be totally different on RU or EU server....

 

This leads to 3 conclusions:

- The home effect is really strong, on RU, USSR is by far the most popular nation, on NA its US on EU its Ger (but less big gap) this means most likely that bigger also start with those nations, so those tanks suffer from worst drivers

- Nobody plays the small nations, China and Japan, these drivers will thus most likely be not equal as the average player

- RU server has ridiculous much more games as NA and EU, which in theory means that a rating should be based on RU server, compare games played for the servers:

RU: 1.345.986.485 (80.1%)

EU:    271.896.805 (16.2%)

NA:      63.054.523 (3.8%)

 

Basing a rating on NA, with a vastly different tank-selection, different mm due to less players and a 20x smaller population means that a rating which works perfect for NA almost certainly wont work for RU server

 

Its like using a small village as benchmark to say something about a whole country, while it might be fairly accurate sometimes, it can also be totally off...

 

ps: This is for the rest not really a suggestion or so, i just wanted to show the significant differences between servers, also tier popularity differs quite a bit from server to server for example

 

edit: tier spread:

Pop_3.png

Link to post
Share on other sites

Expected values as well as the formula are based on NA, EU and RU clusters as far as I know. Quite possibly others too. Expected values come from vBAddict, and the formula comes from Noobmeter.

Link to post
Share on other sites

Expected values as well as the formula are based on NA, EU and RU clusters as far as I know. Quite possibly others too. Expected values come from vBAddict, and the formula comes from Noobmeter.

 

But does VBaddict data match these numbers? (these also have a flaw, its from past 4 weeks, to a big ``flavour of the month``)

 

VBaddict users are most likely totally different from the ``overall user``

 

ps: and ``other`` clusters doesnt matter much, RU is 5x bigger as EU and 20x bigger as NA, ``we`` are totally insignificant. If you combine the 3 servers and compare most popular overall vs most popular RU, the numbers barely change....

 

past 4 weeks RU players played: 214.484.354 games in russian heavys, all NA players played together: 63.054.523 games, ergo, the most played class on RU server is played more as all tanks on NA combined times 4....

Link to post
Share on other sites

The statement "nobody plays the small nations" could be incorrect: as those nations are small you will have less tanks to play, ie 1 tier 8 medium on japan against 2 tier 8 mediums on ussr and germany.

If you do not differentiate it by class it's even worse, because the bigger nations have a greater numer of tanks to choose from. So 1000 battles in german tanks are a lot "less" than a 1000 on japanese tanks, if you get my point.

 

With this comment I would like to point out that your data is biased, you should have averaged the battles played in a nation via the number the tanks in that nation (10 tanks in a nation? divide it by 10; 100 tanks in a nation? divide it by 100, for example). The results might differ a lot with these computations, making the results go a bit better for nations such as china, france and japan.

Link to post
Share on other sites

The statement "nobody plays the small nations" could be incorrect: as those nations are small you will have less tanks to play, ie 1 tier 8 medium on japan against 2 tier 8 mediums on ussr and germany.

If you do not differentiate it by class it's even worse, because the bigger nations have a greater numer of tanks to choose from. So 1000 battles in german tanks are a lot "less" than a 1000 on japanese tanks, if you get my point.

 

With this comment I would like to point out that your data is biased, you should have averaged the battles played in a nation via the number the tanks in that nation (10 tanks in a nation? divide it by 10; 100 tanks in a nation? divide it by 100, for example). The results might differ a lot with these computations, making the results go a bit better for nations such as china, france and japan.

 

thats true, i will divide the gamed played by amount of tanks, and see what that gives.

 

EDIT:

 

I divided total amount of games (in class / nation) by total amount of vehicles in that class / nation, then i compared how much those class got played compared to the average (total games in 4 weeks / total tanks), so on EU server, russian heavies got played 2.44 times as often as ``the average tank``. (its a bit hard to read, but it contains quite nice info imo, see below)

 

Pop_4.png

 

The bottom line contains how much more popular as ``average`` a certain class is (on EU and RU heavys get played 70% more as ``expected``, while the most right column contains the ratio of nations.

 

If there are 5x more german as japanese tanks, then germans should be played 5x more, (then both would be 1) germany is more popular while japan is less popular

 

This shows nicely that:

- On EU server heavys and TDs are more popular as the rest and that the ``main`` nations get played more as the small 4)

- On US server heavys are less popular while US is by far the most popular, ger and ussr tie, france is not really popular (equal to china)

- On RU server they really like USSR tanks and they hate brits, french to suprising well (equal to germans)

 

Since RU server matters 10x more for WG as other servers, this leads to some questions / observations:

- Its logical WG doesnt care for brits, RU players dont care for brits

- Why no more french tanks? They get played just as much as german tanks, which is amazing, since france has no ``heavy heavys``, and thats the most popular class on RU server

- Why doesnt WG (re)balance french TDs? nobody plays those things, they are equal popular as japanese light tanks, a light tank line which ends at tier 4....

 

but when looking at the distribution more question can be asked (and aswered)

Link to post
Share on other sites

They did re-balance the French TDs. They absolutely neutered the tier 10 Foch, which removed any incentive to grind the line. While I personally enjoyed the tier 8, I haven't even purchased the tier 9, since they nerfed them when I was still grinding through the tier 8.

 

The low tier French TDs aren't really "keepers" either, since none of them are "best in tier". The "best" French TD may be the tier 3 premium one.

Link to post
Share on other sites

Glad that something changed since you introduced the "average tank", how you described it.

That's a more unbiased and radable table :)

It's interesting to look at the data, also because the differences between the nations are now not that big, even though the predominance of USSR in the RU server is still very pronounced, like the other predominances you saw in the first post.

Sorry if I made you correct the data, but now I think they might be more correct and readable

Link to post
Share on other sites

Glad that something changed since you introduced the "average tank", how you described it.

That's a more unbiased and radable table :)

It's interesting to look at the data, also because the differences between the nations are now not that big, even though the predominance of USSR in the RU server is still very pronounced, like the other predominances you saw in the first post.

Sorry if I made you correct the data, but now I think they might be more correct and readable

 

Well, a lot of red stays red, especially on EU server it seems to make little difference, on NA server some stuff does change (like RU heavys, from much higher to almost equal to US)

Link to post
Share on other sites

ps: and ``other`` clusters doesnt matter much, RU is 5x bigger as EU and 20x bigger as NA, ``we`` are totally insignificant. If you combine the 3 servers and compare most popular overall vs most popular RU, the numbers barely change....

 

Are those real numbers?

 

Registered accounts have a ratio of roughly 3:2:1 RU:EU:NA. RU does also have a higher proportion of valid and active accounts, but 5x and 20x seems a little unlikely. Maybe if RU has a very high quantity of game-spamming bots.

 

Personally I'd love to collect more RU data, but tanks/stats is much buggier there than on the other servers so I can't get a decent quantity of interval data. Interval data is the best data. I suppose I could use account/tanks for cross-checking.

Link to post
Share on other sites

So... for ultimate balance discussions we must take stats from non native servers to see how they fair to others.

 

USA tank stats from RU server, USSR tanks from NA server.....

 

because yolo statistics (i know it doesnt work like that, just jk)

 

But on a serious note, EU does seem to be the least biased and so maybe the best to look at overall tanks statistic comparison.

Link to post
Share on other sites

Are those real numbers?

 

Registered accounts have a ratio of roughly 3:2:1 RU:EU:NA. RU does also have a higher proportion of valid and active accounts, but 5x and 20x seems a little unlikely. Maybe if RU has a very high quantity of game-spamming bots.

 

Personally I'd love to collect more RU data, but tanks/stats is much buggier there than on the other servers so I can't get a decent quantity of interval data. Interval data is the best data. I suppose I could use account/tanks for cross-checking.

 

They are from:

http://wot-news.com/stat/server/ru/norm/ru

 

Peek online is also roughly 1:5:20 i think, EU has max of 250k, Ru 1.1 million and NA 50k?

Link to post
Share on other sites

This shows nicely that:

- On EU server heavys and TDs are more popular as the rest and that the ``main`` nations get played more as the small 4)

- On US server heavys are less popular while US is by far the most popular, ger and ussr tie, france is not really popular (equal to china)

- On RU server they really like USSR tanks and they hate brits, french to suprising well (equal to germans)

That data also shows that on NA and RU, heavies and TDs are more popular than the rest.  

Heavies are less popular on NA than on the other servers but still the most popular class overall.  

TDs are the more popular on NA than either EU or RU.

 

RU has nearly as much arty as light tank play.  

Link to post
Share on other sites

Ok, I underestimated the difference in account activity between servers. Cross server sample, should be accurate to within a couple of percent:

 

RU: 34.3m accounts, 4.3m active with 8900 average battles

EU: 22.9m accounts, 1.1m active with 6300 average battles

NA: 11.2m accounts, 0.27m active with 5700 average battles

 

Activity defined as having played within the last two weeks.

 

 

Edit: Conclusions on having read the data through properly:

 

1. The class gap between RU and EU is very small. Lights and SPGs at least have relatively similar hp, so that shouldn't throw the average damage out by much. I suspect both classes perform better when there are fewer of them in a game, so lights may do slightly better on RU and SPGs slightly worse. Not a huge difference though.

 

2. National tank popularity shouldn't affect rDmg/rFrag/rSpot/rDef by much, although for extremely popular tanks it does have a flattening effect on expected winrates due to MM tank matching. Hence the russian bias will appear slightly weaker on the RU server.

 

3. The tier 10 popularity gap is a problem, especially with NA data. That will have a knock-on effect right the way through the matchmaking and may have a significant influence on cross-tier results.

 

 

There are other potential differences between servers that aren't covered here:

 

1. Tier skill distribution may be very different between servers, particularly if Russians frequently bot their way up to tier 10. Probably a stronger effect than relative tier popularity.

 

2. Metagame differences. Stats like rDmg are strongly dependent on factors such as cap rate and survival rate. Some classes may also be more effective on a server with a more or less aggressive metagame.

Edited by RichardNixon
Link to post
Share on other sites

2. National tank popularity shouldn't affect rDmg/rFrag/rSpot/rDef by much, although for extremely popular tanks it does have a flattening effect on expected winrates due to MM tank matching. 

 

Could you explain this a bit more?  Popularity might not affect those stats much based on amount of gameplay, but if they are played earlier in player career the average player skill would be lower and result in lower rStat.

 

Also, what do you mean by MM tank matching?

Link to post
Share on other sites

Could you explain this a bit more?  Popularity might not affect those stats much based on amount of gameplay, but if they are played earlier in player career the average player skill would be lower and result in lower rStat.

 

In that case they'd be wrong on any server, just differently wrong. That's why I'm using interval methods now.

 

On the other hand it may be easier to get data for specific rare tanks on other servers.

 

 

Also, what do you mean by MM tank matching?

 

The rule as quoted in the wiki:

"The number of the same vehicle on opposing teams cannot differ by more than 1, except when the vehicles are in a platoon."

 

The end result is that a highly-popular and overpowered tank is likely to be intentionally matched against the same tank, which may cancel some proportion of its advantage if the tank is overpowered for its average player. The old T18 and KV-1S are examples.

Link to post
Share on other sites

Ok, I underestimated the difference in account activity between servers. Cross server sample, should be accurate to within a couple of percent:

 

RU: 34.3m accounts, 4.3m active with 8900 average battles

EU: 22.9m accounts, 1.1m active with 6300 average battles

NA: 11.2m accounts, 0.27m active with 5700 average battles

 

Activity defined as having played within the last two weeks.

 

 

Edit: Conclusions on having read the data through properly:

 

1. The class gap between RU and EU is very small. Lights and SPGs at least have relatively similar hp, so that shouldn't throw the average damage out by much. I suspect both classes perform better when there are fewer of them in a game, so lights may do slightly better on RU and SPGs slightly worse. Not a huge difference though.

 

2. National tank popularity shouldn't affect rDmg/rFrag/rSpot/rDef by much, although for extremely popular tanks it does have a flattening effect on expected winrates due to MM tank matching. Hence the russian bias will appear slightly weaker on the RU server.

 

3. The tier 10 popularity gap is a problem, especially with NA data. That will have a knock-on effect right the way through the matchmaking and may have a significant influence on cross-tier results.

 

 

There are other potential differences between servers that aren't covered here:

 

1. Tier skill distribution may be very different between servers, particularly if Russians frequently bot their way up to tier 10. Probably a stronger effect than relative tier popularity.

 

2. Metagame differences. Stats like rDmg are strongly dependent on factors such as cap rate and survival rate. Some classes may also be more effective on a server with a more or less aggressive metagame.

 

Number 2 might have a lot of influence, if we compare the number of bat chats (relative to the total amount of tier 10s) its:

RU:       10,123 %

EU:          7,902 %

NA:          7,047 %

 

So in 5 games on RU you meet 15 bat chats, while on EU 12 and on NA like 10, thats quite a difference (also because bat chats have fairly much influence, especially if there are 2 or more in 1 team)

Link to post
Share on other sites

I'm kind of curious now as to where the bottom range of the 'top 1%' of players for RU, EU, and NA are, though failing that i'd be curious as to the wr/wn8 values of the 43,000th, 11,000, and 2700th ranked person for RU, Eu, and NA respectively. 

Link to post
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.


  • Recently Browsing   0 members

    No registered users viewing this page.

×
×
  • Create New...