Hello everyone,
as you know, we now have a new Clan Rating available. This post is based on the Russian post , explaining the formulae of the clan rating. I haven’t seen WG EU translate this to English, so… here goes.
The basic formula is this:
where:
CR – clan rating
SR – average personal rating of clan members, calculated like this:
SRi – the value of personal rating of a player i-th player at the moment of the calculation
m – number of players in clan at the moment of the calculation
TR – team rating (see below)
C – coefficient, regulating the influence of SR on CR (currently is 1)
Team rating is calculated like this:
Its components are as such:
bc – average daily amount of battles per one clan player (no matter what type of battle it is or whether it happens on global map)
bCoeff1 – balancing coefficient, default number is 7
mCoeff1 – weight coefficient, default number is 2000
cc – number of players in the clan
bCoeff2 – balancing coefficient, default number is 25
mCoeff2 – weight coefficient, default number is 2000
t10 – average amount of tier 10 vehicles (frozen tech is also taken into account) per one clan player
bCoeff3 – balancing coefficient, default number is 5
mCoeff3 – weight coefficient, default number is 2000
eGK – Elo rating of the clan, calculated based on the Global Map performance like this:
wher eGKni is the elo rating of the clan on global map number (i)
bCoeff5 – balancing coefficient, default number is 2000
mCoeff5 – weight coefficient, default number is 3000
Cpos – clan place in campaign rating
bCoeff6 – balancing coefficient, default number is 300
mCoeff6 – weight coefficient, default number is 2000
After the campaign ends, the value of the f(eC) gradually decreases using the following formula:
where X ϵ [0;30] is the amount of days gone by from the moment of the campain end.
The calculation of Elo rating on Global Map
Step 1
As an example, we will use a battle between clan A and clan B. Even before the battle starts, an expected result of the clan battle is calculated:
where:
Ea – the expected result of the battle for clan A
Ra – current Elo rating of clan A
Eb – the expected result of the battle for clan B
Rb – the current Elo rating of clan B
Example:
Step 2
After the fight is over, the amount of rating points each clan recieves is calculated:
where:
Ka – coefficient for clan A, based on the current Elo rating
Sa – amount of battle points*, recieved by clan A as battle result
Ea – expected clan A result, calculated as a part of step 1
Kb – coefficient for clan B, based on the current Elo rating
Sb – amount of battle points*, recieved by clan B as battle result
Eb – expected clan B result, calculated as a part of step 1
* Win – 1 point, Draw – 0,5 points, Loss – 0 points
Example – let’s assume clan A won the battle.
Step 3
Elo rating of both clans gets updated
Where:
Ra – Elo rating of clan A before battle
Pointsa – amount of points recieved by clan A in battle
Rb – Elo rating of clan B before battle
Pointsb – amount of points recieved by clan B in battle
Example:
Rating of clan A was rounded to the nearest whole number
Rating of clan B dropped to 0, since the Elo rating is supposed to be within the range of 0 to 3000
let the clan stat padding begin!
The statpadder clans are crying red tears, they are so sad that WG doesnt acknowledge their excellence at tier3 and 5 tanks as top clan :D
You can only pad your ELO by winning against higher ELO opponents (only works to a certain point) and only if you are good clan that doesn’t fight CWs much.
Can someone put this into idiot language so I can understand it please
Just remind they’ll give an efficiency rating to clans which depends on players performances in battle and CW :p
And some others things like numer of Tier 10 you have researched,..
With tons of calculum.
play good and have lotsa T10s? or have good players with lotsa tier 10s?
so this clan
http://worldoftanks.asia/community/clans/2000002078-VPA/
which has 792 tier 10s will be OP CR
Most of it is averaging stuff and appointing some coefficients to see which parameter is more important.
The interesting part is the ELO calculation, that’s quite a nice idea for a CW battles. If you are an outsider and defeat a strong clan, you get more points than when you defeat a weak clan. Similarly if you are a strong clan and lose, you lose more rating if it was with a loss to a weak clan.
So the ELO Rating is one of the better things WG implemented? Giving weaker clans a better chance while making strong clans clubbing weaker clans less effective in influencing the rating?
Getting more rating does not mean weak clans get a better shot at winning. It simply represents the value of a win or loss better than a flat reward for winning. Imagine a top clan doing landings only simply to rack up clan rating. With a flat reward system, that would be much more valuable – if you care at all about this clan rating, of course.
+1 (lol)
It’s easiest if you try your brain, however if you don’t know enough mathemathics (how sums, exponents, functions work) then you are screwed and need to learn that before understanding it completely. Basically whoever wrote those formulas uses things like natural exponents, 4th roots to make effects of some things increasese or decrease faster or slower – not to make too big gap.
Those “mCoefficients” are quite important, formulas for them are all very similar, writer/designer of formula wanted to give different importance to things like (ELO rating most, clan wars campaign result second but just little above next things, and then some things like how active players in clan are, how many top tiers the clan has and how many players).
ELO rating is basically changing relative to how much each clan wins (it increases more for underdog or new clan if it’s underdog or new clan winning against good clan or increases slower for good clan, when good elo clan wins against bad elo clan), when it’s two similar clans and one wins it gives it “average” increase in elo (if both clans are equal is looses same as other gains).
It is not same as efficiency/performance, you don’t need performance or efficiency to win, it’s just winning that counts. How you get to it doesn’t matter (it can be cap or some tactical deception etc.).
Ok, we have ANOTHER exact formula.
What we do not have is skill MM.
And never will. Stop crying about it.
I will cry about what I want, trust me.
You DO NOT want skill based MM at all.
IF (they never will) they were to implement this, you would not be jumping onto games every 20 seconds after hitting battle button.
Depending on the population of server you could get super long wait times, THEN you will still end up in the same type of battle after the internal timer is exceeded and it relaxes the criteria to fill the battle.
Also, the Tomatoes (YOU TOO were one once upon a time) would take MUCH longer to “Learn” how to play their respective tanks in game, effectively making the game bad, for longer.
Are you sure u want skill based MM?
Personnaly: no.
I just want different team battles max Tier.
(4,6,8 and why not 5 and 7 :P)
Yes.
I do not want to do more damage than the rest of my team every five battles.
I do not want to play with players, who turn off the minimap or ignore it.
I do not want to play with players, who have 30 000 battles and 29 900 of them with bot.
I do not want to play 3 minutes long battles, where after 90 seconds half of my team is dead.
May I continue?
Why is it so hard for WG to actually create a different ladder where you would be matched with equally rated players?? :|
It isn’t hard for WG to do it. It’s just a godawful idea that should never be done for the well-being of this game.
Tell me why.
how about introducing the skill MM in a new mode? in this way everyone will be happy. :D
Because it will only punish good players. You play good? Let’s toughen up enemy team with better players. If you go with the “balanced teams” enemy will always get someone good to compensate, basically cripling your chance to win because of it. League system, making the teams of e.g only good players? If you are not among the very top players you won’t be able to carry the team at all. Being aggressive? Big no-no, because those enemies will punish you. Ever seen ESL? Cautios play, every damn mistake will cost you and they even have commanders to coordinate the team, so it is easier for them to attack than for random team.
Not to mention the big hit for XP and credit income. You’ll win only 49% of battles, most of them without some significant damage, your allies are not dumb and will do their share too.
Of course if you are 40% bot, it will be a heaven, you’ll win 49% battles…
ok then no skill MM for you noobs… how about fixing the MM? im tired of these rigged matches with 8 heavies and 2 meds (the rest in tds and spg) vs 3 heavies and 7 meds….
or when they balance tds with heavies…or tds with meds…or meds with heavies….or teams with 13 tds in both sides…
I think you are not able to win the game alone. And it is not important how high skill you have. You are still alone vs. 15, it is unwinable.
And the other important thing – skill MM will significantly reduce the pedobearing.
Of course you cannot win alone. What I meant, you can hold that flank while being outnumbered or win 1 vs 2 situation now. Not that possible if you are facing as good players as you are.
And I am not sure about pedobearing. I’ve seen quite some players with 60% WR, all because most of the battles they have are on T18, LTraktor, SU-26 and so on, while they have 45% on T6+ Not sure how would MM treat them…
skill MM = massively increase queueing times, then there’s difficulty in determining skill factor for player. Just imagine warthunder like queue or worse (few minutes) for high and low end skilled players, it might also take some randomness part of game out which is what makes it interesting and good.
Everyone’s stats would even out to the same exact value. A tomato against another tomato can do as much as an unicum against another unicum.
Still, this could be solved by including average difficulty of matches in stat formulas. Or perhaps people could be measured solely by the league they are playing in.
Another problem is that a person may not be equally skilled in various classes of vehicles, let alone various tanks themselves. Different brackets for each class or for each vehicle? How many battles would it take to assess player’s ability then? Or what about sharing an account with a kindergarten kid – it feeds for X battles, gets lower bracket, you rock tomatoes for X battles, get higher bracket, then the kid plays again?
All you opponents of skill based MM make wrong assumptions.
If WR is important factor of matchmaking, then WR will become less differentiated, but will not become uniform.
Why? Imagine everyone has 49% WR. Then teams are random like today. So better players win more often.
Every time you have win ratio above your real skill level your team will be at slight disadvantage and so on.
Change would be from 40% WR to 47% and 63% WR to 53.
The upside however – less battles that have a score 7:1 after 3 minutes.
About increasing queue times – not neccessarily. All depends on how restrictive the skill factor will be.
Imagine such implementation: 30 tanks are chosen to participate in a game like today. Then top tier tanks are divided so the skill on both sides will be equal, then middle tier and lastly lowest tier.
The only delay would be a fraction of the second after players are already getting into the battle.
Of course it would be far from perfect implementation, but still would decrease number of curb-stomp battles.
However – all that said, skill based MM is not worth implementing.
Even with equally skilled players and equal tanks on both sides, luck factor (I don’t mean RNG but simply who will go where) decides too often so change wouldn’t be noticed as much as we would like too.
The amount of skilled vs non-skilled players mean that if you want an equal matchup, your queue time will easily get into hours instead of seconds. According to the WN8 color scale image, you’re in the top 5% of all players if you’re blue, top 1% if purple and top 0,1% if dark purple. Hell, even a 1600 WN8 player is better than 89% of the entire playerbase.
Then you also have to split these people across the various tiers, CW, TC and so on (in terms of removing them from your specific queue). Even if you want something that adds up to a certain number like WN8 (or whatever WG would come up with), you still run into similar problems as you have today: You would curse the presence of unicums in most matches because they are roughly equivalent to two 1250 WN8 players, whom are still better than 75% of the in-game population. If he fucks up without doing thousands of damage and getting a few kills, your team is boned by default.
In essence, a skillbased system only increases queue times by a metric of hundreds (if not thousands) but provides relatively little in return as almost every team you see will still be full of reds and yellows (they make up 75% of the player base). On top of that, battles with very good players often end up being decided faster than battles with bad players because a single mistake or lucky ammorack can send your side back significantly enough to snowball out of control.
Unofficial statistics: 95% of players have less than 1000 battles. This is the playerbase core.
Get two good mates, make platoon, do the damage, get some resistance medal, get full XP for loss if your team fails, profit.
We dont need skill based MM. Only tomatoes need it so they wont get pwned every game. But what the hell, they can bot to WTE 100, arty or 183 and do 1500dmg/game anyway so I guess they are happy as well either way.
Could you imagine T10 tomato battle when skill based MM is introduced? 8 TDs 4 arties and 3 E100/IS7. In less than two minutes all heavies would die and than the rest would camp until the end at their bases. Very amusing.
But anyway as long as tomatoes can report stuff to FTR so SS can fap over unicums, everybody is happy.
I hear a lot of talk about skill-based MM. I have never, ever seen a mathematically literate proposal about how to measure skill in such a system.
After 2nd or 3rd black part I gave up. Simply don’t care.
The only funny thing is that WG doesn’t want to use any rating for matchmaker, but puts one rating after another to the game.
Also doesn’t want to show own rating in battles, because “everybody use XVM” – same as for clans, if somebody wants to know clan strength, will check players’ skill and number of their X tiers, won’t care about such rating.
Lets see:
PR included … BS
Number of battles during a certain time frame … BS
Number of t10s … while important for a clan still BS. A bot clan can have hundrets of t10s and still suck.
place in campaign … BS
Elo rating – good
My verdict: BS
Just go to WoTlabs.net or noobmeter.com and compare clans there, which includes number and type of t10s and avg rating in WN/Eff/PR. That aside no rating is able to measure the skills of the FCs and the teamwork within a group of 15 players.
but Elo Rating for clan – excelent idea.
the only trouble I see is the Ka and Kb coefficients in step 2
looking how eGM changes for some top clans I think it may be too large and ranking jumps too much from day to day.
Maybe lower Ka/Kb would make ratign adjust to changes too slowly, but too big will make rating jump to much so looking at this rating value for 1 day only will have no meaning
Place in campaign is not necessarily BS, it’s not easiest to get a good ranking and practical achievement because there is a lot of competition. Similar for battles in time frame, if you have 0 for weeks your performance might drop a bit. And player PR works for high end players (if you discount team battle PR farmers). This is like clan CW rating, not how much your clan is stronku and efficient in random bob clubbing, which is really irrelevant for winning, there are different styles of play (one for just winning, highest XP, highest damage, not using gold ammo in light tanks for example etc.) . WoTlabs or noobmeter don’t tell you what kind of team you are dealing with (whether it has good FCs, teamwork, tactics, how do players perform against better players etc.).
place is campaign works only for a month decreasing linearly by 1/30 every day so it’s not a big deal
PR is stupid in itself
number of tier 10 tanks doesnt really mean much above some number
number of battles lately says sth about activity of players of this clan, so clan with a lot battles played lately is more likely to have 2 or 3 teams ready at the same time.
But eGM is the only thing worth anything there
Clan Rating Formula – Easy .
( Number of clowns in clan + have many years clowns from clan fap on wot + number of hours per day all clowns from clan nolife in wot / IQ ) * have many euros clan did donate to Serb = Clan Rating
That’s your best trolling?
It’s nice that their new rating system was able to get the math to make the top clans get better ratings. I am so glad that WOT works so hard to keep the big cheating monster clans happy.
Why are you fucking retarded like this? Give me one fucking proof any of those clans are cheating. Otherwise stfu and gtfo pubbie. Jelly moron who can’t grasp the most simplest game mechanics, therefore everyone better than me is cheater.
Does anyone know how do they get Ka and Kb?
Is it (unknown coefficient)*(clan’s ELO rating) or something else? Where can we get unknown coefficient if that’s the case?
I find it interesting that they don’t give us the C factor in the (SR*C)^4 bit of the primary (first) equation.
But it doesn’t matter. The reason they take each of the two terms to the fourth power is to give whichever one is even just a little bit larger most of the influence on the final result. I explain how this works in more detail here:
http://forum.worldoftanks.asia/index.php?/topic/42767-the-clan-rating-feature-is-now-available/page__st__20__pid__1041247#entry1041247
I’ve done a quick, back-of-the-envelope analysis of the top 200 clans from the SEA server by dumping them in to a spreadsheet and just graphing it out. I didn’t need to go any further that that; the obviously close correlation between average personal rating of clan members and clan rating, and the obvious lack of correlation for every other figure in the chart, make it clear that the SR*C term dominates to the point where everything else is basically unimportant. You can have a look here (scroll to the right to see the chart):
https://docs.google.com/spreadsheets/d/15GV_WRMyAhEV1veeckt92LrsCm1vnPWIQ1hKmh7ZPRI/edit?usp=sharing
TLDR: You want to make your clan rating higher? Don’t bother playing CW or anything like that. Just start kicking members out of your clan, starting with those with the lowest PR.
How simple !
And there was me thinking: ” The clan rating formula must be very complicated ! ”
;)
WOW, Einstein is envious on WG.