This page is still being rewritten (last updated 07-02-2015)...

  • General information

  • The formula


  • What is this website?

    This website was founded in 2011 to help players of one little clan navigate through the dangers of Global Map, to help understand which clan is a weak one and should be assaulted and which one is too strong to be attacked.

    Eventually the website grew up and in the end became the most popular clan portal on the Russian Cluster and won the second prize for "Best functionality" on the Wargaming Developers Contest.

     

    Why is the rating better than the official one? Why is it better than the others?

    According to a friend of mine who works at Wargaming Belarus, the official rating was created loosely based on my rating. Although they grasped the general idea, they failed to implement it properly. As a result their rating shows the worst result among the other ratings. What kind of result? Well, it is quite easy to determine the performance of a rating by applying every rating system to clans that participate in Clan War landings. Ratings that could guess more results correctly got better points. And yes, there was a statistical study for every major rating/ranking system that proved that the official rating is the weakest among all the others.

    What is the difference between this ranking and all the others? This ranking was created having in focus not players, but clans. Other rankings use a single player formula, they calculate some value for each player, then one way or another they try to get an average value and then they rank clans by this average value.

    Although this approach is not a failure it doesn't take into account lots of meaningful variables my ranking system uses, variables that allow my ranking system to outperform its competitors in the task of guessing results on the Global Map.

     

    The gist.

    Before digging deeper, it is vital to understand the gist of the formula.

    Firstly, we must accept that tanks on the Global Maps have different value for a clan. Let's say no one uses FV4202, but every clan seeks players with Bat.-Châtillons (as of 21.01.2015). We will return to determining the values later.

    Now, we must accept that the more tanks a clan has, the more versatile the clan is. If the clan has lots of different tanks, then it can use them on a variety of maps and not suffer due to so-called tank 'freezing'.

    So, the first important variable of the rating is FIREPOWER. Firepower basically is a relative quantity of tanks in a clan and every tank adds to the variable according to the tank's value.

    Right, we have tanks, but do we have required ability to use them?

    Then it comes to the difference between players who play in random and players who play in teams. Playing in random battles requires different ability, different behavior and different SKILL. If we look at the profiles of a generic random player and a generic team player we can notice the difference between their WinRates. Player who plays in teams is able to have a higher WinRate than the player of the same ability but who plays only random battles. So SKILL must be based primarily on the WinRates of the players. A statistical study showed the formula of SKILL=(WinRate)^3*(FragsPerBattle)^2*(HitRatio) shows the best results.

    Of course, there we must take into account the tank's values that we mentioned before. No one cares about the WinRate on FV4202, because no commander would take the tank into a team battle.

    Therefore the POWER of a clan is a combination of clan's FIREPOWER and clan's SKILL, or POWER=FIREPOWER*SKILL

    Dig deeper.
    1. THE TANK VALUES

      We started by mentioning the values of every tank.

      How do we determine the importance of a tank? Well, every now and then commanders on the Russian Cluster are requested to complete a survey. The survey asks them "Please, rate the tanks from 0 to 5 according to its importance on the Global Map". The results of the survey are used in the ranking system.

      (here is the picture of the coefficients)

    2. THE LEVELING PROCESS

      Since we are calculating WinRates for every player on every tank we should bear in mind that some players who have only started to play the tank can have very unusual WinRates. For example, you bought a tank and had a very unlucky streak of 20 games without a win. Will your 0% WinRate influence clan's overall SKILL?

      Of course not. The formula has a bit that uses a linear dependence of the influence to the number of games on the tank. A player has to play 300 games on a tank to maximize the influence of the tank on the overall WinRatio of a clan.

      It's probably easier to show it mathematically than to describe.

      PlayerA has 450 battles on BatChat and 270 victories (60% WinRate).
      PlayerB has 20 battles on BatChat and 5 victories (25% WinRate).

      If we wouldn't use the leveling the result would be IncorrectWinRateOfAClanOnBatchat=(25+60)/2=42.5%
      The low WinRate of PlayerB significantly lowered average WinRatio.

      Now, let's do it right way. The formula that uses leveling looks like:

      CorrectWinRateOfAClanOnBatchat=(5+270*300)/(20+450*300)=59.99%

      This way PlayerB with his unlucky streak on BatChat doesn't influence the overall WinRate much, but with every new battle his influence will be rising.

      By the way, players with less than 10 battles on the tank doesn't participated with the calculation of the overall WinRate of a clan for that tank.

    3. ONLINE

      Image a clan that consists of 100 talented players that have allz the tanks in the game, BUT the players that rarely play. They event can't get 15 fighters to play in a ClanWars game. And now imagine another clan which players are less skillful but they play every day and always ready to participate in a team game. Of course the second clan is more powerful than the first one, but how can we reflect that fact in a rating? We have to introduce a variable called ONLINE. ONLINE is a reduction factor. A player who has been online even once during last week is considered to sill be in the game and has the online factor of 1. After seven days missing from the game the factor is reduced according to the formula (here will be the picture) and his influence on his clan starts to diminish. All the parameters of the player (number of games, value of his tanks) are being multiplied to this parameter and as a result being reduced.