The site

I have made this site in order to keep track of the playstrength of players of the online tank action game "World Of Tanks".

In order to get a reliable rating, the best thing to do is to run the client, and by doing so submit your results to the site. If you do not, you'll only get a rating if enough of your opponents or team mates submit your matches.

The "sup" part of the name stands for "Statistical User Proficiency". Ratings are based on win-loss statistcs in individual matches for a given user and the purpose is to quantify the users proficiency, relative to other players at a given point in time.

The site is currently in extreme alpha, meaning that statistics can change arbitrarily, site functions come and go and that the sup-score is not to be trusted (at least not more than efficiency or Win/Loss).

Ratings

Rating is calculated based on actual match results, as sent to the site by a client program run on your computer. The client reads .wotreplay files, extracts relevant data and submits it to this site for analysis.

The sup-rating score is calculated based on a modified Glicko system, and is intended to be something like the "TruSkill"-system made by Microsoft. The main modifications made is support for multiplayer games and WoT-specifics, like weighing of different tanks importance in a battle.

Rating is the relative play strength that sup guesses a player has in relation to the opponents he/she has met. The higher the rating, the stronger the player is. Rating is the basis for all top lists on this site.

Deviance is a measurement of uncertainty that the rating set for a player is correct. The lower the deviance, the more certain we are that the rating is revelant. Generally, a player that has submitted many results to the system gets a lower deviance. Deviance also increases with time since last sighting, meaning that if we have no results for a given player, our uncertainty as to his/her proficiency increases.

For more detailed information about the glicko system, check out its homepage, or for a slightly easier read, check the wikipedia entry

Advantages compared to win/loss and efficiency

Whereas it might be too early days to draw any firm conclusions as to the usefulness of the sup-rating, the system is designed to be an alternative, or possibly an addition to what is currently achieved by considering the win/loss or efficiency of a player. What does the Statitical User Proficiency rating do that the aforementioned do not?

Reduce impact of bad teams - When playing with teams of low proficiency, you are expected to lose. This means that the impact of losing with such a team will have negligable effect on your rating, winning with one will mean a big boost to your rating. So when you start a game and notice that you are playing with complete vegetables, it's a possibility rather than a chore.

Reduce impact of good teams - When playing platoons or company battles, it's easier (provided you team with good players) to attain a high win/loss ratio. This skews the w/l in favour of playing platooned, and encourages stat hunting with other good players, but may also lead to misleadning statistics. Since the sup-rating takes into account the skill of the entire team when doleing out rating changes, you'll have no bonus for playing with skilled comrades - Wins will give little change if you play in a highly skilled team, losses will mean a big hit to your rating.

Old sins go away - The glicko rating system used as foundation for the sup-rating gives lesser importance to old data. This means that your newest matches will have the largest significance and that old matches will have a very small influence (if any) on your rating. You are no longer punished, statswise, for your youthful misadventures.

Rewards macro, disregards micro - Being based on wins rather than kills, damage, cap points, def points etc. the sup-rating rewards team play and tactics. Winning is everything, micro skills are disregarded except when it comes to their contribution to winning (brave last stands count for nothing - neither does capping a few percent before the last kill).

Limitations and caveats

Of course, no system is without faults, and WoTsup has several. A few of them rather annoying. I'll deal with design limitations rather than with bugs, since bugs fluctuate over time too much to have a fixed text about them here.

Skew in submitting players - The populaton of players that submit data to the system is unlikely to be representative of the game as a whole. This means that a lot of statistics will exhibit bias towards, for example higher tier matches. This needs to be taken into consideration when viewing data other than raw rating.

Plans for the future

Various stats. Of course there are many things one might do with a bunch of data. An initial example of that can be seen on the "Match statistics" page. More things of that vein are expected - If you have any special wishes, don't hesitate to send me a message on the official forum ("me" in this case being Erapin).