Ads displayed for guests and not donating members only. Get ad-free by donating. If you have already donated, please read here.
Results 1 to 9 of 9

Thread: Player Battle Performance Study

  1. #1

    Player Battle Performance Study

    I'd like to perform a study of players' performance in individual battles against available metrics including MM ranking of their vehicles. I have examined the replay JSON files for pre-battle stacking information but - if it's in there - I haven't found anything. Is anyone aware of a source for such a MM stacking list?

    And in the same vein, is there an established large database of replay JSON files available somewhere? I have a rather small set of my own battles but I'd like to insure statistical significance of the results with a large sample set.

    Any help would be greatly appreciated.

  2. #2
    Retired Commander Phalynx.eu's Avatar
    Join Date
    Jan 2013
    Location
    Erlangen, Germany
    Posts
    2,127
    What exactly are you searching for? MM ranking for battles? I have over 8 Mio battle results in my database, so maybe this could help?

  3. #3
    Thanks for the reply, Phalynx <The Man himself!> I'm trying to establish expected values for player battle performance on a per battle basis. One of the most obvious would be to look at battle outcome (HP,credits, etc) versus pre-battle stacking of the players' vehicles (MM assigned ranking), player WN7, or player win rate, etc.

    Things like players' global WN7s, etc are pretty easy to come by, but so far I;ve been unable to find the MM rankings for vehicles (top to bottom). I've tried to reverse engineer the list by looking at MM vehicle stackings for many battles (mine) but I simply don't have enough data to do a reasonable job.

    And yes, a database as large as yours would be extremely valuable in doing the study!

    BTW, I can't thank you enough for the work you've put in on your data/site. It's been very helpful in guiding my decision making on vehicle purchases / tech tracks. Donation on the way...
    .
    Last edited by LTCjRet; 2013-09-26 at 16:26.

  4. #4
    Retired Commander Phalynx.eu's Avatar
    Join Date
    Jan 2013
    Location
    Erlangen, Germany
    Posts
    2,127
    Who else did you expected here? I'm a one man show!

    Can provide you a list of tanks with for example the difference to low and top tier in that game, aligned with the count of that. e.g.:

    Tank, +3, +2, +1, 0, -1, -2, -3
    Object 904, 567, 234, 890, 245, 123, 345, 678

    Would CSV be ok for you?

    And thanks for donation, very appreciated!

  5. #5
    CSV would be great, thanks! Or even a MySQL database - I'm somewhat conversant with it.

    Any data on battle results would be appreciated. I've accumulated a small DB of around 400 battles but only those of which I was a part. What I'm attempting is to run a correlation / regression / ANOVA of the contribution the player makes to an individual battle / team vs what the "predictions" say (MM stacking / global WN7 / efficiency / global win rate, etc.

    One of the biggest drawbacks to global ratings is that they are not normalized for the learning curve, i.e. those ratings include the earliest performance (bad) of a player's lifetime. I strongly suspect things like XVM rankings are very poor predictors of battle performance for players with total battles in the ~4000-7000 count range. Rating over the last, say 200 battles and ratings tied to a single vehicle / type should be much better I would think. For example, I'd much rather run into a player with 10k battles than a player with the same WN7 lifetime rating that has only 3000 battles. The lower battle total rating is much more influenced by the player's early poorer performance than the former.

    Of course, I will share my findings with you for whatever it's worth. And I would appreciate your critique of my approach once it's is established.
    Last edited by LTCjRet; 2013-09-26 at 19:00.

  6. #6
    Phalynx, is there access to to your MySQL database (I think that's what you are using?) via the web?

  7. #7
    Retired Commander Phalynx.eu's Avatar
    Join Date
    Jan 2013
    Location
    Erlangen, Germany
    Posts
    2,127
    Quote Originally Posted by LTCjRet View Post
    Phalynx, is there access to to your MySQL database (I think that's what you are using?) via the web?
    That would be a big security hole! No, and it will not be available.

  8. #8
    Retired Commander Phalynx.eu's Avatar
    Join Date
    Jan 2013
    Location
    Erlangen, Germany
    Posts
    2,127
    Quote Originally Posted by LTCjRet View Post
    CSV would be great, thanks! Or even a MySQL database - I'm somewhat conversant with it.

    Any data on battle results would be appreciated. I've accumulated a small DB of around 400 battles but only those of which I was a part. What I'm attempting is to run a correlation / regression / ANOVA of the contribution the player makes to an individual battle / team vs what the "predictions" say (MM stacking / global WN7 / efficiency / global win rate, etc.

    One of the biggest drawbacks to global ratings is that they are not normalized for the learning curve, i.e. those ratings include the earliest performance (bad) of a player's lifetime. I strongly suspect things like XVM rankings are very poor predictors of battle performance for players with total battles in the ~4000-7000 count range. Rating over the last, say 200 battles and ratings tied to a single vehicle / type should be much better I would think. For example, I'd much rather run into a player with 10k battles than a player with the same WN7 lifetime rating that has only 3000 battles. The lower battle total rating is much more influenced by the player's early poorer performance than the former.

    Of course, I will share my findings with you for whatever it's worth. And I would appreciate your critique of my approach once it's is established.
    This approach would also require to have the WN7 for all players. Data that is not available here on vBAddict.

    Please mail me at marius.czyz@gmail.com so I can give you the requested MM data.

  9. #9
    E-mail is on the way - thanks very much!

Tags for this Thread

Bookmarks

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •