Results 1 to 9 of 9

Thread: How well does the database reflect actual player strength?

  1. #1

    Join Date
    20-04-19
    Location
    Norway
    Posts
    6
    vCash
    500

    How well does the database reflect actual player strength?

    Greetings all, and apologies if this post is misplaced or verbose -- I am new to this forum.

    My interest here is not in CM per se, but rather in the fact that some of you are maintaining its database, which, as I understand, also includes some form of assessment of player's abilities.

    Herein lies my interest. My background is as an academic, and at times I have seen the need to estimate the playing strenghts of footballers. That is, when economists and the like perform studies related to sports, they frequently have to compensate for the quality of the players or teams involved. Say, for example when studying the home field advantage and its causes, referee biases, the effect of kick-off times of matches, the effect of managerial changes, the peak age of players, or the effect of team homogeneity on overall performance.

    Over time I have developed a rating system for football players myself, which is based on +/- ratings. I will not explain this in detail here, but include some links in my bio to a webpage and a youtube channel with quite a bit of information, for any nerds out there. I have also seen some academic studies involving the use of other ratings, such as the ratings from FIFA (the game), Castrol ratings, or quite typically the player values listed by transfermarkt.

    Now, here is my question to your community: It would be very interesting to see if the ratings (player evaluations) in the (updated) database of CM are better than the other ratings (including mine), and therefore would provide scientists (if I may refer to economists as such) with a more precise tool to support their studies? A typical method for doing this would be to see if the CM ratings are better at predicting match outcomes than other ratings, when used in a method for creating predictions. This is also interesting in terms of seeing whether human expert evaluations of football players are more (or less) accurate than computer generated ratings.

    I collaborate with several researchers that may find such issues interesting to investigate, and therefore I hereby ask whether any of the CM update team would be intrigued by a possibility to explore some such venues.

    Thanks for reading all of this. Cheers!

  2. The Following 2 Users Say Thank You to FootballPlayerRatings For This Useful Post:


  3. #2

    Join Date
    03-03-12
    Posts
    2,791
    vCash
    7500
    Sounds like the old CA15 sim database
    Long live 00/01




  4. The Following User Says Thank You to Cam F For This Useful Post:


  5. #3

    Join Date
    07-03-17
    Posts
    976
    vCash
    600
    Nice to see someone developing a well designed +/- rating system for football players. I love a good football analytics article and I was at first a GoalImpact enthusiast (and soon later disappointed by the way it's heading).

    I watched some of your videos and found all of them both interesting and promising. I am looking forward to you publishing much more data and results.

  6. The Following 3 Users Say Thank You to Andrea71 For This Useful Post:


  7. #4

    Join Date
    20-04-19
    Location
    Norway
    Posts
    6
    vCash
    500
    Thanks! Indeed, Goalimpact was quite active around the time I started working on +/- ratings. They have now turned commercial it seems, which I think is very understandable, although I prefer a more open path myself.

    Question is, would the CM-type of games become more realistic if players' contributions were dictated by ratings calculated by mathematical methods, or would the ratings calculated by mathematical methods improve if they could utilize information hidden in the databases of games like CM? I would like to know the answer to that.

  8. The Following 2 Users Say Thank You to FootballPlayerRatings For This Useful Post:


  9. #5

    Join Date
    20-04-19
    Location
    Norway
    Posts
    6
    vCash
    500
    Quote Originally Posted by Cam F View Post
    Sounds like the old CA15 sim database
    I have to admit that I have not played CM since the '95 edition (and the Norwegian version of that). Could you elaborate?

  10. #6

    Join Date
    03-03-12
    Posts
    2,791
    vCash
    7500
    The original thread is long gone I think but found this on same page - https://champman0102.co.uk/showthread.php?t=4237
    Long live 00/01




  11. The Following User Says Thank You to Cam F For This Useful Post:


  12. #7

    Join Date
    03-03-12
    Posts
    2,791
    vCash
    7500
    Long live 00/01




  13. #8

    Join Date
    26-09-18
    Location
    Brazil
    Posts
    735
    vCash
    747
    very interesting your rating system and studies

    Im always impressed that CM predicts many results of games correctly. Just an example, in oct2018 update i saw it predict correctly several results of russia world cup matches, even in some unexpected real life results, sometimes even the goal scorer or the times and ways the goals occurred.

    However the data is created and tested specifically for the CM game. The game uses its own method for interpreting and processing this data (for example, the CA15 mentioned above is part of the method the game uses to process the data).

    If you take the data from the CM database and put it into some other method for creating predictions independent of the CM game, then I think it will predict well, but that will depend on the method. Anyway, that would be a really interesting experiment.

    It would be nice to experiment the inverse too, i.e. ratings calculated by mathematical methods being added into players in CM database to see how the game will predict

  14. The Following 2 Users Say Thank You to MadScientist For This Useful Post:


  15. #9

    Join Date
    20-04-19
    Location
    Norway
    Posts
    6
    vCash
    500
    Quote Originally Posted by MadScientist View Post
    very interesting your rating system and studies

    Im always impressed that CM predicts many results of games correctly. Just an example, in oct2018 update i saw it predict correctly several results of russia world cup matches, even in some unexpected real life results, sometimes even the goal scorer or the times and ways the goals occurred.

    However the data is created and tested specifically for the CM game. The game uses its own method for interpreting and processing this data (for example, the CA15 mentioned above is part of the method the game uses to process the data).

    If you take the data from the CM database and put it into some other method for creating predictions independent of the CM game, then I think it will predict well, but that will depend on the method. Anyway, that would be a really interesting experiment.

    It would be nice to experiment the inverse too, i.e. ratings calculated by mathematical methods being added into players in CM database to see how the game will predict
    This is more or less what I had in mind. As part of my research I have some knowledge about useful methods to perform predictions, and as you have discovered I have a means of generating some player ratings (I would be happy to help someone test if they could add something to the CM database). Now, I can think of many fun things to do with access to something like the 15 player attributes as used by CM. As an example, just today I came across a paper that discussed the degree to which player salaries are associated to the attributes of the players in FIFA (the game).

  16. The Following User Says Thank You to FootballPlayerRatings For This Useful Post:


Posting Permissions

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