Announcement

Collapse
No announcement yet.

Team Dynamics Plugin

Collapse
X
  • Filter
  • Time
  • Show
Clear All
new posts

  • Team Dynamics Plugin

    I'm working on a plugin to map the team dynamics. To do this I am trying to keep track of who is on your team more often when you win. Basically, every time you win the "bond" value between you two will increase.

    So of course those that regularally lead their team to victory should have many more people connected to them, but I think that if this is run for a good 100 or so games it should give some interesting results... Maybe we will be able to tell which people seem to just "work" together.
    (This will be done by using an outside program to actually parse the data and then draw it into a net with everyone's name.)


    Now I have some questions about what you guys think of this plugin.. First of all, sound interesting? I don't know if it will ever be on TG, but we shall see.
    Second, what do you think of the idea of strengthining the connections when players win together? What if someone joins midgame or leaves midgame? Should they be count in this? What if they switch teams and the old team that they are no longer on wins?

    Do you think the "connections" should decrease upon losses? or just increase on wins? The point is not to find balance, but to find which people work well together. Losses might not actually mean the people don't work well together, just that things didn't work out that game. When they win it's probably because they worked together. Should bonuses be given for extra long games?

    There are just so many variables to this. What do you guys think? I'm trying to figure out what I want to put into this to make it as interesting and accurate as possible.

    The main problem with this is that there ARE leaders, but I can't make a program to identify who those leaders are. They will probably have more connections and lead people to victory more, but then again, everyone will probably have at least *some* connection with everyone else just because of randomness.


    I think if I get this data it would be pretty cool to render it in 3d with orbs symbolizing each person. The orb would glow golden if there are many connections, or be black for very few. The connections between orbs would be lines of various thicknesses. Probably have an option so that the connections of only the orb you're looking at would be drawn... This could be wicked interesting... Or maybe rather than thickness of the line it could be handled by proximity... So if the connection is stronger the line connecting would be way shorter.... Would cause the ones with lots of good teamwork to be clumped together...

    Give me your thoughts. I have rambled long enough.
    remi.D

  • #2
    Re: Team Dynamics Plugin

    I would suggest adding 1 for a win and subtracting 1 for a loss. Since I would not expect the same number of games to be played across the board this will keep it closer to true. If player1 is in 20 games with player2 and wins 10 but is in 5 games with player3 and wins all 5 you can either have the scores "10" and "5" or "0" and "5"

    In theory with this and looking at the data you would be able to identify people who have a tendancy to win (strong connections with lots of people) and people who have a tendancy to lose (weak connecitons with lots of people). I'm not sure I would recommend this method for that task though, as it would involve looking over lots of data. I guess you could average a person's scores with all the different people they have played with and compare it that way.

    As a curious person, if you do this I would be interested to see the results. I don't know how affective it would be at determining who works well with other people.

    Comment


    • #3
      Re: Team Dynamics Plugin

      The problem with detracting for losses though is that sometimes even with the best teamwork your team can lose.... But then again... your example implies that you played 20 games on the same team... if you lost half with this guy, but won all of them with another guy, what's the likelyhood that you have more teamwork with the other one? Or could it just be luck?

      If yuo played 20 games with a guy, don't you think your teamwork with that one would be better because you've played more games?

      Detracting reverses the effects... I tend to feel that the "bonds" should only weaken as other ones form stronger, that way you can tell more who just happened to be on the same team as you and who is actually on your team often when you win.
      remi.D

      Comment


      • #4
        Re: Team Dynamics Plugin

        Originally posted by blu.knight
        The problem with detracting for losses though is that sometimes even with the best teamwork your team can lose.... But then again... your example implies that you played 20 games on the same team... if you lost half with this guy, but won all of them with another guy, what's the likelyhood that you have more teamwork with the other one? Or could it just be luck?

        If yuo played 20 games with a guy, don't you think your teamwork with that one would be better because you've played more games?

        Detracting reverses the effects... I tend to feel that the "bonds" should only weaken as other ones form stronger, that way you can tell more who just happened to be on the same team as you and who is actually on your team often when you win.
        The point I am making is if you don't either subtract or use a ratio you will not have an accurate way of comparing "player1's bond to player2" with "player1's bond to player3"

        If player1 and player2 are both on the server from 4pm until 8pm and player3 frequents the server from 6pm until 10pm then given random or semi-random teams player1 and player2 will with time develop a higher bond than either of them will to player3. Player3 could have all the teamwork in the world and work very well with those people together and it wouldn't matter because more games would eventually tally up a higher "bond".

        Comment


        • #5
          Re: Team Dynamics Plugin

          You do make an interesting point. The question becomes which way the information would be more interesting. The people that you play with more often you probably are more familiar with and such, but you are right, someone you play with less could in fact have awesome teamwork and everything with you. The solution I think is not what you suggest though. I'm not sure there is a solution because this is another thing that can't be figured out by a program.
          I mean I could also keep track of how many games you played with each person so that you could find the ratio of games you won while playing with the other person...
          The problem with this though is that if you play one game with someone and win, you will have a "perfect" "bond" with them, which is the problem behind this.

          If you play more games with someone, you are likely to have better teamwork. If you win more often with one person you are also more likely... but there are always exceptions.


          edit: This is like the balance plugin, I just have such a hard time wrapping my head around the type of data I would collect.
          remi.D

          Comment


          • #6
            Re: Team Dynamics Plugin

            One thing that is absolutly essential for this data to be meaningful is normalization. If you don't normalize your data, you'll get everyone "connected" to people who happen to win more, with or without the people they're supposedly bonded to (which would be pretty much the entire server). What you want to find out is how much having these two people on the same team INCREASES their respective chances to win. The best way I can think of would be this: after some large number of games, take the win/loss ratio when they are on the same team, and subtract the average of their individual win/loss ratio. This result would be positive when they win more often than you would expect just from their individual ratios, and negative when... well, you know what can happen...

            Note that this approach still has a problem, in that NS is often more affected by the better players (a really good fade/lerk can save his team), so the average of the individual players' W/L ratio is not a good estimate for their expected chance to win. If a more accurate predictor could be determined, it would improve your data, but offhand I can't come up with one. If we were able to come up with a better predictor of some team's chance of victory, it could be used to significantly improve the balance plugin (not that I think there's anything wrong with the current version, just that there's always room for improvement).

            Comment


            • #7
              Re: Team Dynamics Plugin

              I'm not sure if this could actually work for a balance plugin... I mean it could, but it's just way overly complicated for that. I think it's better for it to just be interesting data, especially if I was to see it graphically.

              I definitely see your point. If anyone has any other ideas of how I could figure out how well certain people work together, give me them.

              I definitely like what you're suggesting. It would probably make for more interesting data for sure.
              remi.D

              Comment


              • #8
                Re: Team Dynamics Plugin

                Originally posted by blu.knight
                The problem with this though is that if you play one game with someone and win, you will have a "perfect" "bond" with them, which is the problem behind this.
                Thats the problem that doing the +1/-1 thing would solve. It will still eventually be related to how many games you play with a person but you won't notice it as quickly. If player1 and player2 win 60% of the games together then you would expect with the +1/-1 system for them to have 2 points for every 10 games they play. If they keep playing with that expected win% it will go up, just not as fast as it would with the adding only. Similarly if player1 and player2 only won 40% of the games together you would expect them to lose two points per 10 games.

                Ratios would be the best way to express something like this, and consider that if you had only 1 game to look at it wouldn't matter what system you used that information would be useless by itself. Whether you mark it as a 1, a 100% or whatever it's not useful until there are more games played.

                Comment


                • #9
                  Re: Team Dynamics Plugin

                  It definitely requires more than a simple +1/-1 , there are simply too many variables / data points. You might just end up plotting who is good vs. who isn't, and then when a team is stacked (skill, not so much teamwork wise) you will be able to predict what anyone could. The results would be uninteresting.

                  Its a great idea, but a very hard execution. I like Ag's input.

                  Comment

                  Connect

                  Collapse

                  TeamSpeak 3 Server

                  Collapse

                  Advertisement

                  Collapse

                  Twitter Feed

                  Collapse

                  Working...
                  X