Announcement

Collapse
No announcement yet.

[Discussion] Identifying and Fixing Balance Problems

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

  • [Discussion] Identifying and Fixing Balance Problems

    (There is a TLDR at the bottom)

    So there's been a lot of talk for a while now about balance problems on this server, with some long-term veterans commenting that they believe it has gotten worse now than it was in years past. It's led to some recent threads such as Wyz's introduction of the swap mod as well as Rad's thread regarding rookies. Since those threads are focused exclusively on their respective topics, I thought it might be helpful to create a broader thread that might not necessarily be focused on a specific remedy, but rather verifying what the issues are (and how they come about) and maybe brainstorming a solution, as I think there's a lot of different situations where balance problems come about, so one mod or change in community behavior isn't going to be a catch-all for all balance problems. And in order to come up with a solution, the related issue needs to be correctly identified and detailed.

    Below are some thoughts I have regarding what causes balance problems. I'll throw out a disclaimer now that I readily recognize I may be wrong/analyze these issues incorrectly, so feel free to disagree with any/all assertions (or ignore them all and just post what you think regarding the topic).

    One of the obvious ones is apathy. Sometimes teams are clearly uneven, and maybe someone on the disadvantaged team will complain, but won't attempt to take any action and nothing gets done, or players just "accept their fate" that it's going to be a stacked game. This seems to clearly be something that can only really be solved by change in players' behavior in taking action.

    Another one that is frequently noted is lack of time to properly fix teams. Wyz addressed this with the swap mod, though the issue has still persisted. On the one hand, players are still getting used to properly using the mod to it's full effectiveness so it may just be a matter of time. That said, some have still talked about possibly increasing the 60 second timer regardless of the mod.

    Another issue I've found is properly identifying where the imbalance lies. Pretty much everyone agrees on who the top-tier players are, but anyone else besides that handful of players I've actually noticed some disagreement (even if not spoken and only implied) on where certain player's skill level lie. As an example, people may say "you guys have player X, player Y, player Z, can we get player X in exchange for [low tier player]?" where player X is a top tier player, and players Y and Z are not top-tier but still good, though how good, and if good enough to be noted in such a way to notably impact balance, becomes the issue. I've seen games where the above example played out, but players Y and Z's skill levels were being overestimated (or there were players on the other team that were equally skilled if not better to match them), and it didn't justify trading over a top-tier player in such a way, and the result of the game ended up proving it was stacked in the other direction as a result of the swap. Similarly, I've seen games where certain players were underestimated and teams thought balanced, but at the game's conclusion it became clear certain players should have been given more consideration. Over time I've personally been guilty of every variation of this. I'm not sure what can be done to address this issue, though I wanted to note it regardless.

    I think the final issue I'll note is somewhat related to the previous issue, where you have new players on the server whose skill level is unknown. They may be a complete rookie who is not only new to NS2 but is also unskilled in shooters in general, or they may be a top-tier player (or close to it) who ends up stacking a team significantly. So recently to address this people try to balance based purely on primers/known players, and just kind of hope that all the unknowns are average players that average each other out, but many times that ends up not being the case and it turns out stacked. One possible thought for a solution: a mod that indicates on the scoreboard a non-primer's skill/rookie status, and it would be similar to the voicecomm bubble in that only primers could see it. Now how to estimate skill I'm not sure; first thought was to say to use the hive score since Hive 2.0 is apparently in the game (to the best of my knowledge), but how reliable it is I'm not sure (based on the Captain's board that still organizes by score it still seems dubious), and also since we're unable to view scores ourselves players would have no idea what's a "good" score or a "bad" score. Perhaps display number of hours/games played? Again not a good indicator at all of skill, but at least indicates experience, and the point of this mod is simply to give a general "in-the-ballpark" idea of what skill level that unknown player is at.

    Regardless of the above, would indicating rookie status be a possibility? Because as far as I know the only way rookie status is currently indicated is if that player types in text chat, otherwise they appear like any other normal player on the scoreboard and on voicecomms (didn't their names used to be in green?), and some rookies have been quite hesitant to divulge their rookie status for whatever combination of reasons, even when directly asked. But being able to identify rookies immediately would help immensely in both balancing as well as teammates knowing who to look out for and help (and know not to yell at them as if they're a veteran when they make a rookie mistake).

    TLDR: There are multiple reasons and situations where and why balance becomes an issue. I've noted what I thought some are. Please feel free to disagree/ignore what I thought, post what you think leads to balance issues and brainstorm possible solutions, whether that take the form of a mod or a push to change a certain behavior in the community. Perhaps once a certain issue/solution gains enough interest it can get it's own thread (or it can stay here, however it should work).

  • #2
    What you are talking about is skill imbalances, not gameplay imbalances. I want to make this distinction, if only for me.

    A lot of the issues you are describing are almost impossible to solve. It is really difficult to have players of a very high skill, and I do not even mean Nominous, play well with those of a lower skill. It is very hard to balance team skills when you have one player like Nominous and a rookie who walks in a strait line on the floor towards marines.

    This is not a unique problem to TGNS or NS2. NS2 has a very high skill curve and a very small player base that makes issues like these more visible. This is a problem that plagues NS2 worse than most games.

    TGNS has a multitude of options to show skill. Some servers have show hive skill values. Some servers have given players badges based on their relative ranking in gameplay metrics. TGNS has so far been against showing skill in this way. TGNS even goes as far as to hide player skill metrics from the NS2 end round stats.

    TGNS has also modified the shuffle function most other servers use for team balance to suit this communities requests. I personally think that the default shuffle function is superior. Using the default shuffle would at best be a marginal improvement, and might not even be humanly noticeable.

    You can try to improve the team balancing algorithm all you want, but that doesn't solve the real issue. It is very hard to have skill balanced games with large differences in skill. TGNS is a community for all skill levels, so this is not going to change. It is not feasible, nor appropriate to have a low skill TGNS and high skill TGNS servers.

    The only realistic way to improve this specifically on TGNS is to be vigilant in checking voice comms, and trying to spot clear imbalances. If you think the TGNS team balancing algorithm just is not doing a good enough job, try to get a captains game going.

    Comment


    • #3
      Thank you for two articulate and well-formed posts.

      It's definitely a possibility to return Rookie decoration to some or all player scoreboards. I'd be very interested to hear more folks' thoughts on the pros and/or cons of that. We're not bound to name colors for implementation. We can get more creative, if that helps. Part of UWE's reasoning for removing it was seasoned players discriminating against Rookies, and I'm certain we can work together to keep that under control. The newest "I will hear" gate does briefly show Rookies with a pinkish player name color on the scoreboard, so that might already be a step toward our understanding what we're dealing with as new faces join us.

      Originally posted by Nyon View Post
      [Apathetic complaining] seems to clearly be something that can only really be solved by change in players' behavior in taking action.
      We -- me, admins, and community members -- long ago made several iterations on the balance algorithm. Like... we spent months on it -- with and beside other efforts elsewhere in the community, we definitely made marginal improvements. I'm now content to wait for the Next Big Thing that significantly lessens title-wide player complaints about imbalanced games -- maybe we can adopt that. Until that time, my opinion is that NS2's greatest skill-balancing minds have done just about all they're going to do for delegating roster skill balance responsibility to a computer.

      So we're still innovating on tooling by which we enable knowledgeable, pre-game complainers to provide last mile solutions where the data and logic fail. We're growing our adoption of and iterating the implementations of both "switch" and "swap", including how they manipulate the force-start timer -- and we might yet directly alter that timer itself (I'm still hoping that, in the context of well-informed "switch" and "swap" usages appropriately raising the timer, we needn't raise its start value).
      Steam Community? Add me. | Free Remote, Encrypted Backup

      Darkilla: In short, NS is pretty much really fast chess. With guns. Apophis: I haven't seen anyone say that SM's are better than non-SMs. Nordbomber: This is THE first server I've seen where either side can comeback from out of seemingly nowhere with the right teamwork. en4rcment: I have NEVER experienced the type of gameplay that I have found here. Nightly I am amazed at the personalities and gaming talent. Zephyr: Apophis is clearly a highly sophisticated self-aware AI construct that runs on a highly modified toaster oven in Wyzcrak's basement.

      Comment


      • #4
        I often find it hard to keep track of all the regulars when attempting to choose which players to swap for whom. Players who are stellar on one team might not perform similarly on the other, after all. Instead of relying on memory and intuition, I think we need to be able to visually rank players based on their alien and marine performance. This would make it much easier to identify optimal swaps.

        Something that I've wondered about, which Nordic mentioned, is why the server doesn't show individual hive skill. I assume this has to do with possible embarrassment or discrimination against those who do not perform as well as others. If so, I can see a ranking system like the one I propose being considered undesirable. I don't see what the issue is, however, since this is a mature and friendly community. By contrast, I'd argue that higher skilled players face more pressure and derision (lighthearted of course) during their off days. Anyway, even without individual hive skill or public ranks attached to players, their skill can be estimated by playing games with them, spectating, and by viewing Replays. We already have the gist of individual skill of regulars because the basic scoreboard tells us so (although I'd very much like to see the inclusion of NS2+'s comprehensive end-of-round stats). This ranking system would merely serve as a convenient reminder of separate alien and marine skill for quick manual balancing.

        For starters, there could be five ranks for both teams, colored in green (lowest), green-yellow, yellow, orange, and red (highest). The icons could be based on the Frontiersman and Kharaa team icons over at the NS2 wiki. This results in two columns of marine and alien icons of (probably) different colors for each player on the scoreboard. Simple and easy. The ranks could be determined by players' average KDR, and the data could be taken from the last 20 or so games (keeping the player's performance updated) played only on TGNS. Before a game starts, the scoreboard would sort players on both teams from top to bottom based on team rank. If players are displeased with auto-balance or most players have already joined a team manually, this ranking system would come in handy for swapping players. It would also quickly identify whether captains teams are grossly imbalanced, which does happen.

        Comment


        • #5
          Part of what you are talking about Nominous is something like what the Tactical Freedom server did. https://tacticalfreedom.com/leaderboard
          Their admin has since moved on, so the website is not functional in that capacity.

          We could develop many variations of skill ranking. We could use similar badges to Tactical Freedom, but use hive skill instead. We could use some combination of gameplay metrics. For example, I have playing with a ranking that averages the individual ranks of kills/deaths, score/minute, average accuracy, player damage / minute, and structure damage a minute. This produces a number where lower is better. How we rank players is unimportant right now, but if we want to implement badges that relate to relative skill. If TGNS chooses to show a relative skill badge, TGNS or myself have the data to support it.

          Separate marine and alien skill values is something I have wanted included into the hive skill system for a long time. UWE has so far not attempted it because it is very likely it would force certain players to play one team significantly more often than the other. Even if UWE was able to implement this, its effect on skill imbalances would be highly marginal. That is because the problem lies not with team balancing, but in the difficulty of trying to balance the set of players we have. Try to have a quality game where you are playing any sport with both young children and professional players. That is a similar albeit more dramatic analogy to what trying to balance teams in NS2 is like.

          Comment


          • #6
            Shameless did some great things with data. I miss him, and the server.

            There could 2 small changes to pregame that may help, but I'm not sure how "doable" it is.

            We show lifeform/skill strength in Captains, so I know we're recording SOMETHING there. Maybe, in pregame only, and only to Primers, show those icons inline on the scoreboard, and make it team specific.

            For instance, it I'm on marines, a comm icon is shaded by my name similar to how ti is on the cap menus. As alien, the khamm/gorge as shown. Same for all known players.

            The second, show the teams is order from highest skilled to lowest pregame, only for P signers. Then in the limited time we have to adjust balance, we can quickly glance and respond. I've been guilty of missing a player in a balance judgment, mainly because I still feel the pregame is way too rushed, and I'm trying to assess, process, propose adjustments to my team, THEN type it out and hope we get balance.

            This WHILE we're doing comm checks, people are still trying to switch (which will undo all the balance thoughts), etc.

            Yeah, I feel the 60 second timer is a hindrance.

            Comment


            • #7
              The other issue I see with balance in general is that your weaknesses as a player might be exposed by different types of players on the other team. i.e. some players are better at killing you or finding your gorge tunnel or drawing you out than others. So when you do a swap, its possible that the one person who would make your "skill" level lower, by harassing you more often, is now on your team, and you look more "skillful" via the stats than normal. Like if Nadal plays on clay in tennis, he is usually better than on other surfaces.

              Also we tend to remember certain commander "tendencies" as far as structure placement or what techs to prioritize, which if the other team learns quickly, can be used to devastating effect.

              And we also have map variations, where players inherently play some maps more than others and perhaps intuitively understand the critical points of the map to hold and defend or attack, whereas on other lesser played maps, those points are not regarded highly because the map has not been played enough. For instance we all know that mezzanine is emphasized greatly on Tram, and the loss of that RT location for a phase gate or a tunnel would not be discarded lightly be either commander/team. I know that I have probably heard "mez" on voicecomm 536 times in the last month. On a map like ns2_gorge, however, the map is not played significant gametime on the server, and when it does somehow win a map vote, the relative inexperience of the map in general can affect everyone's perception of whether the game just played was a "good" or "enjoyable" match.

              So, anyway, perhaps we could display a map stat prior to each match, that gives real information about which team wins more frequently for a given map. That way, when players go to choose which team they want to go on, the players that have a tendency to stack will choose the team that has the greater winning percentage. This can give primer signers the ability to choose the underdog team and help determine whether or not the statistic is defeatable by skill. This would put more focus on playing the map, and give players something to talk about after the game as far as if the map has enough elements in it to be "fair", or if it is too skewed one way or another. This approach has the benefit of not changing anything at all about the current way the game is played other than information about the map given. You can still play on either team regardless of hive skill or rookie status and we don't have to change the hiding of individual stats.

              As with any proposal, there will always be pros and cons, and proponents and detractors, but I thought I'd throw this idea out there in case others think it might be a good idea and lend a supporting voice. Or a non-supporting one. No harm either way.

              Comment


              • #8
                Originally posted by Nominous View Post
                I often find it hard to keep track of all the regulars when attempting to choose which players to swap for whom.

                ...

                Before a game starts, the scoreboard would sort players on both teams from top to bottom based on team rank.
                The scoreboard is crowded to the point that some folks already tolerate overlapping/unreadable content on the scoreboard due to their screen resolutions. Meanwhile, scoreboard modding isn't trivial. Data-driven, pre-game scoreboard sorting is less likely to cause UI problems. That's not to say we can't further decorate the scoreboard -- it's just complicated, is all.

                Originally posted by james888 View Post
                Separate marine and alien skill values is something I have wanted included into the hive skill system for a long time.

                ...

                Even if UWE was able to implement this, its effect on skill imbalances would be highly marginal. That is because the problem lies not with team balancing, but in the difficulty of trying to balance the set of players we have.
                Yeah. I'm definitely not feeling any "improved roster automation through better data" vibe these days -- I'm very open to someone else spearheading that particular improvement. But I don't mind using team-specific data when sorting players on the scoreboard, to make manual swaps easier.

                Originally posted by rad4Christ View Post
                Then in the limited time we have to adjust balance, we can quickly glance and respond.
                Originally posted by smiley View Post
                we could display a map stat prior to each match, that gives real information about which team wins more frequently for a given map
                I'll see about pre-game scoreboard sorting of players based on their team-specific TGNS win/loss rate, maybe even on the current map specifically (I know that wasn't quite your implementation intention for the map data, smiley -- I'm thinking of showing what you had in mind another way, but I'm not sure yet how). That should help the switch/swap usages (and, as I've said recently, we'll continue to tweak the timer adjustments there).
                Steam Community? Add me. | Free Remote, Encrypted Backup

                Darkilla: In short, NS is pretty much really fast chess. With guns. Apophis: I haven't seen anyone say that SM's are better than non-SMs. Nordbomber: This is THE first server I've seen where either side can comeback from out of seemingly nowhere with the right teamwork. en4rcment: I have NEVER experienced the type of gameplay that I have found here. Nightly I am amazed at the personalities and gaming talent. Zephyr: Apophis is clearly a highly sophisticated self-aware AI construct that runs on a highly modified toaster oven in Wyzcrak's basement.

                Comment


                • #9
                  Manipulating data is always fun, but presenting that data in a way that is both easy to read and useful is hard. Reading through all of the comments above, it seems like a hybrid approach would be able to address most suggestions.

                  Such as:

                  - Keep track of skill in some type of stat for both marines and aliens seperately for each player (possibly in a last 20-50 rounds limit like suggested above so 1 really good or really bad game does not stick around too long)

                  - Use this data to break people into maybe 4 or 5 different skill groups. (With hive data, say <1000 = 0, 1000-1999 = 2, 2000-2999 = 3, 3000+ = 4, but the resulting tiers would really depend on how the "skill" is calculated)

                  - Display this this skill ranking as a two digit number in the scoreboard so an average player may show up as a 23 but a high skill player would be more like a 55, where the first number is marine side and the second number is alien side

                  - Sort the scoreboard by hive skill so you also have a general ranking of each player as well

                  This would let people know relative rankings or players on each side and who to swap to gain the best balance. This may cause somebody with imbalanced skill to play 1 side more than the other, but since the switching is optional, it is really up to them if they are ok with the change.

                  I don't think adjusting the balance algorithm will help much unless you change it to also re-assign players that are already on a team. I personally like this as it is now, because sometimes I really want to play a specific side.

                  Comment


                  • #10
                    Two things regarding balancing as it stands

                    1) BUG: Recording "badge" icon in tab screen currently obscures part of player name.

                    a)This making it difficult to quickly read a players name so that a swap command can be entered. Sometimes the spelling on player names are "unique" :)

                    2) Game start timer is too short.

                    b) Since there is no visible indicator for player skill we have to use our memory. Then we carefully weigh if we think teams are stacked since if we are wrong we will in fact create the problem we are trying to solve. Then we have to find two players we think can swap. Then we type in the command for the suggested swap. Doing all of these in a responsible manner can sometimes be accomplished in the time frame but I haven't been able to accomplish it successfully myself. I've tried in a few games to no avail and I'm a fairly quick typer.


                    Some possible solutions

                    1) Fix recording badge/icon or remove it.

                    2) Extend timer

                    3) Create some kind of pause for timer. (It can be keyed to primers only, and or allows any primer to unpause the timer as well.)

                    3) Enable ALL talk before round starts then we might not need to perform manual swap commands. We don't generally use it for planning unless it is captains night it's too short on a normal game.
                    Last edited by popemonster; 01-06-2017, 08:34 PM. Reason: EDIT: accidentally hit submit too early.

                    Comment


                    • #11
                      You only need provide a portion of each player name -- enough to uniquely identify that player by name (wyz, bria, nomi, rad4, etc) -- maybe you already knew this.

                      I can totally hide that recording badge pre-game. I won't do that unless we decide it really is necessary considering my comment above (which might have been already known to you).

                      We're still tweaking how "switch" and "swap" commands raise the timer, and players are still learning how it works. The current design is that the timer raises when players move due to either "switch" or "swap", and the timer also raises when someone proposes a "swap" -- but I need more testing. I've already got plans to pre-game sort the scoreboard teams by each player's historical team-specific competitive strength -- this will streamline everyone's identification of possible balance problems. I'm still aiming for solving the "not enough time" problem by making it easier to identify problems and raising awareness of how the chat commands interact with the timer.

                      If process efficiencies continue to result in insufficient time, a timer" chat command that extends the timer by 10 seconds or so would be fine -- creates additional time for typing late chat commands. I'll also see if I can detect on the game client that a player is /about/ to send a "swap" chat -- I could extend the timer based on that.

                      I think we're better off avoiding pre-game alltalk.
                      Steam Community? Add me. | Free Remote, Encrypted Backup

                      Darkilla: In short, NS is pretty much really fast chess. With guns. Apophis: I haven't seen anyone say that SM's are better than non-SMs. Nordbomber: This is THE first server I've seen where either side can comeback from out of seemingly nowhere with the right teamwork. en4rcment: I have NEVER experienced the type of gameplay that I have found here. Nightly I am amazed at the personalities and gaming talent. Zephyr: Apophis is clearly a highly sophisticated self-aware AI construct that runs on a highly modified toaster oven in Wyzcrak's basement.

                      Comment


                      • #12
                        Thanks Wyz, I didn't realize that only a portion of the name was required, that should make it easier. I know you've put a lot of thought into the balancing mechanics as they stand and appreciate all the hard work. Just wanted to throw in my 2 cents on the topic. I realize there is no silver bullet for the problem and any solution will be imperfect.

                        Comment


                        • #13
                          Pre-game Marine and Alien rosters are now sorted descending by team-respective TGNS win/loss rates, with 20 games needed for any player to sort beyond the default rate (0.5) -- rates are not shown... only used to sort. When the game begins, the scoreboard resumes stock sort behavior. Only the scoreboards of Marine and Alien clients are modified -- players in Ready Room and Spectate only ever see stock behavior.

                          Consider this information when deciding who to move pre-game. I'll continue this week observing and (if necessary) further tweaking "switch" and "swap" impacts on pre-game force-start timers.
                          Steam Community? Add me. | Free Remote, Encrypted Backup

                          Darkilla: In short, NS is pretty much really fast chess. With guns. Apophis: I haven't seen anyone say that SM's are better than non-SMs. Nordbomber: This is THE first server I've seen where either side can comeback from out of seemingly nowhere with the right teamwork. en4rcment: I have NEVER experienced the type of gameplay that I have found here. Nightly I am amazed at the personalities and gaming talent. Zephyr: Apophis is clearly a highly sophisticated self-aware AI construct that runs on a highly modified toaster oven in Wyzcrak's basement.

                          Comment


                          • #14
                            Also, the Recording/Streaming badge should be hidden from the scoreboard pre-game for sub-1080p resolutions, so that should aid switch/swap ease of use.
                            Steam Community? Add me. | Free Remote, Encrypted Backup

                            Darkilla: In short, NS is pretty much really fast chess. With guns. Apophis: I haven't seen anyone say that SM's are better than non-SMs. Nordbomber: This is THE first server I've seen where either side can comeback from out of seemingly nowhere with the right teamwork. en4rcment: I have NEVER experienced the type of gameplay that I have found here. Nightly I am amazed at the personalities and gaming talent. Zephyr: Apophis is clearly a highly sophisticated self-aware AI construct that runs on a highly modified toaster oven in Wyzcrak's basement.

                            Comment


                            • #15
                              Thank you very much Wyz, not sure how many people appreciate the work you put in. I'm currently back in school to get into the medical field but have 10 plus years in ATM tech support and worked regularly with programmers on front end database design/implementation. Very crazy to have someone who can implement and make these kind of changes as quickly as you can once you are convinced of a need.

                              Comment

                              Connect

                              Collapse

                              TeamSpeak 3 Server

                              Collapse

                              Advertisement

                              Collapse

                              Twitter Feed

                              Collapse

                              Working...
                              X