Autobalance - why are you so weird

[Dr.GreenThumb]

I'm trying to get some insight into how autobalance works exactly. I've noticed on occasion the following (at random times):

- autobalance is asleep and refuses to balance out teams - a game will be 9:2, with zero action from autobalance till the end
- autobalance is overactive - I get moved 3-4 times during a game, the match may not even be that out of balance to begin with
- autobalance is just weird - game is over, I am looking at the final scoreboard, 10:8 final score, yet autobalance moves me to the other team
- autobalance is too late - game is 9:3, only three players alive from the losing team, a player gets moved over from the winning team 10 seconds before the game is finished.

None of this is a problem, I just don't fully understand what these accomplish

It can be quite annoying sometimes to be in the winning group with a bunch of friends, doing team work only to be all of a sudden be thrown into the enemy team, so from that point on one has to go against the soldiers that one went to war with ...but I can understand that as I'm sure it is very difficult to adjust the balance as people leave mid-game (boooo) while others join, etc.

Yet, my personal opinion is that it's sort of pointless to move anyone around post round 8, but definitely pointless -during- round 9. I mean at that point it is what it is, autobalance shouldn't change the outcome of the game in round 9. Not to mention moving a player while on the scoreboard...

Again, all this is just my opinion and would love to get some adminless insight

Cheers!

[aimer]

It would also be nice if players would join as frozen in the current round when they connect or swap team.
This would prevent people to cheat the game by reconnecting immediately after and would also prevent strong players changing the outcome of the current round as soon as they connect adding imbalance.

[Dr.GreenThumb]

Interesting idea.

Join an existing game frozen - for new connections - yes that I think would make sense. But not when autobalance moves you to the other team automatically... it already feels like punishment to leave your team, let alone you end up frozen up... but if you are a new joiner, yeah that is tolerable considering how some users leave and join right back in like you wrote.

Of course, there may be other negatives we aren't thinking of or none of this may be feasible as is.

[adminless]

well, I already answered about those late game team changes on other occasions, it's true that they may make no much game logical sense but they serve a clear and very important purpose, mitigate end game drama as well as condition the teams for the next game. I remember on early versions of the auto balance that it gave some more freedom for manual adjustment, people frequently bitched and complained about it times ten near non stop.

for the rest, I'm sorry but I don't disclose exactly how things works, mainly due security reasons, but well I can tell you that for the most part those look like normal game situations to me. I mean, I think that it's unrealistic, artificial and it would be boring/people wouldn't even accept it either to have continuous endless 10:9 games where winning would become more a matter of luck than anything else. it's not the goal of the auto balance to blindly enforce a balanced game at all times and nothing more. if I would like to do that, I could do that very easily but that would basically trash the whole server. then the auto balance looks forward more into conditioning a fair game overall while minimizing player dissatisfaction and, in my opinion and according to the stats, it generally success at that task.

for example your current stats as of now reads that the teams you've been on won 49.4% of the times and scored an average of 8.41 points with a standard deviation of just 2.09 points at a change rate of 20.4%. or in other words, that your teams won half of the games and typically scored between 10 and 6 points and you are changed (or you change yourself also) one in each five games. those are very good stats in fact, the high change rate (which is typically more in the 10% range) is explained as obviously mid rank players have higher chances to be changed as changing a low rank doesn't make any difference and the game isn't typically that unbalanced as to change a top rank.

as I typically check the servers I find the teams most of the times generally balanced. what I find that happens often is that players/teams under/out perform and that some times leads to unbalanced scores but just like I said, that's normal, that's at the end what the game is about, winning, there's no much to do about it. as for starting frozen that's a killer for obvious reasons and it's not even needed as there's already a re-connection timeout of about a minute to account for that issue.

[Dr.GreenThumb]

Ok makes sense, thanks for the insight. This is really all I was asking for to understand how it works at a high level, and that there is a reason behind the things that may seem illogical. I'm also happy to read that I don't get shuffled nearly as much as the average. I never change myself, I don't see the point.

As usual, looks like there is a lot more on the backend that an average player realizes

Thanks for the hard work adminless!

[Dr.GreenThumb]

A follow up question that sort of ties into this subject. Apologies if this makes no sense, or if it's simply not doable or if it shouldn't be done for reasons I am not thinking of.

Let's say a player gets moved by autobalance from red to blue. But that player doesn't want to be moved, and they immediately move to spectators. Then, after a minute or so they join back red, because they want to play with their friend. Autobalance doesn't really do much about this after the manual switching back, and not surprisingly the game is lost as the player did not honor the autobalance's move.

Is there any way to make a rule that says "if you are in spectators and you've recently been moved by autobalance" - if this condition is true, then to join back the source team you've been moved from - the time to allow that increases from the default 1 minute or so to let's say 5 minutes. but, if you want to join the team autobalance moved you to, you can do so right away (if the balance conditions still allow). I'm thinking this is to discourage the "manual override" some players do and by that they also somewhat ruin what autobalance is trying to achieve.

Maybe this is silly, I'm mainly just asking, I'm sure you've already thought about it...

Thanks

[adminless]

don't worry that's no silly, it's a good suggestion in fact. at the moment it's not implemented because I don't really find it necessary, I believe that this level of autobalance should suffice. think that besides the minute of delay people who goes spec immediately throws off their score and they don't even know if a minute later the team they want to join will be available plus all the shame they take now and on top of that even if by chance you still succeeded you'll be candidate number one to be moved again (the "ping-pong" effect, the auto-balance already accounts for that). if you think about it I believe that those are enough penalties already as to make it in-viable in the long run, in fact, as I see just after some days of the new leveled up auto balance the team hijacking has been gone almost completely.

as for that like I said it's a good suggestion and may be at some point in the future I may end up adding some variation of that for a new level of auto balance. however that isn't "perfect" either, first of all yes many people go spec in a attempt to hijack teams but not everybody and by being too restrictive you'll also get rid of legit guys going to spec. secondly, you don't really know how the server is gonna look like one minute ahead, let alone 5 minutes later, you may be locking someone to a team where now he is no longer needed. third and foremost, all that can be worked out to some extend and you may expend lot of time in perfecting for people ultimately just re-connecting all the time until they success so overall it does also have some traps of its own, it's not so straight forward as it looks. like I said, yes probably for a level three auto balance code, it may be the way to go but so far we better work with what we have now.

[Dr.GreenThumb]

Nice, glad to hear team hijacking is almost completely gone. Thanks for the explanation.

[PacMan]

Dude just bumping auto balance topic

ok you made a change so we cant switch, but some times now 5mins will pass playing 6v3, or other unbalanced teams, bots turning up late, even tho bots not needed etc theres a delay with the auto balance you need to fix it

[adminless]

"bots turning up late" because "playing 6v3" (humans), I think that your post answers your own question and five minutes is clearly a exaggeration, bots very rarely (may be once a day if ever) stay for more than a minute let alone five. humans can't be switched at convenience at any given time as they can be engaged on combat in that moment, hence the delay. the only player that can be added/removed at any given frame at convenience is a "imaginary" one (a bot), hence when the auto balance can't meet its needs from the human players because they are engaged on combat it adds a imaginary one until the human player selected can be changed this way meeting both conditions at any single given frame. that's a perfectly valid solution so unless I missed something (i.e. there's a bug, in such case let me know) you're not counting all players when you say that you played 6vs3 at any single given frame.

following this, it's more than a well proven fact that the fully auto balance works as can be seen on last game records.

• Monday 4 of November of 2019

  Code: Select all

  red won 46.21% of the games and averaged 8.243 points
  blue won 53.78% of the games and averaged 8.117 points
  score distribution was:
   - 20.16% of the games scored 19 points
   - 20.16% of the games scored 18 points
   - 18.48% of the games scored 16 points
   - 13.44% of the games scored 15 points
   -  8.40% of the games scored 17 points
   -  7.56% of the games scored 14 points
   -  5.04% of the games scored 13 points
   -  4.20% of the games scored 12 points
   -  2.52% of the games scored 11 points

• Tuesday 5 of November of 2019

  Code: Select all

  red won 47.86% of the games and averaged 8.085 points
  blue won 52.13% of the games and averaged 8.538 points
  score distribution was:
   - 21.36% of the games scored 19 points
   - 20.51% of the games scored 18 points
   - 17.09% of the games scored 17 points
   - 13.67% of the games scored 16 points
   -  8.54% of the games scored 15 points
   -  8.54% of the games scored 14 points
   -  6.83% of the games scored 13 points
   -  3.41% of the games scored 12 points

• Wednesday 6 of November of 2019

  Code: Select all

  red won 69.04% of the games and averaged 8.976 points
  blue won 30.95% of the games and averaged 6.992 points
  score distribution was:
   - 18.25% of the games scored 17 points
   - 16.66% of the games scored 19 points
   - 14.28% of the games scored 18 points
   - 12.69% of the games scored 16 points
   - 12.69% of the games scored 14 points
   -  8.73% of the games scored 15 points
   -  6.34% of the games scored 13 points
   -  6.34% of the games scored 12 points
   -  2.38% of the games scored 11 points
   -  1.58% of the games scored 10 points

• Thursday 7 of November of 2019

  Code: Select all

  red won 54.91% of the games and averaged 8.303 points
  blue won 45.08% of the games and averaged 7.926 points
  score distribution was:
   - 22.13% of the games scored 19 points
   - 18.85% of the games scored 16 points
   - 13.93% of the games scored 17 points
   - 12.29% of the games scored 18 points
   - 10.65% of the games scored 14 points
   -  9.01% of the games scored 15 points
   -  5.73% of the games scored 12 points
   -  4.09% of the games scored 13 points
   -  3.27% of the games scored 11 points

those are some perfectly realistically good results.

[PacMan]

your just to hardcore for me! anyhow maybe exggerrated alittle , but ask the other players , logs can look great but actual it effects games in my opinion,

theres only one way to sort this out you and me 1v1 you kill me and you win, you see i am full of great ideas,

[adminless]

look, know this, no matter what you do, one way or another, people is always going to complain no matter what so the point here is not whether they complain or not, if not whether those complains are right and make sense or not instead. people dissing the fully auto balance because it added some "imaginary" players (bots) for a few seconds, or may be even a minute at worst if you want, to cover up the time it took to materialize the optimum real human player team change is clearly in the wrong and that's essentially what matters here. it has no tangible negative impact on the game whats-o-ever and it just gets the job done as it should and that's an actual fact I can back up as you can see.

btw also for the record let me tell you this. yes, the guys now dissing the fully auto balance for this reason, of course I know them, they were the sames consistently exploiting it for their own benefit to get on the better team and/or to play with their friends and actually break the team balance and the ones that ultimately motivated these policy changes. so better get your facts straight and your own opinion and don't trust/believe in everything people throw in a public server chat.

[PacMan]

theres only one way to sort this out you and me 1v1 you kill me and you win


ok cool dude i being educated!

[Dr.GreenThumb]

I like it when the bots join, it's quite funny to see them ..makes me nostalgic. And of course from the previous explanations with the difficulty to balance during combat I think the bots are a really smart workaround. Did notice though when I spectated the bots that they are pretty easily picked off. Is joining them at hardcode level too strong?

[adminless]

yes very likely. the bots are more or less tuned for the average public random player skill level, or in other words, for them to be able to play the game but without being too "annoying" for anybody.