team balance on FreeFun

[imprecise]

Hi Adminless

I'm tired of playing crappy matches because of the team balance that puts all the good players in one team and all the players of a lower level in the other team. Or that puts the pro players in the same team against average players. Fortunately this doesn't happen in every match, but unfortunately it happens a little too often in my opinion...
Adminless could you please try to modify, improve the distribution of good players and pro players in the team balance. At least on the FreeFun server.
Thank you for taking my request into consideration.

[adminless]

I guess I'll take a further look at that but from a quick look up now the server history for registered matches of at least 4 or more players (i.e. 2vs2 and onward) gives a average combined score of 16.57 with a standard deviation of 2.21 meaning that on average a game is expected to score 10:7/10:6 and typically within 10:4 and 10:9 which actually makes any score bellow 10:4 and statistical anormal result. data computed over 1339 recorded games so it should be a meaningful and significant sample. overall in a quick analysis those look like perfectly fine and realistic numbers to me. I mean, you can see the raw output as it came out of from the database computation, the scores were so close a third float was even needed to tell them apart.

Code: Select all

+-------+--------+--------+---------+---------+----------+----------+
| games | avgRed | stdRed | avgBlue | stdBlue | avgScore | stdScore |
+-------+--------+--------+---------+---------+----------+----------+
|  1339 |  8.284 |  2.321 |   8.282 |   2.319 |    16.57 |     2.21 |
+-------+--------+--------+---------+---------+----------+----------+

[...jutuli...]

Balance is really weird sometimes. It often happens that one team has all the people and the other team has half the bots, and it lasts for two rounds. Then the balance recovers and distributes people evenly between the two teams. I remember just yesterday this happened, then Impre started to get upset about balance - it was a T2 map and one team had 2 or 3 bots, the other only humans. Too bad I didn't think to take a screenshot, but it really happens quite often and it's not a problem to lose two rounds because of it...

[adminless]

I think that has also been discussed/explained other times, that's the way it's, the auto balance can not start breaking teams/games and swapping people like that in the middle of a play every time some new player joins and/or some random leaves. it just doesn't work like that and that would be a much worse situation that having to go some bad rounds as nobody wants to be switched like a dozen times in a game for the sake of it so the auto balance has to wait for proper opportunities like when the key players die to make adjustments so yes it can take some time. in that context then yeah it can perfectly happen that a game that was randomly assembled 3vs3 ends up being played 3vs3bots before the losing team had a chance to break it around but that's the way it's, you can not start changing people mid play just like that as I explained so in such cases you'll have to wait until game ends for the auto balance to correct which hopefully in such case it should hit fast. it's what it's, some public server games not some sort of competitive organized games so you can not expect every single instant to be nose bleeding. that said, I insist when you look at the numbers as I posted it's a proven fact that the auto balance works and it delivers sensible results. just for comparison I just made the same computation now for the ranked games, 417 in total, that feature not one but two layers of human level balance (first only top or better players go there to begin with and second those games are setup with a balance target) and the results weren't even really any better overall (blue team score avg/std was pretty much the same however the red team score was slightly imbalanced there) and only marginally better in some aspects (16.69 ±2.07 against 16.57 ±2.21 that's not a significant difference) so I don't even think that you can get much better numbers than that even if you were to manually review/organize every game, which is not even realistic. you can check the raw ranked output in that aspect by yourself down bellow if you want, those were actually in fact fantastic results for some random public games as you can see there.

Code: Select all

+-------+--------+--------+---------+---------+----------+----------+
| games | avgRed | stdRed | avgBlue | stdBlue | avgScore | stdScore |
+-------+--------+--------+---------+---------+----------+----------+
|   417 |  8.408 |  2.142 |   8.285 |   2.273 |    16.69 |     2.07 |
+-------+--------+--------+---------+---------+----------+----------+

[Lisa_Ann]

Balance is really weird sometimes. It often happens that one team has all the people and the other team has half the bots, and it lasts for two rounds. Then the balance recovers and distributes people evenly between the two teams. I remember just yesterday this happened, then Impre started to get upset about balance - it was a T2 map and one team had 2 or 3 bots, the other only humans. Too bad I didn't think to take a screenshot, but it really happens quite often and it's not a problem to lose two rounds because of it...

Yes, I can confirm the phenomenon, in fact in the morning hours there are 4 players and for a while I am alone with 2 sticks and on the other side there are 3 players. And then I switch over if I can win...

[imprecise]

Hi Adminless

I remember that this has already been discussed several times but unfortunately since there has been no change I am not satisfied with a certain number of games played. I know you can't solve everything but perhaps you could consider improving the algorithm that manages the distribution of players in order to generate slightly better balanced teams.
I was sure that you would answer me with figures and statistics, it seems logical since this is what allows you to analyze the games played, but know that this does not always represent the experience on the server... As a player on FreeFun I can confirm that a certain number of games are unbalanced due to team balance for different reasons. This happens when several players leave the server to join the UnFreeZE server and in this case the distribution of players is distorted, it also happens in the opposite direction when several players arrive on the FreeFun server, the distribution of these arriving players is not always the most appropriate. Sometimes this works out after 1 or 2 turns of play and sometimes it takes more than 5 turns of play to see a change happen and not always the best change or the one hoped for.
As for the distribution of pro players or good players, it is not uncommon to find them together in the same team, but I concede that this is not systematic, it makes for painful games and if in stats you observe results that satisfy you, it is because in this case some pro players and good players lower their level of play somewhat, while the other players try with difficulty to survive with the hope of gaining 1 point. Even if we like the challenge, it still has to be possible to achieve it...
You say that you don't want players to be swapped several times in the same game, but that's what happens from time to time, I've already seen players move from one team to another at least 4 or 5 times in a game, that but also happened, it's not always pleasant but I think it's less worse than playing pros against noobs and finishing with a score of 10 - 1
Thank you for responding seriously and precisely, with supporting figures, and thank you for all the work you do for these servers. I would sincerely like you to try a change, an improvement to my request, if it doesn't work it will always be possible to return to the current state I imagine

[adminless]

as I said I might try to look at it further at some point but also as you have noted yourself and even as the data shows itself if after being discussed other times I didn't really make any further changes it's because I don't really see much room for improvement in that regard. for the rest I know, as I said, you can not pretend/expect every moment in a public server to be top notch, people come as people go, there's really no much you can do about it and start making more switches because of that I don't really think it will make things any better if not probably the contrary. in any case, thanks for the input.