I believe it would be a welcome addition to some hosts to have an option for forcing even teams. Meaning that a game would refuse to start until teams have equal number of players. It's not a problem in regular casual games, but it was a huge disadvantage for teams during high rank games today. I'm not sure why, but as it is an extra commander and one less player to resource share with is considered equal to the highest rank player on a team. This just not true in practice and makes no sense since the highest ranked player doesn't get resource income of two players.
+4 / -0
|
quote:
is considered equal to the highest rank player on a team |
To the best of my knowledge this is incorrect; the balancer attempts to make the average rating on both teams equal.
+0 / -0
|
I have manually added up ratings for exactly one game with uneven teams and came to the conclusion I had written, that the highest ranked player was counted twice. I don't know how it works internally and my sample size is lacking. Regardless, in high rank games a missing player means missing micro/macro, which is irreplaceable in such games.
+2 / -0
|
I agree that just not playing such games is preferable to trying to find some way to balance them.
+1 / -0
|
I much prefer even numbers each side, but I do wonder if the dual comm situation is a niche that with sufficient practice someone could reliably achieve a decisive effect with. There have been a few occasions where I've had the extra comm inflicted on me and have been able to leverage that extra early game weight to my advantage or realised afterwards that I could have used it (e.g.   B789973 9 on Trojan Hills v05) It doesn't make up for lacking an extra pair of eyes and hands a large majority of the time but it isn't always a nightmare. Perhaps it could be made less nightmarish still if some effort was made to compensate for the smaller team's APM penalty with more income. It's not enough to simply match the other side. As it stands, the game grants 2 m/s to each player regardless of whether they have comm or not and I seem to recall reading that the player with 2 comms doesn't get extra innate m/s. Perhaps a multiplier could be applied so that with just one comm you get your 2m/s, but each additional comm multiplies that figure. Perhaps with 2 comms you could get 6m/s. Maybe the APM deficit is worth even more metal than that? Anyone who plays the game for a while knows that there are actually at least four resources in this game, not two. Time and attention are at least as valuable as e and m.
+0 / -0
|
quote:
I much prefer even numbers each side, but I do wonder if the dual comm situation is a niche that with sufficient practice someone could reliably achieve a decisive effect with. |
That's not even the issue. The issue is that one player gets 2 coms and 2 facs but only 1 player worth of resources to actually spend on units. Even if you're technically strong enough to stand in for 2 players it doesn't do you any good.
+1 / -0
|
quote:
I have manually added up ratings for exactly one game with uneven teams and came to the conclusion I had written, that the highest ranked player was counted twice. I don't know how it works internally and my sample size is lacking. |
I wouldn't trust the precision of any user-facing numbers when talking about what the backend of the rank system thinks that it is doing. Giving the double-commander player two resource shares sounds reasonable. It would at least incentivise the experience.
+2 / -0
|
IF the system is in fact counting the two comm player twice, I think it's a little ridiculous. IMO the two comm player team should have a slight elo edge as it's far harder to compensate for the lower player count, especially on certain combinations of map and team size (the bigger the map size, the more of an advantage the normal team has). In my experience, if the lower player count team does not achieve great trades initially or gain initiative at the start, they'll fail. This is less important in larger team sizes and far more important in lower team sizes. Getting stuck with weaker players to offset the rush potential seems a bit backwards. I'd prefer to see uneven games play out more than just a rush or resign thing and I feel that getting more competent help (even a very slight edge) could help improve the experience for the uneven team player and the state of the game overall in these situations (turning it from a "rush success or resign" into a normalish game with less rush success emphasis).
+0 / -0
|
Why are you theorizing on things clearly defined in the code? Teams are balanced by average Elo. The second Commander going to the strongest player is purely a bonus, actually it should be split up evenly among all players. E.g. go to the most average player in the team.
+2 / -0
|
http://zero-k.info/Battles/Detail/7924382 coms => push goliath => win bridge => gg
+0 / -0
|
quote:
Why are you theorizing on things clearly defined in the code? |
"Self-documenting" means not documented tbh.
+2 / -0
|
I'm just having a déjà vu, does nobody remember that thread?
+1 / -1
|
I don't like the solution of giving a player twice the resources of other players. In casual lobster pots it doesn't really matter that there's a missing player since most of the time the other team only gets one more newbie which becomes a penalty since they can't utilize their share properly.
+0 / -0
|
|
|
Even if it were the objective     DeinFreund's logic is fatally flawed; there is no particularly good reason to believe that in a game where the number of players on each team is uneven giving the commander to the most average player is a good idea with respect to any useful metric. Note that I do not consider "the average rating on each team" a useful metric in this case. It does a reasonable job of standing in for "overall skill of the team" and hence determining "the predictions of the balance algorithm" when the teams are even, but when one team has more people and can split their attention more ways the assumption doesn't hold up. (The game does have to be small enough and/or of high enough skill level that the APM factor actually matters.) EDIT: To take an edge-case example, in a game with three players rated 1800, 1900 and 2000 it seems fairly obvious to me that the most balanced and reasonable teams will be the two weaker players against the stronger player.
+1 / -0
|
I am not even sure if taking average rating is the way to go for even teams either. I have played and watched many, many clusterfucks and I think I can conclude the following: 2600 + 1000 is not going to win against 1800 + 1800 with a 50% chance it SEEMS. It´s way less in my perception. Since i am aware that this is pure guesswork, is there any way to get statistic data about this?
+0 / -0
|
Honestly I don't know how the game should behave for uneven player counts. All I can tell you is that the rating system assumes every player to contribute equally. All attempts I've made at incorporating skill disparity into the weighting have been unsuccessful. I have never specifically applied any rules for uneven teams, so feel free to experiment around. Maybe there is indeed some kind of penalty for teams with less players that would work. In general I'd prefer what the OP proposes: Not playing with uneven teams.
+1 / -0
|
Everybody has some "perception", the hard part is checking it and generalizing it (my perception: not all high rank ELO players do equally well when playing small teams, as it requires some different skills). Isn't the extra commander given because of the income per commander? Such that both teams get the same income in the same way. I agree though that data would greatly help everybody trying their "perception". Now that we anyhow have a public ladder, would it be ok to have a simple downloadable form of all the battle results (if someone would put the effort to write the code for the export)?
+0 / -0
|
|
|
Scrappers are one solution, but will be probably bad for server, inexact (everybody getting it at different times) and dependent on site design format. Was asking if anybody sees any issue with something embedded in the website and generated regularly, like once a week or a month. Of course someone should make the effort to implement it, but first should be clear if it is a good idea...
+0 / -0
|
