Best Of

Just a few thoughts, including a summation of what everyone has been discussing. The best way to introduce these changes is via server options. If any of my interpretations re: moultano's proposals are wrong, please let me know.

Each player will have two skill values. The first is equivalent to the current skill value. The second is an offset that turns the average into their skill for each team when added or subtracted.

marine_skill = skill + offset
alien_skill = skill - offset

The skill values this learns without the offset are backwards compatible, so shuffle algorithms will not need to be modified. Despite the fact that each player will still only have one skill value for balancing, there is still a benefit to doing this. A player’s skill will no longer randomly walk based on their team assignment. It will always be the average of their marine and alien skill, rather than being effectively a weighted average of their skill for the teams they’ve played on recently. Additionally, by collecting this data, and returning it from hive, we can open it up to modders to balance teams more aggressively if they desire.

1. Introduce the following shuffle modes:

Aggressive:
Shuffle is determined by averaging player skills which have been calculated using the offset. Of a player’s two skill values ([skill + offset] and [skill – offset]), whichever value results in the lowest team standard deviation will be used. For a given server population, this system seeks the most statistically "even" game possible with the lowest possible standard deviation for each team. Servers running this mode will see team win-rates move toward 50/50 (subject to the limits of hive accuracy). Better players will generally be forced to play on the team they're worse at (and vice versa).

Moderate:
Shuffle is determined by a player's [skill + offset] and [skill - offset] values, but with the server’s total average offset subtracted before shuffling. For a given server population, this system seeks a blend of the aggressive and classic modes by attempting to reduce the severity of certain drawbacks of either. Servers running this mode will see team win-rates that reflect the average difference in skill on the server. Most team assignments will be random. Exceptional players may still be forced to their worse team.

Classic:
Shuffle is determined by using each player's average skill (without a + or - offset). This mode is essentially the current shuffle. For a given server population, team assignments will be uniformly random regardless of individual player skills.

2. Allow server operators to select the mode they want to be the standard shuffle mode for their server.

3. Allow server operators to force shuffle using whichever mode they want, even if it's not their server's standard shuffle mode.

4. Indicate a server's standard shuffle mode in the server information window of the server browser.



1. Give players the option to set a default team preference in their settings.

2. Give players the ability to select a team preference on a per-round basis. Do this by allowing them to switch their default preference (if set) at the start of each round, before a shuffle vote. To select or update a preference, there would be a simple on-screen interface allowing them to select marines or aliens. If they've set a default team preference in their settings, it will be selected by default. If they haven't set a default preference in their settings, no option will be selected. If they join a team before the shuffle vote is complete, that team would be locked in as their preference.

3. Was the player placed on their preferred team? After each shuffle, record the resulting user statistic as a preference_mismatch value. The goal is to keep track of the historical percentage of rounds that a player isn't successfully placed on their preferred team. Update this value once per shuffle. Commit this data to the historical total only once per round, at the end, using the most recent value generated. If a player switches teams after getting placed on their preferred team, it shouldn't be recorded as a preference mismatch. If a player is forced onto their non-preferred team because another shuffle vote happens, then it should be recorded as a preference mismatch.

4. Prevent gaming the system by only recording preference values for rounds longer than a specific duration - maybe 5 minutes?

5. Let server operators set a 'team preference' flag. This would introduce team preference data to their chosen shuffle algorithm. Not every server would have this flag enabled. In servers ignoring team preference data, your team preference and result would still be recorded, but it would have no influence over that particular server's shuffle results.

6. Don't use a strict karma system. On servers running this mode, allow operators to designate a % value as the server's team preference override threshold. The system would correct the problem of a player consistently being placed on their non-preferred team by giving them their first choice in the event that their historical preference_mismatch value crosses the threshold. In the event that multiple players in a server cross this threshold in a given round, randomly assign them to their preferred teams until the maximum allowed standard deviation is reached (see next point).

7. Allowing per-user team preference to impinge upon the shuffle algorithms comes at a cost in the form of increased standard deviation, and thus reduced balance (with respect to each algorithm's goal). So server operators would need to be able to establish a maximum allowed standard deviation value which would limit preference-honoring whenever doing so would raise a team's standard deviation above that value.

Pack it away boys, it's solved

SquishpokePOOPFACE wrote: »

Pack it away boys, it's solved

Fair enuf, I'll make a note of this in the book of "git gud scrub!"