Return of Reckoning

Spawncamping solo queuers night after night and hoping we get a good fight will have our crew more likely disbanding for good then us loosing outnumbered vs random every now and then

zumos2 wrote:Another issue I see is people leaving pops because they know who they are facing. I don’t know the reason why the RoR-team made it so you can see who you are facing when the queue pops, but I think it would be better without it. Making a good matchmaking system when people just leave when they see hard opposition is never going to work. And that is exactly what I would like to see: a good match-making system. I’m not an expert in this, but I think there are definitely opportunities in making a good matchmaking system.

This was debug text for my benefit that I never got around to removing. It appears if a scenario pops with at least one group having been matched. It can be removed if it is causing issues.

zumos2 wrote:<role balancing>

The scenario manager algorithm attempts to balance roles only when adding solo players. I'm planning to refine this a little for groups, both by using the composition of a group as a weight when adding to the queue (to ensure that for example a 3 DPS group is prioritized below a balanced group) and by ensuring that the composition which comes closest to resolving any existing role imbalance in a scenario is used for a given player count fill slot, rather than the composition which includes the highest number of players in an individual group.

The matchmaker logic for groups is, broadly:

- For both sides, try every unique combination (i.e. NOT permutation) of group sizes which add up to a number between 2 and the maximum player spots left in the scenario, using the first available groups in each sublist and prioritizing combinations with the highest individual value for size

For example for 6 players:

[6] Priority 1
[4, 2] Priority 2
[3, 3] Priority 3
[2, 2, 2] Priority 4

- For any unfilled slots, use first group value encountered when descending from that slot (for example, empty 5 and 6 player slots above a 4 would use the 4 combination that was found)

- For any such slots, check with the solo queue whether or not any shortfall in players could be addressed using solo players, and if so, mark these slots as usable

- For a given slot X on realm A, check whether there exists such a slot Y on realm B which can be used with X such that the final playercount in the scenario is within 1 on either side, and if no such pair exists, return

- If a slot contains only groups, add immediately

- If a slot requires solo players, add the groups first, then attempt to pull solo players to match any role imbalance

Someone will probably ask: Why only match the first groups? Two reasons:

- It's a scenario queue, not a scenario pool - priority is given to those who queued first to preserve the important Queue Time element

- Optimization - Trying every permutation of groups in the queue, which would optimally solve for balance, is expensive, especially if balance needs to be compared (vector magnitude operations)

This 403 crap when attempting to edit posts makes me really salty, as I'm a compulsive editor

It's important to note here that since larger compositions already have priority over smaller ones, you can help your realm and the balancing by forming larger groups with more balanced compositions, as the system will not optimally handle smaller groups. A [6] is very likely to be balanced, whereas a [2, 2, 2] could easily be composed of 5 DPS and a healer who are trying to dodge the solo queue logic which is weighted against the large numbers of DPS.

Have you considered dropping the FIFO for a more scenario pool like version? Long term I think most would gladly trade insta with bad comps for high quality fights.

Risk of dropping FIFO is that as mentioned the queue system is violated. I intend to partially violate it via the measures I indicated above (penalizing badly-composed individual groups and prioritizing fragmented compositions over more contiguous ones if such would result in better role balance) but the idea that you may never actually get a pop at all if your composition is bad does not sit well with me, because the queue is not transparent.

add a persistence bonus

Just a thought.... Would not MORE people opt into doing Scenarios if the rewards were equal?

1. Allow 40 > players to get the same (end of SC) bonuses as < 40
2. Allow for officer medallion and anni gear drops (advanced to conq then inv then wrld etc as they become available)

Before going through the hardship of defining and coding brackets (a work heavy solution) why not try a more easier one?

I have a feeling more people would start doing sc's, thus allowing for faster que times, more people would form grps, thus allowing for more balanced play.

just an idea ... don't kill the messenger

The problem is equal size random groups with good comp will loose 999/1000 vs our team.

A super queuer can match premades vs premades more often make good compositions. Neither of this solves the problem of unbalanced fights happening way to often.

Giving random teams a numerical advantage when they are fielded vs premades can give random teams and premades higher quality SCs.

Bottom line premades such as ours must loose more often if a mixed queue is ever going to work

Azarael wrote:

Spoiler:

zumos2 wrote:Another issue I see is people leaving pops because they know who they are facing. I don’t know the reason why the RoR-team made it so you can see who you are facing when the queue pops, but I think it would be better without it. Making a good matchmaking system when people just leave when they see hard opposition is never going to work. And that is exactly what I would like to see: a good match-making system. I’m not an expert in this, but I think there are definitely opportunities in making a good matchmaking system.

This was debug text for my benefit that I never got around to removing. It appears if a scenario pops with at least one group having been matched. It can be removed if it is causing issues.

zumos2 wrote:<role balancing>

The scenario manager algorithm attempts to balance roles only when adding solo players. I'm planning to refine this a little for groups, both by using the composition of a group as a weight when adding to the queue (to ensure that for example a 3 DPS group is prioritized below a balanced group) and by ensuring that the composition which comes closest to resolving any existing role imbalance in a scenario is used for a given player count fill slot, rather than the composition which includes the highest number of players in an individual group.

The matchmaker logic for groups is, broadly:

- For both sides, try every unique combination (i.e. NOT permutation) of group sizes which add up to a number between 2 and the maximum player spots left in the scenario, using the first available groups in each sublist and prioritizing combinations with the highest individual value for size

For example for 6 players:

[6] Priority 1
[4, 2] Priority 2
[3, 3] Priority 3
[2, 2, 2] Priority 4

- For any unfilled slots, use first group value encountered when descending from that slot (for example, empty 5 and 6 player slots above a 4 would use the 4 combination that was found)

- For any such slots, check with the solo queue whether or not any shortfall in players could be addressed using solo players, and if so, mark these slots as usable

- For a given slot X on realm A, check whether there exists such a slot Y on realm B which can be used with X such that the final playercount in the scenario is within 1 on either side, and if no such pair exists, return

- If a slot contains only groups, add immediately

- If a slot requires solo players, add the groups first, then attempt to pull solo players to match any role imbalance

Someone will probably ask: Why only match the first groups? Two reasons:

- It's a scenario queue, not a scenario pool - priority is given to those who queued first to preserve the important Queue Time element

- Optimization - Trying every permutation of groups in the queue, which would optimally solve for balance, is expensive, especially if balance needs to be compared (vector magnitude operations)

I'm gonna read it carefully tomorrow and see if I can think of some concrete improvements.

Bozzax wrote:The problem is equal size random groups with good comp will loose 999/1000 vs our team.

A super queuer can match premades vs premades more often make good compositions. Neither of this solves the problem of unbalanced fights happening way to often.

Giving random teams a numerical advantage when they are fielded vs premades can give random teams and premades higher quality SCs.

Bottom line premades such as ours must loose more often if a mixed queue is ever going to work

6 man premade's are designed to take on bigger groups!
A well coordinated 6 man group is going to lick their chops for the chance of taking on 12 pugs !
If 6 vs 12 ever is implemented chances are 6 man premade's is still going to be annoying against pugs in sc play !