Please upgrade your browser for the best possible experience.

Chrome Firefox Internet Explorer
×

Time to dicuss the matchmaker?

STAR WARS: The Old Republic > English > Galactic Starfighter
Time to dicuss the matchmaker?

LordFell's Avatar


LordFell
12.17.2017 , 10:40 PM | #11
Quote: Originally Posted by HrRav View Post
And yes, 4v4 matches with a counterpart to the ground game grouped rank would probably help as it would give the best players a queue of their own.
To some degree, it's difficult for the game to give true rankings. Like... let's say the game wants to place 2 aces, four "genuine flyers" and 2 zero-killers in a match. I think.... by some metrics that game would consider me an ace. I've been flying a long time, most of the ships in my hanger are Mastered, I've scored Aces in matches (infrequently) and am frequently "promoted" to Ops Leader.

...but I would NOT agree that I'm an Ace. I belong in that middle, "genuine flyer" group. I had a match last week though, where I scored 11 kills and 9 assists. Did I have a good flight? Probably... also, did I gorge on zero-kill feeders? Also, probably.

The presence of CXP farmers is skewing our numbers.

Drakkolich's Avatar


Drakkolich
12.17.2017 , 11:51 PM | #12
Quote: Originally Posted by LordFell View Post
To some degree, it's difficult for the game to give true rankings. Like... let's say the game wants to place 2 aces, four "genuine flyers" and 2 zero-killers in a match. I think.... by some metrics that game would consider me an ace. I've been flying a long time, most of the ships in my hanger are Mastered, I've scored Aces in matches (infrequently) and am frequently "promoted" to Ops Leader.

...but I would NOT agree that I'm an Ace. I belong in that middle, "genuine flyer" group. I had a match last week though, where I scored 11 kills and 9 assists. Did I have a good flight? Probably... also, did I gorge on zero-kill feeders? Also, probably.

The presence of CXP farmers is skewing our numbers.
The only thing the matchmaker looks at is highest requisitioned ship (Highest being mastered), numbers of games played on the character and how long you were in queue.

Because of this if a player can't hit anything at all but has played 1000 games and another player has a 95% win rate and like 10 kill/death ratio but has only played 500 games. The matchmaker considers the 1000 games player "better".
DrakolichDrakolích
The BastionTwitch Stream

HrRav's Avatar


HrRav
12.19.2017 , 05:06 PM | #13
Quote:
The only thing the matchmaker looks at is highest requisitioned ship (Highest being mastered), numbers of games played on the character and how long you were in queue.
So what should the matchamker look at then? I would suggest two main attributes:

  • Accountwide (or at least legacywide) kill/death ratio - Reasonable measure of player skill level. Could also include assists in order not to punish players who go for support ships. This ratio will also be self-adjusting---if you are put in too tough matches your kill/death ratio will slowly decrease and eventually the matchmaker will match you against less deadly opponents.

  • Requisition on your top three ships - If two equally skilled players are playing against each other, but one has a new character with ships straight out of the hanger and the other has fully mastered ships then the latter player will have a definite advantage. This could be compensated by multiplying your kill/death ratio by a factor depending on how much requisition you have on your best ships. For example, if you have zero requisition you could get a 50% reduction of the kill/death ratio VALUE considered by the matchmaker.

LordFell's Avatar


LordFell
12.19.2017 , 06:01 PM | #14
That sounds good. Give each character a ranking based on Legacy Wide KDA, Mastery Percentage of ships in hanger... and maybe a "hot streak" past ten (or twenty) game Win/Loss.

...I think Legacy KDA is probably the biggest factor that should determine a players overall ranking when match making though.

phalczen's Avatar


phalczen
12.19.2017 , 06:16 PM | #15
I doubt a legacy wide value could be considered for probably the same reason that ELO for the ground is character specific. I would imagine it has something to do with database lookups. Just consider how long it takes to open your legacy bank these days or how you have to open, close, and then re-open crafting crew skill windows to get the system to display the right number of things you can craft. We know that the matchmaker doesn't even attempt to pop a match unless there are 8 people in queue on both sides. Once you reach that critical mass of players queued, the match doesn't pop instantly ... probably because its figuring out the other factors that Drako mentioned. Adding in a database lookup to compute K/D ratio or worse (K+A)/D ratio legacy wide, and then use those figures to pair the groups, probably causes an unacceptable delay in queue popping. Your ground ranked pvp ELO can't even be calculated unless you have completed ten matches for that season, suggesting that ELO is computed post match, to minimize impact on time-to-queue-pop. Then that variable is "attached" to your character the way your discipline is, and then with wait time determines when your queue pops and with whom. Much easier to compare a static number that only is updated after the match.
If you think I've made a good contribution with this post, I kindly ask that you use my Refer a Friend link! Here is more information about the program.

Ramalina's Avatar


Ramalina
12.20.2017 , 01:58 PM | #16
@placzen

The 10 match requirement for ground PvP rankings doesn't really have anything to do with the time it takes for loading and processing data. Ranking algorithms need to have a certain minimum sample size of match outcomes before they can estimate player skill with enough accuracy to be useful for matchmaking. It would still require you to complete those 10 matches even if data access and computation were instantaneous.


The compelling argument against a legacy based ranking function for GSF is that you'd want to recycle as much of the ground PvP ranking code as you could to save time, and because of class differences the ground PvP ranking system has to be character based, not legacy based.

It's not about hardware time, it's not about our time in queue, it's about payroll time.
"A padawan's master sets their Jedi trial, Rajivari set mine."
- Zhe Lian, Sage.

Twitch

phalczen's Avatar


phalczen
12.21.2017 , 10:06 PM | #17
Quote: Originally Posted by Ramalina View Post
@placzen

The 10 match requirement for ground PvP rankings doesn't really have anything to do with the time it takes for loading and processing data. Ranking algorithms need to have a certain minimum sample size of match outcomes before they can estimate player skill with enough accuracy to be useful for matchmaking. It would still require you to complete those 10 matches even if data access and computation were instantaneous.
I know that. My point was that the variable is calculated after the first ten and after each subsequent match and is then "static" while you are queued. You cannot change disciplines or utilities while queued for ground pvp, just like you cannot change your ship builds in the hangar while queued. They become "static" variables as far as the matchmaker is concerned. Similarly, your ELO is not calculated when the matchmaker is attempting to form a match. It's just "looked up" along with your discipline.
If the matchmaker had to check each person's legacy in order to form a valid match, that would generate extra overhead, and I think that would slow down queue pops. I could be wrong though.

Quote: Originally Posted by Ramalina View Post
The compelling argument against a legacy based ranking function for GSF is that you'd want to recycle as much of the ground PvP ranking code as you could to save time, and because of class differences the ground PvP ranking system has to be character based, not legacy based.

It's not about hardware time, it's not about our time in queue, it's about payroll time.
Perhaps. If ground ELO was legacy based, a player could get onto a as-yet-unranked alt and be matched with higher ranked people, resulting in a much larger boost in ELO if they won. I'm not sure that would necessarily be exploiting the system, but I doubt its intended. But hey, they don't enforce a gear rating for ranked ground pvp, and its incentivized such that the rewards enable acquisition of the gear you should already have in order to compete at the highest levels ... its sort of backwards. So who knows what they really intend with the system beyond an attempt to match the highest skilled players.
If you think I've made a good contribution with this post, I kindly ask that you use my Refer a Friend link! Here is more information about the program.

Verain's Avatar


Verain
12.22.2017 , 08:39 PM | #18
Nothing should be based on legacy or account. Everything should be based on character. This is a free to play game, and GSF is the most free to play part of it, given that it offers ALL of GSF to a totally f2p character. That means if someone wants to smurf, they'll just spam no-account accounts.

If hangars were account wide, or accounts cost money to create and maintain, you could make a case for this. But none of that is true.
"The most despicable person on the GSF forum."

HrRav's Avatar


HrRav
12.23.2017 , 03:38 AM | #19
Whether or not account/legacy wide rating is feasible is not the crucial part of the matchmaking. (I guess most pilots have more or less the same stats on all characters anyway.) As I see it, the main issues are that the matchmaker needs to use an appropriate skill measure and that it should not start a game as soon as there are 16 players in the queue if there are other ongoing matches on the server - in that case it is bettter to wait until that game has ended and try to balance the teams from a pool of 32 players.

Flusssaeure's Avatar


Flusssaeure
12.25.2017 , 04:53 AM | #20
Well after depending on time slot and server there often isn't more than one gaming running at the same time and people who qued while the game is running have to wait until the game ended that enough people are in que to start a second game I don't think that having to wait that a game ends will change the matchmaking enough to make a difference. Especially when you consider that not every one of the first game will que again because daily full / game was to one - sided / having to go to lunch; dinner; bed and not every one who listed later is patient enough to wait for the full 15 Minute duration.

And what about the prime time when two or three games running at the same time? Should one time out match than prevent every one on the server from getting a match because the other games ended earlier but still the people from the earlier ended matches and the player qued later have to wait for the time out matches to get a new game.

The biggest sin of the matchmaker today is that it priofities new player and groups at the same time. Because it forces veterans with many matches to group to get a match at all but then pairs them against a random group of new players who also have priority. This together with the small player pool is one of the main sources for unbalanced matches.