38 replies to this topic

[Funkadelic Mayhem]

well?

Edited by Funkadelic Mayhem, 30 October 2013 - 04:21 AM.

[Joseph Mallan]

Only reason I say no is that we don't have the ability to drop with the same players every time. So ranking a team Elo would be worse than figuring how to implement Short Run SPC.

[Kunae]

Joseph Mallan, on 30 October 2013 - 04:22 AM, said:

Only reason I say no is that we don't have the ability to drop with the same players every time. So ranking a team Elo would be worse than figuring how to implement Short Run SPC.

While I see where you are coming from, it isn't ranking the team as a whole, it's just giving you an Elo score that is based on your play time when you are in a group.

[Funkadelic Mayhem]

Kunae, on 30 October 2013 - 05:57 AM, said:

While I see where you are coming from, it isn't ranking the team as a whole, it's just giving you an Elo score that is based on your play time when you are in a group.

Yep, this!!

[FinsT]

Funkadelic Mayhem, on 30 October 2013 - 06:10 AM, said:

Yep, this!!

Nope, not this. Like Joseph said, we play in different groups. And mind you, personal performance is MUCH dependant on group's performance. So, if half the time i drop with 3 blind monkeys and other half of the time i drop with 3 super-pilots, - my "team-based" elo will average to... average. How does that reflect inability of blind monkeys - or super-skills or 3 super-pilots i drop with half the time? It doesn't.

See, makes no sense.

[Funkadelic Mayhem]

FinsT, on 30 October 2013 - 06:21 AM, said:

Nope, not this. Like Joseph said, we play in different groups. And mind you, personal performance is MUCH dependant on group's performance. So, if half the time i drop with 3 blind monkeys and other half of the time i drop with 3 super-pilots, - my "team-based" elo will average to... average. How does that reflect inability of blind monkeys - or super-skills or 3 super-pilots i drop with half the time? It doesn't.

See, makes no sense.

you must only drop on comstar with randoms... ?

[Roland]

Someone else recently described the correct way to do this.

You basically generate an Elo score for each specific combination of players.

The result is that when you play with a new group, it'll start you in the middle, and you may get some bad matches, but since Elo tends to adjust itself very quickly, this isn't a problem.

The end result is that you have a specific Elo rank for not just "group play" but for each specific group of people that you play with.

I believe someone mentioned that this is how other team based games use Elo.

[Kunae]

Roland, on 30 October 2013 - 06:30 AM, said:

Someone else recently described the correct way to do this.

You basically generate an Elo score for each specific combination of players.

The result is that when you play with a new group, it'll start you in the middle, and you may get some bad matches, but since Elo tends to adjust itself very quickly, this isn't a problem.

The end result is that you have a specific Elo rank for not just "group play" but for each specific group of people that you play with.

I believe someone mentioned that this is how other team based games use Elo.

While that sounds best... can you imagine the database bloat that would cause for PGI?

[MonkeyDCecil]

All premades should be put into the highest ELO. I say this since voice coms give and unfair advantage.

[Joseph Mallan]

MonkeyDCecil, on 30 October 2013 - 07:00 AM, said:

All premades should be put into the highest ELO. I say this since voice coms give and unfair advantage.

[Joseph Mallan]

Funkadelic Mayhem, on 30 October 2013 - 06:26 AM, said:

you must only drop on comstar with randoms... ?

No true. I dropped with the Law for months and I can say without a doubt, that certain members Inflate the groups Elo to where I become... a liability to the group as a whole. Soon as that individual leaves an someone "more my speed" shows up Our game gets... easier.

[Roland]

Kunae, on 30 October 2013 - 06:45 AM, said:

While that sounds best... can you imagine the database bloat that would cause for PGI?

Dude, it's rows in a database.
Databases are MEANT to deal with that kind of thing. That's why they exist. That's how they work.

The idea that it would take up too much space is silly... Buy another hard drive. The cost of the space increase would be trivial. The idea that it would cause a performance problem is equally trivial. If it does, it means your database is garbage, either in design of the database or in the database itself. Get a better database or get a better database designer.

Much MUCH larger games, with much MUCH larger player bases, do it as described.

[Kunae]

Roland, on 30 October 2013 - 07:08 AM, said:

Dude, it's rows in a database.
Databases are MEANT to deal with that kind of thing. That's why they exist. That's how they work.

The idea that it would take up too much space is silly... Buy another hard drive. The cost of the space increase would be trivial. The idea that it would cause a performance problem is equally trivial. If it does, it means your database is garbage, either in design of the database or in the database itself. Get a better database or get a better database designer.

Much MUCH larger games, with much MUCH larger player bases, do it as described.

3 little letters for you...

P G I

I wouldn't be inordinately surprised to find out their back-end was written in MS Access.

[Almond Brown]

Easier way is to have just a 1 v 1 Pool. Losers goes to the UnderHive and need a 25 Wins in a Row streak to get back into the 1 v 1 Trial Pool again. Win and you move up, lose, back to the UnderHive. Just be sure your Leet arse does not have a bad game on your first visit to the 1 v 1 Pool. LOL!

It is a game ffs.

Edited by Almond Brown, 30 October 2013 - 07:48 AM.

[Screech]

The acoustic version.

[Magna Canus]

I think the big question here is... is this really necessary? Would development time spent on this make the game any more enjoyable than time spent on something else (CW for example)? I say no in both cases. I can think of thousands of "nice to haves" that have a higher priority.

The current ELO system is pretty borked as it is and you want to apply it to groups as well?

[Chronojam]

The newest FPS game EA has released, Battlefield 4, has had some good attention given to its matchmaking system. The developers have gone with a hand-tuned solution instead of trying to use a raw rank or Elo due to the pitfalls of relying on an Elo metric in a teams-and-squads environment.

Quote

Q: Microsoft is planning to take on some of the issues of online play with a ratings system which will separate players who have been reported as abusive, disruptive or otherwise unpleasant. Does that hold any concerns for you from a matchmaking perspective?

Patrick Bach: Yes, but we are adding a skill system. Previously we only had rank, but rank can just be a reflection of time. So, rank doesn't really show how good you are, just how used you are to the game. We've been looking at various ways of calculating actual skill - skill being more of a subjective thing that we're trying to objectify and measure. We'd previously used an ELO system, which is used in chess rankings, but that didn't work out. We've actually created our own system, now, which seems to work - we'll have to see when we launch the game. You'll actually be matched based on skill, not on rank. We're looking into more ways to do that, because it's more fun, playing with people at your level.

[Funkadelic Mayhem]

Quote

Q: Microsoft is planning to take on some of the issues of online play with a ratings system which will separate players who have been reported as abusive, disruptive or otherwise unpleasant. Does that hold any concerns for you from a matchmaking perspective?

Patrick Bach: Yes, but we are adding a skill system. Previously we only had rank, but rank can just be a reflection of time. So, rank doesn't really show how good you are, just how used you are to the game. We've been looking at various ways of calculating actual skill - skill being more of a subjective thing that we're trying to objectify and measure. We'd previously used an ELO system, which is used in chess rankings, but that didn't work out. We've actually created our own system, now, which seems to work - we'll have to see when we launch the game. You'll actually be matched based on skill, not on rank. We're looking into more ways to do that, because it's more fun, playing with people at your level.

Very interesting. The ELO matching system feels for 56k dial up. Its good to see modern game developers thinking outside the box and pushing towards the future.

Edited by Funkadelic Mayhem, 31 October 2013 - 09:35 AM.

[HammerSwarm]

This is so simple to fix PGI Let Me Help Yah! (Contents of this and all other forums posts are free to use)

ELO(TEAM) = ((ELO(P1) +ELO(P2)+ ELO(Pn)....)/(N))*X

Start X at one and bump it up as necessary based on telemetry as to how much being in teams matters

Average team ELO * Average team victory increase

Example
Team A has players of the following skills: 1, 4, 5, 10
Teams win on average 25% more than individuals
((1+4+5+10)/4)*1.25 = 6.125 for Team ELO

Team B has players of the following skills: 1, 1, 1, 13
Teams win on average 25% more than individuals
((1+1+1+13)/4)*1.25 = 5 for Team ELO

[Abivard]

Dump ELO and bring back Weight matching.

I am not sure ELO matching really exists, PGI tells us it is there but won't show numbers. I really do not see it in pug drops, Did not really see it in premades nor 12 mans. Seems totally arbitrary as to who gets matched up.

If ELO really exists, then the player base would seem to be to small to support it in actual practice, so we have this monstrosity that more often than not leads to stomps.

Maybe the stomps balance out as W/L for you, maybe they don't, but the point is stomps should be RARE! not common place.