Pages: [1] :: one page |
|
Author |
Thread Statistics | Show CCP posts - 0 post(s) |
A B Ablabab
The Unholy Legion Of DarkStar DARKSTAR ARMY
61
|
Posted - 2013.07.10 02:06:00 -
[1] - Quote
I think the guys writing matchmaking are screwed, there's just an ungodly amount of junk to take into account.
SP won't work: Because players can easily spunk SP into useless junk, then have be stuck in a tier that's too high, and we'll have ENDLESS threads of 'I want a respec or I quit', as if we don't have enough of those already.
New players espically wont feel the need to specialise. They want to try new things, and that's only good for the game (Less ARs running around)
Fitting won't workYou can buy and fit suits while in battle, any system which takes suits into account will have people AFKing for the first minute so they can throw together a proto suit and suicide.
How do you balance someone with a proto fit that's BAD. Seriously I know in these forums a lot of people are calling each other out over bad fits, but just look at EVE. It's really easy to fit badly, or worse fit badly for the way you want to play. It's no good putting me up against real proto stompers when I've put all my SP into my brand new proto fit nova knife heavy. Know what I'm saying?
Suits won't work.On top of that we have to take into account suits. It's easy to think something like: "Just rank EHP vs DPS, then match on player skill" But if that ever happens no one will EVER want to run heavy. High damage weapons and solid EHP with no accounting for the drawbacks of the suit.
So. The question I ask y'all is what metrics can you really take into account? How do you spot a stomper from a nova knife heavy? How do you know when a player is having a bad day / running a bad fit / is just bad.
I await the knowledge of the crowd. |
Pregnancy-leave rituals
Mannar Focused Warfare Gallente Federation
660
|
Posted - 2013.07.10 02:08:00 -
[2] - Quote
Nerd. |
Pregnancy-leave rituals
Mannar Focused Warfare Gallente Federation
660
|
Posted - 2013.07.10 02:09:00 -
[3] - Quote
But what you said is true. The matchmaking is so bad that joking about it would only produce bad jokes. |
A B Ablabab
The Unholy Legion Of DarkStar DARKSTAR ARMY
63
|
Posted - 2013.07.10 02:10:00 -
[4] - Quote
Pregnancy-leave rituals wrote:Nerd. Yessir, by trade. Blowing people the f**k up is only a hobby :) |
Dr Allopathy
Raging Pack of Homos
110
|
Posted - 2013.07.10 02:14:00 -
[5] - Quote
If CCP attached something like metadata or w/e to different values in the game (stats from data drawn at the end of the match, skills spec'd into per player, etc), then they could probably come up with some algorithm which draws good probability from a selection of players on how well they will face another selection of players. |
A B Ablabab
The Unholy Legion Of DarkStar DARKSTAR ARMY
64
|
Posted - 2013.07.10 02:19:00 -
[6] - Quote
Dr Allopathy wrote:If CCP attached something like metadata or w/e to different values in the game (stats from data drawn at the end of the match, skills spec'd into per player, etc), then they could probably come up with some algorithm which draws good probability from a selection of players on how well they will face another selection of players.
+1
That's actually a great idea. Use a system like spam filters do that'd require that CCP go the route of 'recording everything that moves', that might be difficult in such a busy system.
If it worked we'd probably find some utterly strange correlations that suggested better players. I'm enough of a nerd to think that'd be fun to do even if it didn't fix matchmaking XD |
hooc order
Deep Space Republic Top Men.
563
|
Posted - 2013.07.10 02:32:00 -
[7] - Quote
A B Ablabab wrote:I think the guys writing matchmaking are screwed, there's just an ungodly amount of junk to take into account.
SP won't work: Because players can easily spunk SP into useless junk, then have be stuck in a tier that's too high, and we'll have ENDLESS threads of 'I want a respec or I quit', as if we don't have enough of those already.
New players espically wont feel the need to specialise. They want to try new things, and that's only good for the game (Less ARs running around)
Fitting won't workYou can buy and fit suits while in battle, any system which takes suits into account will have people AFKing for the first minute so they can throw together a proto suit and suicide.
How do you balance someone with a proto fit that's BAD. Seriously I know in these forums a lot of people are calling each other out over bad fits, but just look at EVE. It's really easy to fit badly, or worse fit badly for the way you want to play. It's no good putting me up against real proto stompers when I've put all my SP into my brand new proto fit nova knife heavy. Know what I'm saying?
Suits won't work.On top of that we have to take into account suits. It's easy to think something like: "Just rank EHP vs DPS, then match on player skill" But if that ever happens no one will EVER want to run heavy. High damage weapons and solid EHP with no accounting for the drawbacks of the suit.
So. The question I ask y'all is what metrics can you really take into account? How do you spot a stomper from a nova knife heavy? How do you know when a player is having a bad day / running a bad fit / is just bad.
I await the knowledge of the crowd.
Just make a meta score for levels in core skill (damage and health maybe some other crap). the score would be a number. Put low numbers in one tier put medium skills in another tier etc. The code to do this already exists to an extent with the academy. |
GTA-V FTW
Mannar Focused Warfare Gallente Federation
381
|
Posted - 2013.10.07 23:47:00 -
[8] - Quote
A B Ablabab wrote:I think the guys writing matchmaking are screwed, there's just an ungodly amount of junk to take into account.
SP won't work: Because players can easily spunk SP into useless junk, then have be stuck in a tier that's too high, and we'll have ENDLESS threads of 'I want a respec or I quit', as if we don't have enough of those already.
New players espically wont feel the need to specialise. They want to try new things, and that's only good for the game (Less ARs running around)
Fitting won't workYou can buy and fit suits while in battle, any system which takes suits into account will have people AFKing for the first minute so they can throw together a proto suit and suicide.
How do you balance someone with a proto fit that's BAD. Seriously I know in these forums a lot of people are calling each other out over bad fits, but just look at EVE. It's really easy to fit badly, or worse fit badly for the way you want to play. It's no good putting me up against real proto stompers when I've put all my SP into my brand new proto fit nova knife heavy. Know what I'm saying?
Suits won't work.On top of that we have to take into account suits. It's easy to think something like: "Just rank EHP vs DPS, then match on player skill" But if that ever happens no one will EVER want to run heavy. High damage weapons and solid EHP with no accounting for the drawbacks of the suit.
So. The question I ask y'all is what metrics can you really take into account? How do you spot a stomper from a nova knife heavy? How do you know when a player is having a bad day / running a bad fit / is just bad.
I await the knowledge of the crowd. So did you get what you want?
|
Dimmu Borgir II
Bannana Boat Corp
158
|
Posted - 2013.10.07 23:54:00 -
[9] - Quote
A B Ablabab wrote:I think the guys writing matchmaking are screwed, there's just an ungodly amount of junk to take into account.
SP won't work: Because players can easily spunk SP into useless junk, then have be stuck in a tier that's too high, and we'll have ENDLESS threads of 'I want a respec or I quit', as if we don't have enough of those already.
New players espically wont feel the need to specialise. They want to try new things, and that's only good for the game (Less ARs running around)
Fitting won't workYou can buy and fit suits while in battle, any system which takes suits into account will have people AFKing for the first minute so they can throw together a proto suit and suicide.
How do you balance someone with a proto fit that's BAD. Seriously I know in these forums a lot of people are calling each other out over bad fits, but just look at EVE. It's really easy to fit badly, or worse fit badly for the way you want to play. It's no good putting me up against real proto stompers when I've put all my SP into my brand new proto fit nova knife heavy. Know what I'm saying?
Suits won't work.On top of that we have to take into account suits. It's easy to think something like: "Just rank EHP vs DPS, then match on player skill" But if that ever happens no one will EVER want to run heavy. High damage weapons and solid EHP with no accounting for the drawbacks of the suit.
So. The question I ask y'all is what metrics can you really take into account? How do you spot a stomper from a nova knife heavy? How do you know when a player is having a bad day / running a bad fit / is just bad.
I await the knowledge of the crowd.
War points or skill points are the only measure of how much experience a player has, it doesn't matter that someone has wasted their SP, that's their problem, that person will generally still have as much in-game experience as someone with an equal amount of none wasted SP!
War points are the purest measure of balance and could be averaged out over all of that players characters to show a true median of their experience because that's what we want from matchmaking, matching equally experienced players, not equipment and abilities etc. |
Atiim
Living Like Larry Schwag
66
|
Posted - 2013.10.07 23:55:00 -
[10] - Quote
Match people based on War Points, as I feel that this is the only true test of skill. |
|
Zeylon Rho
Subdreddit Test Alliance Please Ignore
2682
|
Posted - 2013.10.08 00:26:00 -
[11] - Quote
Fitting could be an self-selected means of match-making and be fairly functional. MLT/STD, ADV, and PRO/no-restriction matches. They could also gate the matches by Meta-level.
The newbies could opt for MLT/STD matches, and while vets could do that too - they wouldn't be able to field fits with any gear better than STD on it.
Vets would still have a passive advantage if they played lower-gear tiers, but newbies wouldn't have to eat the combo of passives + PRO gear.
You could have matchmaking within a gear-gated system, and make an internal player rating based on K/D, WP, SP, and W/L. I'd give further weight to pre-existing squads on a team (current matchmaking ignores whether or not players or squaded I've noticed), and you could probably make some sort of ELO score as well. Like if the OP and I had similar base scores, but in 9/10 matches against him I lost, then my rating drops.
Gear-based provides a simple solution that will work 100% of the time for gear-issues. You can't field better than STD/MLT? Done. It doesn't fix other issues, but unlike most solutions it always fixes the gear disparity.
|
Onesimus Tarsus
Kameira Lodge Amarr Empire
657
|
Posted - 2013.10.08 00:37:00 -
[12] - Quote
KDR. |
Vrain Matari
ZionTCD
1048
|
Posted - 2013.10.08 01:04:00 -
[13] - Quote
How does one elicit variables, most likely composite, from a giant pile of data?
In this case, since CCP has all the data and can compute all the conditional probabilities, I'd be tempted to go Bayesian on its ass, using maxent priors.
Come to think of it, E. T. Jaynes would have been a hellacious merc to face. I'd pick him for my side ;) |
lrian Locust
BLACK OP SNIPERS General Tso's Alliance
153
|
Posted - 2013.10.08 01:08:00 -
[14] - Quote
Onesimus Tarsus wrote:KDR. Then you'll get snipers fighting snipers, and logi's fighting logi's. |
Delta 749
Kestrel Reconnaissance
2046
|
Posted - 2013.10.08 01:09:00 -
[15] - Quote
Fittings would work if you do it right, all it takes is barring you from selecting any fitting that goes over X meta level when you spawn Then all your standard, advanced, proto fittings would be left alone on your list and just off limits for that match if you choose whatever tier
Hell the tier system as well is relatively simple in design, just let someone toggle which tier they would like to be placed in like we do with game modes now for instant battles while leaving FW and PC as free for alls and scale after battle rewards accordingly
And if everyone decides they only want to queue up for standard meta level battles then that should tell you something |
lrian Locust
BLACK OP SNIPERS General Tso's Alliance
154
|
Posted - 2013.10.08 01:10:00 -
[16] - Quote
Zeylon Rho wrote:Fitting could be an self-selected means of match-making and be fairly functional. MLT/STD, ADV, and PRO/no-restriction matches. They could also gate the matches by Meta-level.
The newbies could opt for MLT/STD matches, and while vets could do that too - they wouldn't be able to field fits with any gear better than STD on it.
Vets would still have a passive advantage if they played lower-gear tiers, but newbies wouldn't have to eat the combo of passives + PRO gear.
You could have matchmaking within a gear-gated system, and make an internal player rating based on K/D, WP, SP, and W/L. I'd give further weight to pre-existing squads on a team (current matchmaking ignores whether or not players or squaded I've noticed), and you could probably make some sort of ELO score as well. Like if the OP and I had similar base scores, but in 9/10 matches against him I lost, then my rating drops.
Gear-based provides a simple solution that will work 100% of the time for gear-issues. You can't field better than STD/MLT? Done. It doesn't fix other issues, but unlike most solutions it always fixes the gear disparity.
Then you'll get routined players skilling in passive only, who still have a huge advantage over newberries.
|
Vrain Matari
ZionTCD
1049
|
Posted - 2013.10.08 01:17:00 -
[17] - Quote
First question is what is the matchmaker trying to equalize? Win/loss ratio? Isk efficiency? KDR*win/loss? A combination of these?
What is the target the matchmaker should shoot for and how do we measure it? What variable(s) measure a fun and fair match?
Or maybe turn it around - what variables predict a redline? |
Crash Monster
Snipers Anonymous
1580
|
Posted - 2013.10.08 01:26:00 -
[18] - Quote
Nerd checking in. What do you want?
Anyway, ditch the matchmaker and have people self-segregate due to risk/incentive. Might take some work to find a way to keep people from moving up too early as they have an incentive to do so but it will help keep higher players out of lower reward areas. Especially if the only real difference is the caliber of play. |
Crash Monster
Snipers Anonymous
1580
|
Posted - 2013.10.08 01:30:00 -
[19] - Quote
Hmm, on a truly nerdly bent...
Create a Bayesian analysis engine to see the impact that various players have on the outcome and mix and match players until the win/loss impact is reasonably balanced. To get more sophisticated ratchet the results based on the last 10 games to compensate for recent player style or equipment changes.
Bonus points for separating squad vs solo impact analysis. Also bonus points for compensating for the fact that a person with less impact in high SP gameplay is likely to have more impact in a low SP gameplay. CCP, pay me, give me a stats dump, and I'll build it.
I'd like a nerd factor rating please... |
Maximus Stryker
Villore Sec Ops Gallente Federation
626
|
Posted - 2013.10.08 01:39:00 -
[20] - Quote
Atiim wrote:Match people based on War Points, as I feel that this is the only true test of skill. I think you would need to do war points devided by minutes played to get a better rating of player. A player who got 50,000 war points in a week is probably better than the player who got 75,000 war points in a month. |
|
pseudosnipre
DUST University Ivy League
269
|
Posted - 2013.10.08 01:50:00 -
[21] - Quote
Every spawn, recalculate
PLAYERmeta = .99*PLAYERmeta + .01*SUITmeta
use this moving average to balance teams or gr poo up players together in tiered battles.
Also, I like system security limiting suit meta...just throwing it out there |
Garth Mandra
The Southern Legion The Umbra Combine
139
|
Posted - 2013.10.08 02:38:00 -
[22] - Quote
I'm tempted to say something along the lines of Elo rankings.
This system was intended for 1v1 games like chess but it may be extendible (I know League of Legends uses it).
Quote:After every game, the winning player takes points from the losing one. The total number of points gained or lost after a game is determined by the difference between the ratings of the winner and loser.
This could be used to set up teams with the same total ranking (with some modifier for premade squads).
Elo ranking would take a while to put into effect since initially everyone has the same rank. We'd have to wait for the good players to pwn some n00bs before they went up in rank and started playing each other.
A problem with this system is the fact that a lot of noise would be generated by the 15 other guys on your team. So perhaps this system in combination with another may work.
Another important requirement for this to work is to make winning matter. If people aren't trying their best to win then the ranking system would suffer. |
knight of 6
SVER True Blood Public Disorder.
346
|
Posted - 2013.10.08 02:41:00 -
[23] - Quote
Maximus Stryker wrote:Atiim wrote:Match people based on War Points, as I feel that this is the only true test of skill. I think you would need to do war points devided by minutes played to get a better rating of player. A player who got 50,000 war points in a week is probably better than the player who got 75,000 war points in a month. I'd say on a 5 match average or something. reason being if you are AV specced and nobody is vehicle specced than you don't work to your level. also logis earn massive wp they would need to be balanced so they are put on all levels of play perhaps not evenly but you know what I mean. |
lrian Locust
BLACK OP SNIPERS General Tso's Alliance
155
|
Posted - 2013.10.08 02:49:00 -
[24] - Quote
Crash Monster wrote:Create a Bayesian analysis engine to see the impact that various players have on the outcome and mix and match players until the win/loss impact is reasonably balanced. To get more sophisticated ratchet the results based on the last 10 games to compensate for recent player style or equipment changes.
Then how can you measure the impact that people have on a game? A pilot might have a huge impact, although you'll hardly see him in the hacking and killing stats. And how about that speed-tanked scout who sacrificed himself at the start of a Domination game, but gave his team a huge advantage by capturing the objective?
I like your idea, but perhaps we're making it too difficult than it needs to be. |
lrian Locust
BLACK OP SNIPERS General Tso's Alliance
155
|
Posted - 2013.10.08 03:10:00 -
[25] - Quote
How about the statistic "average total meta level per spawn"? It works like this: 1) add the meta levels of all equipment, dropsuit and weapons per spawn, 2) add all the numbers of all the fittings together, 3) divide it by the amount of spawns. Here's your average. Calculate it over the last 30 days, and/or or 20 games, or whatever, and off Scotty goes!
This has the following advantages:
- higher level fittings will be matched with opponents with higher average meta levels - a pubstomping squad's team will be filled with newberries to level out the averages, levelling the playing field - Newberries will therefore still be able to earn ISK and SP - AFK'ers use low-quality suits and will count as less of a force in a game - increase and decrease in difficulty will be gradually - the playing field will be leveled when it comes to gear and experience, but skill can still make the difference - it works with any amount of players - it's easy to calculate - also works for new players entering during a game - a players value can be calculated prior to matchmaking, so the matchmaking process is fast
To avoid a scenario where vets with maxed passive skills and militia gear will still outperform newberries with similar gear, PPC can choose to add additional meta points for completed passive skills (e.g: a level 3 electronics skill will add 3 meta level points).
Appreciate any feedback!
|
Virtual Riot
The Vanguardians
48
|
Posted - 2013.10.08 03:13:00 -
[26] - Quote
Its not even hard. Just do it how LoL does it for their normals.
Match players based on their number of wins. You are only matched against players who have a similar number of wins as you.
Boom, fixed. Like seriously, its that fkin easy I dont know why this has not already been done. |
lrian Locust
BLACK OP SNIPERS General Tso's Alliance
155
|
Posted - 2013.10.08 03:17:00 -
[27] - Quote
Atiim wrote:Match people based on War Points, as I feel that this is the only true test of skill. That really depends on the role you're playing. A sniper will get less WP, although he might pin the enemy down and soften them up for the guys on the ground. On the other hand, logi's can rake in the WP and might have to fight mercs that are relatively much stronger than they are. They'll have to fight an uphill battle and lose on ISK due to the high cost of their fittings.
|
lrian Locust
BLACK OP SNIPERS General Tso's Alliance
155
|
Posted - 2013.10.08 03:20:00 -
[28] - Quote
knight of 6 wrote:I'd say on a 5 match average or something. reason being if you are AV specced and nobody is vehicle specced than you don't work to your level. I agree, it has to be an average, or the skill levels will be all over the place. There need to be more games though, and in a limited period (say, 20 games in max. 30 days?).
|
|
|
|
Pages: [1] :: one page |
First page | Previous page | Next page | Last page |