Upcoming matchmaking system
dePARA
Join Date: 2011-04-29 Member: 96321Members, Squad Five Blue
Join Date: 2011-04-29 Member: 96321Members, Squad Five Blue
I have a question @UWE
How is the upcoming matchmaking affect the current community based server structure?
Are these matchmaking servers hosted by UWE?
There many admins like me who spend a lot of money for hosting servers.
Im doing this since 2,5 years now for example.
If players only use the matchmaking system this would be a slap in the face for every admin out there.
So how is UWE going to handle this?
And why should community admins continue to spend money for empty servers?
How is the upcoming matchmaking affect the current community based server structure?
Are these matchmaking servers hosted by UWE?
There many admins like me who spend a lot of money for hosting servers.
Im doing this since 2,5 years now for example.
If players only use the matchmaking system this would be a slap in the face for every admin out there.
So how is UWE going to handle this?
And why should community admins continue to spend money for empty servers?
Comments
Matchmaking is an abomination that doesn't work unless you have a HUGE pool of players to draw from (like consoles do, and even then it has it's problems) - With the extremely small community that NS2 has there is virtually zero chance of matchmaking working effectively. - They are wasting their time.
I say give it a week after it's released, and those who try it will gladly be flocking back to the server browser and rented servers.
If they actually try to force us to use this garbage (by removing the browser) NS2 will be completely dead within days.
Err, i doubt its the sort of matchmaking your thinking of. I think Its more of a Gather system like the one on the ensl website, just built into the game. Enabling everyone to join into a organised 6v6 game - The way ns2 is meant to be played.
Defiantly not a waste of time and its actually one of the most anticipated features to come to ns2, It is the best thing they can do to bridge the gap between competitive and public.
So, back to OP's point, it would be interesting to hear the ins-n-outs of the way its going to work!
it's too much hassle to balance the teams manually, and some people forcably pick the stronger team because they'd rather caress their ego than have a good, fair and challenging game.
Out of like 10 recent messed up ELO games I played only 1 was not grossly imbalanced.
A good Elo system needs to track at least 60% of all games, sadly ns2stats e.g only covers like 20% atm.
ELO is the best reasonably implementable (ie not omnipotent) matchmaking system that I've seen. Your problem isn't with ELO, but with your expectations from a simple matchmaking system.
Like any matchmaking, ELO requires a huge playerbase to be effective.
Lastly, whilst I appreciate the game is designed to play 6v6 for the ultimate test in skill and tension. I very much enjoy my pubs of 8v8 or 9v9. God forbid the 20 players servers are fun too. So, please don't take anything away. Just give us more choice.
Sorry but I can just /facepalm.
Using an unmodified version of the ELO formula that was designed for chess, which is a 1vs1 game and also has a chance of draw, for something like NS2 is ludicrous.
Just basing the rating on round outcomes alone is a very bad choice. I can suck at the game and sit there almost afk and win a game, or fight to the finish and lose a game. Also, I'm not sure giving newbies a default rating of 1500 is sane.
Huge playerbase wouldn't change anything. You don't have access to the playerbase when setting up the round - again, this is not chess nor is there a matchmaking system .. yet. You only have access to a couple of randoms plus maybe some regulars on a given server. Even if you had, it would still suck because it takes quite a few played rounds for the rating to converge on your win:loss ratio (yes, win:loss, not skill). But with a very low chance of there being a fair, balanced round the algorithm probably messes ratings more up than you can imagine.
The only way I can see ELO "working" is due to random chance or a high number of skilled regulars on a server. So tbh I'd rather see a proper random function...
And what would else would you weight ratings on aside from simple win/loss? Any other weighting, as per kills or damage done or healing done or any of these per minute are going to introduce heavy biases. They also simply aren't comparable over varying game sizes (ie getting 50 kills in a 12v12 is a lot easier than 50 kills in a 6v6 in the same timeframe) and aren't comparable over varying game lengths (ie getting 5000 player damage per minute is a lot easier when you spent 30 of the 40 minutes shooting onos when compared to just shooting skulks).
In fact, I think weighted team-based ELO (as I call it) is the best matchmaking system. It works by taking your team's average skill (Ts) and the opposing team's average skill (Os). Then it predicts a winner/loser based on the difference in average team skills. If Ts > Os and Ts wins -> smaller point gain for Ts players and smaller point loss for Os players. That's simple team based ELO. The weighted part can be used to allocate points or minimize gains by using the variance of the respective teams and comparing them with the Player's skill. This can be used to prevent high ranked players from gaining any points by beating teams far weaker than them, even if the average team skills are somewhat even (ie stops pros from pubstomping rookie servers and constantly accruing skill.... as is a problem in our current hive's stats ranking system). I could pound out specific formulas and numbers if so inclined, but I'm really not interested in mathing out something which nobody will use.
Weighted team-based ELO is relatively simple and it's not arbitrary. That means that it doesn't care about game sizes or player roles. If you think player roles should be accounted for, then please explain how you will normalize whatever metric you determine (kdr, pdmg, sdmg, healing, score, etc) such that it does not vary across the player count and game time spectrum. Anything you come up with is going to be extremely subjective and only an effective indicator of skill in a specific circumstance. You can try to best-fit it, but it will still be arbitrary.
Also, apparently it is "Elo" because it is named after Arpad Elo. I don't feel like editing the 15 uses in this post.
Huge playerbases are pretty much the one and only requirement in any matchmaking system. Mediocore systems will still work given a large enough pool of players to pick from. If you ever notice matchmaking systems people talk about in a good way..... they're all games with huge player bases.
If you're talking about splitting teams on a single server then I think that's a rather impossible task. Systems like these just don't work with such a small pool of players. Shine administration has a bunch of random-based options and none of them are much better than just allowing players to stack whatever they want. I'd think altruistic captains picking teams is probably the best matchmaking system for a small group of players, but that's not very likely to happen. And even with altruistic captains picking teams (ie reddit pugs, map testing games, or the few times I've seen this in the wild), you still wind up with frequent one sided victories.
Also isn't the starting value 1200? Or is that just in LoL?
It doesn't really have to be a net zero some thing either. You could start it at 0 and never go negative if you wanted. I don't think that would still be a true Elo rating, but you can apply the same principles of probability and rating gain/loss.
At any given time there's really only 200 or so people playing in my area, and how many want to play comp 6v6? Is matchmaking ONLY comp 6v6? Will individual skill matter at all in pubs?
Honestly I don't really think the game needs this right now, and with the current playerbase I don't really see it working out. But I'd love to be proven wrong.
I would take a guess and say since it uses the hive that the match making system will use the same servers that get white listed for hive. Servers get whitelisted after, I believe, 50 matches with out any mods.
Seems like a waste of time to me. Is anyone going to seriously sit in a lobby and wait? The time to get into a game is already high, how does match making shorten it?
Would be better and more productive to refine the skill rating server browser column, and get quick join to take the skill ratings into account.
But again @hugh we would love to hear some juicy gossip on your views of what's important and what's achievable. This is the only thing keeping me interested in long term ns2 to be brutally honest
Some info about it and exactly what is changing would really be nice.
I don't see 20+ players suddenly wanting to play 6v6, and I never notice many populated 12 player servers, but they might all be passworded. edit: If you look at a selection of players on http://hive.naturalselection2.com/ their game history shows that they like to play on only a handful of servers. Typical MM will stop that happening.
I would like to see just a more advanced "vote random teams" but that takes into account skill/winrate and possibly locks players who want to stay together + gives people a favoured team choice.
it's a bit of luck that sponitor covers 100% of games then, isn't it.
UWE has the tools to make this work, it doesn't have to be done through a 3rd party.