test content
Logo
What is the Arc Client?
Install Arc
Options
thinking about matchmaking filter
ursusmorologus
Member Posts: 5,328 Arc User
I hear a lot of people asking for matchmaking, which I assume means auto-balancing the teams in a queued match. What criteria are you looking for? Seems like the only thing that makes sense is something like kill-death ratio for the toons, possibly weighted by the K/D ratio of the ship, with high-low distribution so that the sum K/D for each team is about even. Maybe you have a guy with 1.6 and another guy with 1.3 then everybody else is 1.0 with a couple of .3 noobs at the bottom, mix it all together so you end up with ~1.0 on each team. Just thinking about this, nothing more.
[img]http://i183.photobucket.com/albums/x57/UrsusMorologus/MechWarrior Online/Catapult_K2 sig.jpg[/img]
Post edited by Unknown User on
0
Comments
The way it stands, even if they fixed the teams queuing against pugs bug, there's still no mechanism other than your place in line to determine the makeup of the 2 teams. People that want to troll the queue can still easily exploit that fact. As I see it, without some form of matchmaking the chance of balanced games is not any better than it was before. I'm sure that Cryptic could come up with something that functions much better than Hilbert's balancing tool (and it's pretty good). They have access to so much more information than us players do.
Anyway, yeah. Matchmaking.
I'm kinda over ever hoping for any semblance of balance to these matches thought. Geko described MMORPG balance as "the dragon that everyone chases but you know, some imbalance is there by design and needs to be there. Imbalance adds a fun factor."
The abundance of universal consoles and the imbalance they create in relation to pvp is out of control and there's no stopping them in sight. Universal consoles only being able to be used in device slots I think would help? Who knows.
It adds a $ factor for them no doubt.
These devs are pretty clueless so I have no hope for there game anymore either. It is funny to see devs say something like that though... its pretty much an admission that things are way over the line. We all know that yes developers of games that are out to make money factor in a small amount of inbalance.... to come out and say it though means they understand they have went a little to far.
All well... plenty of other good games around that are closer to an ideal amount of unbalance. Worse case I'll just spend more time playing chess or something. lol
Better gear is a certain degree of imbalance too.
Assuming that's what he's talking about, in context.
Holy... :eek:
Man that's some really impressive work, congrats!
Tried it (even if it put three out of three Scimitars, a healer and a vaper together) and it might need a bit of refining, but I would look forward to have something like that implemented.
still, it gave me quite balanced results in a few attempts.
Currently I attempt to account for changing builds by limiting the number of matches that are used to calculate the healing/damage/etc. averages for the balancing algorithm; this way only somewhat current matches are being used instead of the entire player history which might include many different builds.
[/Random Thought]
If you look at the table on the TrueSkill page, it seems that estimating a player's strength gets a lot harder when one has two teams with 5 players than if one had a free-for-all.
Unfortunately this is not my area of expertise, but if anyone else has experience with such rating systems, I'd be happy to provide the win/loss/healing/damage/etc data that I have collected if it could help create a better matchmaking algorithm.
(Also note how all these systems seem to be based on Win/Loss whereas my system currently completely ignores this.)
Please re-read what I fixed. IF you trot it out enough eventually it may happen, eventually.
I really haven't gone into depth into this, but my impression is, that these systems work for teams, but it takes longer for the values to stabilize. It may not work if you are always in the same team, then it has no real chance to differentiate your skill levels - but once you leave the team, your score will likely change (probably first the confidence - because it expected you to win or lose, but you didn't, and the first times this happens, it will lower the confidence, as it happens repeatedly, confidence will grow again.)
I think 5 vs 5 man teams are probably not as bad as something like 12 vs 12 man teams in Mechwarrior Online.
No system is perfect. But you don't need perfect. (Perfection will automatically fail the moment you don't get the right players in the queue - if there are 3 excellent players, and 5 mediocre, then it can't help but create an imbalanced match, for example, even if the scores have been calculated perfectly.)
I think most systems assume that it's better to abstract away from specifics. Elo rankings were introduced for Chess - they could have tried to include statstics like "How often did he put the enemy into Chess" "how many points worth in figures where still on the board at the end, how often did he use a weak figure to capture a stronger, or what not" - but in the end, the real result that matters is the outcome, not how you got there.
For teaching you to get better at the game, you need the other statistics. If you realize that your healer heals only 50 % of what the top rated player who usually uses healer builds has, then you know that's a weakness. On the other hand, if you see that you have less DPS then him or that you die more often, you know that's what you need to improve. But the rating systems aren't aimed at improving yourself, they are aimed at creating "fair" matches or predict the likely outcomes of matches.
I can see win/loss working fairly well for 1v1 games and for all games where each player is "interchangeable" (e.g. Unreal Tournament).
I wonder: FPS have roles (medic/sniper/...) and certainly there is a balancing algorithm for FPS multiplayer matches. But maybe they skip the role part in the balancing process because people can swap roles after every respawn (maybe within certain role limits for the team)? This way players would be interchangeable again (which is not the case for STO). I haven't really played FPS in well over a decade, so I don't know how it works these days or if players are locked into a role for a long time.
They probably do it with more modern ones, as they have been aggressively tracking stats since the black ops 1 era, but prior to that there was no auto-balancing that I am aware of. Players in my old COD4 clan used to switch sides manually in the middle of the game to balance it out if it looked too lop-sided. Usually it was 10v10 matches so there was less steamrolling, but it still happened often enough that we had a casual protocol to deal with it.
The classes were very neutral back then too. You could spawn with a hmg and pickup a lmg on the ground, no difference. Classes didnt have different armor or anything, just different kits or camo if anything.
So if your own match-maker yields better results in predicting the outcomes of battles and creating "even" battles then an Elo based calculation, then you probably have sufficient evidence for your point.
But my theory is that roles are not that important. If anything, then the only thing important is that all roles are covered in a team, but beyond that, win/loss based Elo ratings will suffice.
Did we have those? I don't remember anymore. I only remember trying to gather as much of the cryptic information about the game mechanics as I could, thankfully having other people doing the dirty work for me.