I've mostly played the All Welcome teams lobby, but I decided to try 1v1 matchmaking and wow it is awful. After a 10min wait I was matched with a player 500 ELO above me and looking at post game it showed I had a 2.8% chance to win! How is that "similar" skill? I think I'll just play this game in PvE and forget PvP.
+1 / -0
|
Playing against better players, especially in 1v1 is a great way to learn to play better. At the end of the game, always ask for tips or advice. The matchmaking community is relatively small, so it has to sometimes stretch the allowed elo difference to find a game at all. Just keep playing and you will improve! Also keep in mind that matchmaking and casual elo are different, so that may make the elo difference appear larger depending on which type of elo you are looking at. (Also it's technically called WHR now instead of ELO but whatever... I still call it elo too)
+0 / -0
|
I can watch other matches to learn. And by stretch you clearly mean break - 2.8% win chance isn't just a stretch, unless your talking some new neologism.
+0 / -0
|
MM is "broken" because not enough of the population uses MM and instead piles into the all welcome room. When 5-6 people out of 130 some odd people are playing 1v1 MM and they tend to all be good, it drives off the median players resulting in an increase in activation energy. There are means of solving this though and requires experimentation. Simply put: people won't play MM because they fear getting bigger opponents so those who do end up joining MM face the bigger opponents, despite the median population representing a massive majority.
+1 / -0
|
Another problem is the matchmaker is pretty aggressive when it's your first game and you don't have a rating, this is for calibration purpose but it will result in your first game or two being possibly pretty hopeless. Once your rating is known and stabilizes, the matchmaker will not send you into quite so lopsided games. This isn't ideal and it would be nicer if you could pick your starting level instead or the matchmaker used casual rating for the initial games, but that obviously requires work. The matchmaker will not propose battles where you are known to have less than a roughly 10% win chance, and will take a few minutes to reach this level. It's an unfortunate trade off between fair games and actually getting a game given the low population size, and the lobby doesn't make this trade off apparent at all. What you can do is cancel queuing and re-queue after 2-3 minutes to prevent the matchmaker from pairing you up against much stronger or weaker opponents, but letting you set a maximum rating deviation explicitly would be nicer. There's definitely room to make the matchmaker behaviors explicit, maybe just informing people the first couple games are calibration games would help (Starcraft 2 has the same problem when newcomers enter their ladder), or show the maximum allowed WHR variation given how long you've been queuing for but there's no super clear way to do that ("waiting for 4:23, matching 20-80% win chance" maybe).
+2 / -0
|
The first few games will be calibration games since you have no matchmaker rating. The matchmaker does not say this because there is no mathematical distinction between calibration/normal, as the uncertainty about your rating will just tend to go down as you play. Putting an, essentially, fake "calibration game" sticker over your first five games would do a lot for communicating how the system works and making it more fun to interact with. In general, the MM bounds are being pulled in both directions by people on one side who just want games, regardless of how even they are, and people who only want a game if there is a decent chance of it being good. I've tended to argue on the side of lowering the match range because I think if people trust the MM to make decent games they are more likely to stick to it. I'll lower the matching width slightly.
+3 / -0
|
    GoogleFrogI've always advocated for players to be able to customize their matching width. If they want closer matches, they should be able to narrow the width. If they want faster matchmaking, they should be able to widen it. As opposed to many games where the width automatically widens over time, if you don't find a match. This should be a mutual thing, of course - a match must satisfy both players' width settings. The player should be able to set a default, and when they click to find a match they should have buttons to widen / narrow the matching width for that particular search. Bonus points if the system is able to give an estimate on how long it might take to find a match at a particular width, although maybe Zero-K isn't populated enough to meaningfully calculate that.
+1 / -0
|
Making the matching width configurable makes sense (and I'm willing to work on it) since the behavior already exists and can be enforced if you have sufficient knowledge of the matchmaking algorithm and are willing to cancel your queue after X seconds. Being able to set a "difficulty" level in the queue should encourage people from queuing up (I don't mind getting the uphill climb games, but don't always want to play such games, so don't queue much). The problem I see with customizable matching width is you need to go down to +/-100 to have a game no worse than 66/33%, there's fairly high rating uncertainty still due to low game numbers, and this is a very small number relative to the user population. So if too many people set low bounds, there's a chance matchmaking would fail to ever resolve, so there should probably a mechanism to "nudge" people to increase their matching width if they're missing out on games. Or just be conservative and only allow a fairly high minimum (say 250/300).
+3 / -0
|
Sounds fine if you can figure out how to implement the infrastructure side and design the lobby UI. A minimum gap of 250 should be fine, as a 20% win chance is still pretty uncertain. See https://www.walkofmind.com/programming/chess/elo.htm
+0 / -0
|
Update: I played several more MM games and it still seems not to work well, although, it's generally me who receives the benefit against weaker players. I'd say about 40% of the games are reasonably fair. In fact, I played one player that I beat 2x and who beat me 1x so that's a pretty fair match-up.
+0 / -0
|
The population size isn't anywhere near big enough to ensure that any game will be fair, but adding an option to configure a matching range to search for will help increase the number of even-ish games (at the cost of reducing the total number of games that can be played. This trade is probably worth it for you, and I think it's worth implementing in the game to at least give users some control over their matchmaking experience). In https://zero-k.info/Battles/Detail/924583, you got a match as far away from you as the matchmaker currently allows, and that's almost certainly going too far into the "get a game at any cost" direction. Even with a large user population and configurable matching width, I think the nature of Zero-K will result in relatively swingy games even with two equally skilled players. The game lacks a strict execution barrier and has a wide range of maps and factions, so some games will be won based on strategy (e.g. picking a better build on a map than the opponent) or luck (one player blunders basically). The win chance is pretty accurate for a series of games with the same opponent once your rating is stable, but the outcome of any individual game will be fairly random. So while lowering the maximum allowed deviation will help reduce the number of poor games, they can't be avoided fully. (This is compared to a game like Starcraft 2 where the APM barrier more or less guarantees you're not winning if your mechanics aren't as good as your opponent, other than by a lucky cheese they fail to defend. It is much simpler to propose an even game where a large part of the difficulty is in fighting the game UI itself and the elo rating can build on that. APM really doesn't translate as well into a victory in Zero-K, and the variance in performance of most players across games is pretty high.)
+2 / -0
|
I generally am against more complicate setting both because it is a nightmare to do UI and I would assume that it would be harder to maintain and easier to break the core problem is the low population pool and I fear the bandaid could be worse than the current system.
+1 / -0
|
Increasing the user base is the most desirable outcome but that's obviously the hardest one, so is there any other action we can take to increase the matchmaking population? The MM ladder currently has 271 users versus 1490 in the casual ladder, which is pretty good percentage wise already (a few players are only present in the MM ladder and not casual, but that's not a big percentage). There's been about 1000 1v1 games in the last 30 days (plus a handful of team MM games), so it may be possible to encourage existing players to play more to have a healthier MM pool, and improve retention for MM users.
+1 / -0
|
My mm rating is very low and I'm being matched with greys.
+0 / -0
|
