Making sure there is always an open slot on a viable game of your skill level (and saving the game)
moultano
Creator of ns_shiva. Join Date: 2002-12-14 Member: 10806Members, NS1 Playtester, Contributor, Constellation, NS2 Playtester, Squad Five Blue, Reinforced - Shadow, WC 2013 - Gold, NS2 Community Developer, Pistachionauts
Creator of ns_shiva. Join Date: 2002-12-14 Member: 10806Members, NS1 Playtester, Contributor, Constellation, NS2 Playtester, Squad Five Blue, Reinforced - Shadow, WC 2013 - Gold, NS2 Community Developer, Pistachionauts
I've been thinking a lot about what I'd most like the new dev team to work on to retain and bring in new players, and this is what I've come up with.
1) Officially support high player count servers. (at least up to 32 players)
2) At map change, if the number of players is above 28, split the server into two groups by skill and send half of the players to a new server.
This ensures that if there are enough players for viable games, there are as many viable games as possible, and there is always an open slot on a viable game. This will also converge to making games of different average skill so that as many games as possible are fun.
(This system would of course be optional for the server owner, but I suspect servers that join the pool will get more use than servers that don't.)
Games lose players due to friction. The biggest source of friction is being unable to find a good game quickly. We can't do matchmaking with the number of player we have, and we can barely seed servers. This fixes both.
1) Officially support high player count servers. (at least up to 32 players)
2) At map change, if the number of players is above 28, split the server into two groups by skill and send half of the players to a new server.
This ensures that if there are enough players for viable games, there are as many viable games as possible, and there is always an open slot on a viable game. This will also converge to making games of different average skill so that as many games as possible are fun.
(This system would of course be optional for the server owner, but I suspect servers that join the pool will get more use than servers that don't.)
Games lose players due to friction. The biggest source of friction is being unable to find a good game quickly. We can't do matchmaking with the number of player we have, and we can barely seed servers. This fixes both.
Comments
By "server" do you mean the actual physical hardware? Or the players and map? You'd be retaining both of those.
Suggestive and informative methods however, are proven.
You want players dynamically mixed with similar skill levels, SUGGEST it to them and INFORM them why that could be a good idea.
You want servers seeded more easily. Add something fun to do while waiting.
Or make this a reality.
Servers fill proportionally to the number of online players, it's not the system that needs fixing.
If it were all behind matchmaking and invisible to the user that'd be one thing, but giving them a different server than the one they clicked on creates anger imo
Good on ya for thinking outside the box tho
It seems like that poisoned this idea with a bad implementation. I hated it too. The problem was that it would dump you into an empty server immediately.
This version would only happen at map change so you would always get to play to the end of the round in the server you clicked on. It would only split the next match if there are enough players for 2 good games. It would essentially be exactly like half of the players that aren't of your skill level leaving the server, leaving you with a small evenly matched game. The player wouldn't notice any difference.
This could be a great use of servers whose behavioral expectations were consistent (for me, that's synonymous with "unmoderated", but maybe someone else has another route to consistency). Involving community servers, even with the up-front coordination and blessing of their owners, introduces variable behavioral enforcement and might create a scenario wherein players are suddenly being held to new behavioral standards they didn't anticipate when they actively chose their original/entry server at the beginning of their play session.
/ducks
The first one, the wild-west pub servers should be kept and protected. They are a place to play with and against random people with diverse skill-levels. A place for friends to meet and play. A place to try out fun and/or important new mods (hello @Mendasp !).
If you try to force upon these places some kind of "Big-brother knows best" cutting and splitting, you will get a lot of 'I want to play with friend(s) x,y, WTF!' and 'I dont mind that I suck and they don't, I still enjoy it!.
That being said, a ranked play mode, where you can be matched up with and against random players of roughly your own skill-level would be something that (especially new) players massively profit from.
In order to be accepted and successful, I think a ranked play mode would have to meet the following goals:
I am sure I will get a lot of 'the playerbase isn't big enough for that kind of matchmaking' type of feedback.
I just don't buy it. If something like ENSL gathers keep working with a small player base, then basic matchmaking should be possible if you try hard enough.
The issue with matchmaking is that you have to have something to do while you wait that's better than playing something else, or jumping into a random server, especially if the wait time is really long like it would be with ns2. The goal of this is to solve that by letting you play ns2 while you wait (albeit in a larger than normal game.)
I expect lots of the main servers would turn it off, and that's fine.
I'll clarify how I envision it:
1. You start NS2
2. You click Ranked Match, Quick Play or whatever button it will be
3. (optional) the matchmaking system informs you about expected wait time
4. You join a pub server, or navigate to whatever other part of the game to kill the wait time
5. During all that you stay signed up for the ranked match / You are able to unregister from it at any time prior to a found match (see gather functionality)
6. Once a match is found, you get a notification, get an option to switch to the match's server
Both nosgoth and ns2 are asymmetrical games with melee vs ranged factions. Nosgoth has a similar skill curve, similar skill gaps, and a better learning curve. When I lasted played it was in july, where there was an average of 773 people playing daily. Ns2 currently has about ~200 playing daily. Nosgoth had/has about 4x as many players that ns2 has had in recent times.
Matchmaking made me quit that game. In that game I think I was a bit below average skill, which means I should have the more potential players to be match made with because of similar skill. At peak play times I would sit in a queue for 15-30 minutes waiting to be put into a lobby. Then I would wait another 15-30 minutes in that lobby until it filled up with all 8 people for 15 minutes of gameplay.
You would think if the game spent that much time trying to find a better match, it might have really balanced teams. After waiting so long for a game I still would end up with horribly balanced teams most games. Nosgoth is more of a deathmatch where the first team to 30 kills wins. Most games would end up 30/10 which is a huge stomp. By most games I mean 3/4 in my experience.
Ns2 has a smaller playerbase and a much harder learning curve. I honestly do not think traditional matchmaking would work for ns2, unless it was designed around it from the beginning. Even then, nosgoth shows that is not a sure thing.
I am extremely happy with the server browser system we have now.
The reasons gathers work in ns2 even with such a small playerbase is because it is an emergent matchmaking. Only certain kinds of players seek out ns2 gathers. Typically they are high skill, but not all. Even if they are not high skill they are without a doubt someone who is willing to learn. If they are not, they wont do gathers for long. With gathers there is this sort of self selection matchmaking. On top of that, gathers often employ a captains like team picking which is a second layer of matchmaking. Thirdly, you can play random games while you wait for a gather.
Maybe, just maybe the old "organized play" tab on the main menu would work well for matchmaking if it worked how you said it would. This would not be traditional matchmaking. Maybe the faster development and willingness to experiment might try this.
@CmdrKeen , are these two points what you imagine for matchmaking?
Still, what I personally dislike about it is the varying 16-32 player max of those servers after / before split. In my experience, <20 players servers tend to die quickly and >24 player servers make NS2 a much less strategic game (which I personally dislike, but many (especially casual) players seem to enjoy).
And I think systems like this need to focus on the benefit of the casual player more than the comp players. As @Nordic said, the comp players have the gathers as a place of high skill.
Your method would accomplish a basic splitting into an upper and a lower skill-group. And I think that may really be all that needs to be happening. If you seperate players into 3 groups (low, average and high skill), then I think it is fine for average players to play with both groups, and it should be avoided that the lower and higher skill groups encounter each other.
One of my points, the one concerning the quality of servers, still stands. If players get the impression that they are being switched to a much worse performing server, they will likely reject the game mode altogether.
@Nordic : I have no idea how exactly the matchmaking in Nosgoth was implemented. As I understand, matchmaking is always doing a trade-off between waiting time vs. quality of matches. I am pretty convinced that the matching-quality wouldn't have to be extraordinary high in NS2. As written above, I believe it to suffice when low-skill players do not encounter the high skill. To clarify: Say you want to match a 10v10 game, then you wait for 40 players to register, and then fill two servers with them, splitting the skill-ordered list of players in two.
To do it like I imagined it (some kind of asynchronous overlay/menu) or like @moultano thinks, I am not sure what is best!
I think that may be too much in the beginning. This would make it even harder to find a match! Could be considered upon success though.
I'm all for boosting the competetive side, but this should not be at the cost of quality in pubs.
How would that work with community servers?
In Janus VR you type the name of the website you want to go and it opens up the portal infront of you, but there are already set portals in place, it's sort of like that half-life mod which allows you to play any hl1 mod map inside of it.
Maybe we could have portals that lead into a server? i wonder what sort of damage the servers would take on there part for hosting something like this rofl. Actually i revise this, maybe a global server could be hosting a readyroom and updating a global state to each of the local servers vice versa every few ticks to minimise impact.
sounds like a huge undertaking...
You preload all the servers readyrooms in there current state on joining, if they have changed since you connected, it just loads them when you enter the portal to make it look seemless.
A pug would require 1 host who acts as kind of an admin/facilitator. The host does the following:
- Initiates the lobby.
- Chooses what player count it will be (6 v 6, 8 v 8 etc.)
- Chooses where the server will be hosted.
- Can choose a minimum or maximum hours played number
- Can join the lobby as a player
Once the pug/lobby has been hosted, players can join. The players can:
- Vote for a map
- Be skill matched based on some skill system or manual skill matching with the help of the host (who should be able to see hours played).
Once the pug is full to the required player count, the game starts.
The host of the pug should obviously be someone reliable and able to make changes necessary to ensure a good game is had. i.e. they can't
just be a random person who wants to initiate a pug but rather someone known by the community to be able to fulfill the role.
It might be easy to say, "Well just go over to ensl.org and do a gather", but honestly I can imagine that is much less accessible for some
people who would much rather engage with something built into the main client. Also, I think that knowing where the server will be hosted
before joining the lobby is good.
How much growth? The example I used, Nosgoth, has 4x the playerbase we have now yet that is not enough.
We don't know if it will be possible to revive the playerbase, and many people claim that the game is just a "niche shooter" with a high learning curve. Unfortunately that hypothesis was never really able to be tested properly due to all the other barriers to enjoyment that were present upon release.
So perhaps for now it is better to not be too ambitious about the whole match making thing, but it is something that I would want for NS2 future if we ever get to a player count where it would work. As for how much growth is needed, i really have no idea.
I want to repeat (since nobody countered my hypothesis with technical arguments yet).
I really don't see why it couldn't work. You need to let 40 ppl register and then GOGO. What is the problem?
Just because Nosgoth didn't manage to find a solution, doesn't mean there is none. Maybe they implemented their match-making with very harsh matching-criteria (i.e. (translated to NS2) 'Matched players must be within 200 skill range' OR '... within 100 ping range' etcetc.
If matchmaking's only mission (at first) would be to split 40 people into two groups, without any further constraints, then I think that would very well be possible.
Here is a post from the developers of Nosgoth that is semi technical. I don't know for a fact, but it was like nosgoths matchmaking had very harsh criteria and eventually gave up and just made incredibly imbalanced teams.
Your scenario is to let 40 people register in some sort of gather like system and then split them into 2 groups. Is that correct?
I do not think that such a system would be much different than public ns2. I think it would only be a slight improvement. This improvement would entirely come from having a larger sample of players to sort from. Sorting 40 players into 4 teams of 10 is more likely to produce near skilled teams than as it is now with 20 player server sorting 2 groups of 10. I agree this is totally possible assuming players actively use such a system. The more players who use it the better it would work.
Let me try describe what the I think the problem with that is.
The first problem is skill variance. The whole point of a matchmaking system is to reduce skill variance. This means you want as near skilled of teams as possible. We can only assume that random players from all across the skill spectrum will try to use this system. This means you could have any mix of players from a hive skill of 0 to 4000.
Ns2 has such a wide skill gap because it has those high skill ceilings most of us enjoy. A 3000 skilled player outclasses a 2500 skilled player, and just dominates a 2000 skilled player. It is the same with a 1500 skilled player, a 1000 skilled player, and a 500 skilled player. This is assuming relatively accurate hive skill values. This creates a lot of the problems public ns2 has. Not as skilled players get annoying with higher skilled players just being able to dominate, and higher skilled players getting. Even a difference of 200 hive skill is noticeable, or at least I think it is.
In your last post you said being within a 200 skill range would be a harsh criteria. That is half of my point here. The point of matchmaking is to get as near of skilled teams as possible. I think the ideal criteria would be within 200-300 hive skill. As you said that would be harsh, because it would be hard to get a mix of players like that. If you have a more relaxed criteria like 500, it might still be hard to find that mix of players within 40 people. A more relaxed criteria like within 500 would still have some severe imbalances. That is because as I described, a difference in 500 hive skill is a lot of skill in play.
I don't think the system you are describing would have any criteria such as being within 500 hive skill. It would just find the best teams possible with the 40 people. As I said before, that is more players to sort from which increases the chances of a near skilled game. As I also described, I think it would even be hard to find teams that have a low enough skill variance to have near skilled games. The system you described would just be a better shuffle system similar to what we already have. It would be an improvement but it would not really fix anything.
40 players is not enough to find near skilled players. I use nosgoth as an example to show that ~800 players is not enough to find near skilled teams. Kartoshka said matchmaking could work in ns2 if there was some growth. I agree, but I think it needs to be a lot of growth.
For matchmaking to work I think NS would need at least 800 average daily players and a non traditional matchmaking system similar to how you described. Even then 800 daily players might not be no where near enough.
I can envision "What is love", "Rick roll" and "Oooooo you touch my tralala" being blared out upon joining unless voice chat is disabled but then you have another problem, endless spamming walls of text.
They can be nice if they enabled local voice chat only so the entire server doesn't feel the wrath of ding ding dongs.
Which mods, if any? And which settings for those mods?
Which performance score is acceptable? Which ping? How many interrupt warnings should we allow?
And yes, I am ignoring custom rates on purpose. Lets not spend pages of talk on that. Lets just say that it differs per server (hardware) and move on.
I can connect to 5 servers blindly and get so many different performance indicators, pooling them all together would be meaningless.
And with the local client having a big play in performance also, as well as route to server from a client, we can not just make players vote for servers. Gods, how many times ive seen folk play on servers which were limping performance wise, its quite baffling.
And to get rid of folk blaming servers which actually run great as far as can be seen, we would really need to wait for all the hitreg fixes.
Yes, that would be the most minimal kind of matchmaking (just to get it rolling). Mid-term goal should be a system that scales with growing usage numbers (increasing its matching-quality in the process). It should also minimize waiting time at the expense of matching quality because the average player won't be patient enough to wait for ages for the perfect match (tm).
Exactly.
You are right, a 2000 skill player may dominate a 1500 skiller. Even more so vs. a <1000 skill player or a rookie. But if I compare it with what we have right now, I think we can only win. The question is not if such a system would be perfect in the beginning but if it will increase the probabilty for better games (tm) - especially for the more inexperienced players (on who we should focus on if we want to grow the player base).
As I said, I believe it is not so important to keep average skill players (~1300) away from the veterans (2500+). But if new players do not have to encounter the vets in the beginning, that would be a big win.
Yes, the system may match 3000 skill players in the same server group as a 1300 skill player quite often, but I believe even that would occur less often than currently.
Less rookie vs. veteran, less frustration, better retention.
@DC_Darkling: I agree excellent server management is a key requirement for any of this to work. If I were to deciede, I'd try to recruit well known server admins (whoose servers have been used consistently) as well as include comp players in the discussion (since they should be most expert on what really matters). I believe in meritocracy, where those who contributed the most valuable work in the past get to deciede on what to do next.
I hope that will be less so when all the hitreg issues still in the game have been fixed. Although most of the bigger servers these days are actually not doing bad.)
The next biggest barrier to such a system would be servers as DC_darkling brought up.
Beyond that it becomes social. Everybody is used to the server system we have now. Even though hive works on similar principles to matchmaking in games from Riot, Valve, and Blizzard there appears to be a large number of people who think hive is a failed system. Allowing people to sit in a queue would while playing in a server be a start, but how would you encourage people to use this matchmaking system?
The underlying crux of the issue you are trying to solve is that the expectations vary greatly when a player decides to load up NS2. This is normal and healthy in any game. The unfortunate thing is how the "wild west" server browser is set up, it ends up pleasing nobody, satisfying few, and disappointing all.
For example:
-Some just want to play specifically with a group of friends (Does NS2 have a party system? No.)
-Some may have just completed the tutorial and want to learn the ropes under no pressure (Is there some kind of gating system for them to ease into multiplayer? No.)
-Some want a competitive game with high skill players with reliable teammates focusing on the current meta. (Does a ranked queue exist that ensures the majority of players would be trying their best? No.)
I've said this before but the fact that the extremely misleading and harmful "rookie-friendly" distinction still exists is a travesty. When setting expectations should be such a priority as outlined above, this is just a slap in the face to anyone that ever cared about player retention.