Match Seeding feature?
TinCan
Join Date: 2006-12-11 Member: 59010Members
Join Date: 2006-12-11 Member: 59010Members
When 316 rolled out I eagerly started up the game in hopes of finding a match seeding feature. Can anyone elaborate on the progress of this idea? Is it still moving forward? Stuck searching for a server that is not full and has players I see 15 to 20 players in the browser that apparently are also looking and I think, heck, that's a full server right there just waiting to happen.
I wonder if rookies will be treated differently?
Will we see something on the next patch?
What are the drawbacks?
What is the delay?
I wonder if rookies will be treated differently?
Will we see something on the next patch?
What are the drawbacks?
What is the delay?
Comments
@McGlaspie
@McGlaspie
Originally, Match Seeding was slated and being tested to roll out with B315. However, we cut it in order to publish 315 faster, as there were a number of lingering issue with it. It just wasn't ready yet. B316 was never intended to be a big feature patch, as we wanted to get it out (and address long running problem of Boneshield) before the Summer Sale hit.
Yes, Match Seeding is still very much in the works. I'm basically doing almost all of the work on it. The core features works fairly well. It was pretty cool to see 8+ playtesters get put onto the same server within about ~50 seconds from start to finish (time of button press to people connecting to the server). The main goal is to roll out Match Seeding (note, I'm not using the phrase Match Making; there is a distinct difference in functionality) in phases. This is not a concrete plan, instead should be viewed as milestones, which could change as more functionality gets added and tested. For example, Phase 1 was originally slated to just be an "enhanced" version of Quick Play (e.g. Groups of players, instead of a single client). However, taking into account our time estimates for B317, I'm probably going to roll in Phase 2 in addition to Phase 1 features.
It has the capability to search and filter results on just about any number of parameters (e.g. Skill, Geographic regions, etc.). However, a lot of this will likely not be directly accessible on its initial release. Depending on its adoption/usage from our playerbase, we may or may not add more and more granularity to it (think fully automatic searches vs server-browser like searches). The reason being, is when taking into account the limited number of players looking for a game at the same time, their skill-ranges within that small subset, and then the available server-slots at the time of searching...the viable number of match-ups (player -> groups -> server-slots) gets smaller. It's distinctly possible, under manual control, to filter out perfectly viable games if the constraints are too narrow.
The underlying algorithm used when running a search takes both available player-groups, server slots, and takes Rookies status into account. It can also take many other factors into account (Server quality score, individual player skill, as well as the Group's mean skill, and a server's average ping for all members in the group). This will of course take time to tune, and we'll most likely have several public playtests to get "live" data for tuning it further and just stress testing it.
I don't know if this will make it into B317 or not, but my intention is to have both a automatic / public interface to it, as well as a manual invitation-only method (i.e. invite your friends to a group and then search for a server that fits your group). Unfortunately, this will not automatically allow a group of to always be on the same team, at least not yet. There are some especially annoying/complicated caveats that need to be examined before we can look at adding such a feature. But we certainly _are_ looking into things like this. The primary concern (for now), is to quickly and easily) get geographically and roughly similar skilled players into a group and all onto the same server, within 60 seconds. Keep in mind, the timing and efficiency of this process entirely depends on how many people use the feature at roughly the same time.
If you want more nitty-gritty details of what's going on with it, take a look at this card: https://trello.com/c/Pjj02lkw
I hope this answers any questions you folks have, cheers!
Indeed. The reason the playtesting example I cited worked, is because everyone pressed the button at essentially the same time. So, the speed of this new system will _entirely_ depend on how many people are "in queue" at once. Which also comes right back to how many people adopt it as the primary method of finding a server to play on. We're not going to be removing the server browser or anything like that; however, and food-for-thought, any one who jumps straight to the server browser will in essence be diminishing the match-seeding system efficiency.
It will replace Quick Play entirely for the most commonly used form of it. This is the simplest method for it to be accessible. The reason being is the system has a series of failover states it'll hit when X thresholds are met. Here is a simplified example of how it works:
So, as a result, it will still cover the use-cases of Quick Play for single-users, but also seamlessly fold in group-based functionality too. The only real downside to this is it will take a little bit more time for a single-user to get placed onto a server. However, considering it has the potential to massively increase the rate servers are populated, I feel this is worth the trade-off. If all else absolutely fails, we can add either modal behavior to the Quick Play button, or a second one for single-user only (although, I'm very much against this approach).
Assuming it makes it into b317, Private and/or Friends-Only Groups will be a full manual process, so it'll require the Lobby owner to invite specific users. However, this doesn't exclude turning a private group into a public one (which would then lump it into the above process). Now, things get really interesting when we take a looking at group-linking. This would open the door for things like two distinct groups of players being cross-linked via an orchestrator user (mostly handled by the system), which allows for two private groups to be treated as one (i.e. clan gathers or pcw type thing). However, to really leverage linked-groups well, we'd have to modify how servers get viewed by the system, so specific servers could Opt-in to being available for this group-linking match seeding method. For the sake of clarity, I won't be adding in group-linking for b317 and possible/maybe won't add it at all. It really comes down to how much the match-seeding gets adopted by our entire playerbase. If no one uses it, even after core features are solid, well, it just won't merit any more development time. Why? Because the current team's man-hours are at a premium due to limited availability. So, we cannot afford to spend time fleshing out huge complex systems that aren't being commonly utilized.
Also (even though you said it won't be implemented yet) how do you think a system where friends would be placed on the same team work? I have the feeling the ready room will have to be skipped completely in order to achieve that. And does the lobby select their prefered side on their own?
No it would not place you on the SCC server auto-magically. It's possible it could, but unlikely, because it'll be looking at all potential servers and weighting them accordingly. So, if the SCC server was the "best match" for you, then you'd be placed there. The use-case you're describing is mostly outside the scope of match-seeding. My apologies if I came across as accusatory with my remarks about people going straight to the server browser. Nothing hostile was intended there, instead, I was just trying to illustrate part of the possible reason why click-to-InGame time could be increased.
Interesting prerequisites. Possible, but not on the initial release. I do intend on ensuring the system will take server favorites into account (but not as a primary weighting factor). Your second item is unlikely to happen, sorry. Heh, props on the specificity. The match-seeding feature won't have personalization options like that. At least there are no plans for anything along those lines. Thanks for the feedback though, I've made note of them. Perhaps something along these lines could be in a Phase 3 or 4.
Come on 320, give me what I crave~