Let’s Talk About Team Matchmaking

Going over the link referrals during my absence, I noticed a pair of reddit posts complaining about Team Matchmaking.  I’ve stated in the past that TMM ought to be the premier form of competitive matchmaking, but it’s also pretty clear to me that TMM as-is is not working, so let’s look at what’s going wrong and how these problems might be addressed.

The most obvious issue that comes up is that Dota’s Team Matchmaking doesn’t feel much like a ladder.  Teams appear to receive an initial placement based on the combined rating of the individual players.  As a result, the top of the ladder is full of teams that have barely played any games, and you encounter scenarios where the #8 ranked team on Dotabuff is 2-3 and the #514 ranked team is 180-29.  And of course this is an entirely reasonable outcome for a ranking system, but on the player’s end it feels unfair and makes your ranking seem predetermined.  How psyched can you get about a ladder where you can play 200 games at an 86% win rate and still be under hundreds of teams that are basically inactive?

My suspicion is that basic matchmaking “cheats” by attempting to predict a player’s placement in order to get them into appropriately ranked matches as soon as possible.  This rarely becomes an issue because matchmaking is almost entirely opaque in that it never really reveals what it’s doing and why.  TMM attempts to do something similar, but to do so it needs to do it in the open, and this undermines the ceremonial aspects to having a visible ladder in the first place.  Essentially, team activity needs to play a more important role in raising a team’s rating, not because it makes the system more efficient (it probably does the opposite), but because it incentivizes activity itself, which is something TMM desperately needs right now.

With that being said, the much bigger problem with TMM  is simply that not enough people play it, and just changing the ladder system won’t be enough to fix this.  For this, my earnest suggestion is to start letting the battle points flow.  I don’t claim to be an expert on the Dotaconomy, so I’m not going to make an estimate on how much extra BP the system can bare, but if the compendium can offer weeks of battle point boosts, then certainly the coffers can be opened a bit to encourage more team formation.

Specifically, what I’d like to see are a couple windows during the week that offer higher boosts and the first game played during these windows offers a large flat BP bonus on a weekly limit.  It’s similar I suppose to World of Warcraft’s weekly dungeon quests, but for significantly different motivations.

First, by tying the biggest boost of the week by playing just a single game, you most effectively motivate the people least likely to play TMM on a regular basis.  Casual groups that aren’t very organized might be motivated to TMM instead of just joining the regular queue just for the item boost, and individuals without a team can go to their preferred Dota chat during these windows knowing that teams missing a member might be looking for a 5th for that night.

Second, if the BP boost windows are on a regular basis, you greatly simplify the scheduling a group needs to do for TMM.  Sure, it sucks for people who cannot make the scheduled windows, or for that matter the same windows as their teammates, but one of the biggest failure points for a group event is having weakly defined meeting times so that you never have the full team available.  By limiting the options, you honestly make it easier for people to negotiate the option that works best for them, and you help them turn it into a regular event, which then becomes habit forming.

Third and finally, if all the teams are scheduling for the same group windows, the quality of matchmaking during these windows will be greatly improved.  The big issue with TMM is that it’s never going to have as many players as regular matchmaking, and those players will be further divided by 5.  This means that forming quality matches at any given point in the day will be much harder, and if a weaker team has trouble finding quality matches they’ll simply stop playing, which then makes it harder for the slightly less weaker teams to find quality matches.  By encouraging teams to queue all at the same time, you sidestep these issues which will hopefully keep people playing this mode instead of going back to regular matchmaking.

In addition to BP boosts, what I’d also like to see is some relaxation on team size limits.  In my experience in other leagues, the rule of thumb was that roster sizes needed to be double the required players in the game in order to minimize team no-shows.  That might be excessive for TMM, since outright noshows aren’t an issue, but I’d still like to see team rosters expanded to 7 or 8 players.  This will likely deteriorate the qualify of matchmaking some, but that deterioration should be offset by the benefit of having more active teams.  Having some extra roster room would make team formation feel less set in stone.  People could experiment more with who they invite to teams or which teams they join without feeling that a particular team will cease to exist if the same 5 people aren’t always around.  In addition to this, teams could be given a stand-in slot, but that admittedly opens another can of worms that may end up being more trouble than it’s worth.

All of this is in addition to normal ladder things like regular resets and awards for the top x% of players, but I believe that these two changes would be good first steps in order to encourage more people to actually use TMM.  It’s true that both will make it harder for the system to effectively matchmake teams, but let’s be honest, even a perfect matchmaker is going to create stomps, so the average player won’t be able to tell the difference.  On top of that, there’s a clear tradeoff that exists between rules that optimize for matchmaking quality and rules that make the system appear more exciting or that simplify the social aspects of team formation.  It’s entirely possible that a sub-optimal set of matchmaking rules for TMM might actually have the best matchmaking if they’re successful in driving more people to play TMM.  After all, the best matchmaking in the world is useful if an acceptable matchup isn’t going to enter the queue in the next half hour.  Any plan to improve TMM has to be focused around getting people to use the mode more, and it’s my opinion that targeted BP/item boosts and simplifying the social aspects are the two best tools available for accomplishing this.

Advertisements

7 Responses to Let’s Talk About Team Matchmaking

  1. jimmydorry says:

    There is currently no incentive to actually play any team games once you have played the three required (if you are only interested in the ladder rankings). If I recall correctly the TMMR is the average rating of each player… and winning or losing only changes that rating by a few points. However pub-stomping can drastically change the players’ individual ratings making it the fastest path to the top. Many of the top teams employ this method. When their rating from pubs has risen sufficiently, they can just make a new team, and enjoy the new rating. Now if you add ladder prizes, this will only be exploited more.

    The rest of your suggestions may improve the situation (assuming Valve fixed the fundamentals), but as you mentioned would introduce a lot more error as well. This would be less of a problem on the solo pub ladder, as you would expect people to play enough games to cancel it out… but on a team ladder there will be far less games played (as it requires a lot more effort to get a team together) and any increase in the uncertainty of ratings will take longer in real time to be fixed, and have a larger impact for a longer period of time.

    I personally believe that teams should be both explicitly and implicitly defined. If you party in a full stack, you should go into the Team Match Making, and you should get a rating for that team (Starcraft style) that you can view in your profile. Your suggestion to increase the team sizes would help, but it’s still a pain to create a team… especially if you only envisage playing a few games with that stack.

    Another thing worth mentioning is the frustration that comes from forming a team, and not understanding why you are playing against better or worse teams, as there are no metrics to gauge the individual player levels in your own team.

  2. Sheep ♥ says:

    I believe that just giving people a “status” like Bronze Silver Gold would be a huge incentive. And of course, having the means to change that status.

    • phantasmal says:

      The thing to remember is that some groups will not realistically have the means to change that status, and they are aware or will eventually become aware of this. It’s still important for the health of the ladder to keep them interested in playing, and that’s why I suggest participation rewards in addition to the usual top x% status awards.

      • DeathBot says:

        Well I feel like TMM (or another option) should be the one place that does have a revealed MMR. The whole point of TMM (in my head at least, another option is viable as I mentioned before) is to be the place where players can grow as team players and get actual team practice instead of playing pub games. I know I’ve talked about playing some cevo before, but it was actually really nice to see that, even though we weren’t great, we were still doing decently well and progressing. The whole point of the handful of tournaments that we played was to practice and see if there was any potential for us to win anything, and this was incredibly cool.

        Another option that I think would fulfill what both of us are asking for would be to run tournaments based on ingame dota teams. Have short leagues or tournaments or even weekend game drives where you get a point for losing a game and multiple points for winning one. The benefit of running tournaments would be that they could have teams with higher MMR scores get byes in tournaments, have them be preferred with side/pick order, etc. With all the items that they’ve got going on anyways, it would be an incredibly cool way to add incentives to the whole process, and Valve could get away with running tournaments essentially for free.

      • phantasmal says:

        Quick response because I’m supposed to be writing something else. Two big dangers involving tournaments.

        1. Tournaments are often designed in a double or single elim style in order to play everything out under a time constraint. While this is acceptable for major events between top end competitors, it really sucks for minor events composed predominantly of newer players or teams entering the scene. I’ll probably get more shit from FGC partisans for this, but smaller local fighting game events often have the problem where every single setup is devoted entirely to a tournament, leaving no room for casuals to the side and so that new players show up, go 0-2, and leave feeling a bit disillusioned about the whole thing. With an online Dota event we have all the time in the world, a setup for every single player, and first and foremost we’re trying to get good retention rates (particularly from the bottom end of the bracket). Given this, these kind of tournaments should ideally be designed similar to a Swiss-system tournament with a substantial tolerance for no-shows.

        2. Any tournament has to be very careful about using items as a motivator for winning. If they’re too restrictive, they won’t actually reward anyone but the people we already know are capable of winning tournaments. If they’re gated in a way that you try to have an amateur-only league, you’ll encourage rampant smurfing attempts. I don’t mention any awards/rewards for winning in the OP, largely because I assume that they’ll be there, but it’s important to keep in mind that they can undermine the validity of the league if implemented improperly.

      • DeathBot says:

        Yeah I definitely understand a lot of that. With regards to the first point, citing cevo again (this is my only experience with actual low level comp so I’m gonna lean on it forever) they ran it as a league for a while that would culminate with a top whatever bracket. Games would usually be like once or twice weekly, and rather than being at a specific time the TOs would tell you to have your match scores in by a certain date, and in this case all the communication would be done on the cevo website so they could see which team wasn’t able to get together. Between that and the team roster having a maximum of 8 (maybe? maybe 9 idk) it was a lot easier to get a lower skill stack to come together.

        I guess the way to solve this problem could be to make it a badge thing instead of an item thing. Either give out badges that could be applied to items and make them strange or whatever, or maybe just give out items for participation and then make top 4/8 have their items marked differently to show that they did well. And really the best solution to smurfing would be a games played minimum played to join/create a team, and then a smaller one for TMM games played to join a tournament. I don’t know how well it would work, but I think it should.

        It’s obviously a deep topic and I think Valve is smart enough to be thinking about it, but it’s gonna require a lot of management to make it really work.

      • phantasmal says:

        Well, along those lines, what I’d like to see is two separate efforts. The first is to make TMM a more inviting queue for the average team. The second is to give the playerbase the tools to easily create their own leagues that interact directly with the Dota 2 client. The easiest way to find out the ideal league structure is to get people to start making them in all sorts of variation and then let a sort survival of the fittest take over. Chances are that the ideal league structure is highly subjective to the actual community is based around, so there wouldn’t even need to be a one sized solution that has to support say, the ixdl (both open and closed), somethingawful in-houses, NeoGaf, TeamLiquid, the multitude of leagues that would spring up around reddit, etc. Give the players a solid toolset and the way to use API returns to track the league progression and people will start building stuff.

        And with that being the ultimate goal, TMM should ideally function as a way to get players interested in organized 5s with the fewest barriers to entry possible so that the interest in created that will be able to sustain the player-run leagues, with the hope that eventually some subset of the best of these player-run leagues will eventually hit the majors and maybe even become TI contenders.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: