Smurfing is Bad

You heard it here first

Smurfing - the practice of highly-rated players circumventing the matchmaking system to match against lower-rated players - comes up regularly as a point of contention in real-time strategy communities. I’ve seen it come up in every game and every franchise. It’s a topic that gets emotions running hot! And it always seems to attract a vigorous, protracted debate on whether and when it’s OK.

I think discussions around smurfing would benefit from focusing on potential improvements to a game’s automated matchmaking system, rather than getting caught up in an endless debate over whether or not the practice is fair. I think that way for several reasons; but mostly, it’s because I think smurfing is bad, and I think it’s bad in a not particularly nuanced way. People should just avoid doing it, and it really isn’t more complicated than that.

Automated Matchmaking

To offer some context on my perspective, I think smurfing often gets framed in the wrong way, and I think this is an obstacle to productive debate. I see this as similar to how unworkable definitions of game balance can make that a difficult subject to reason about, too. When you begin a discussion from a bad starting point, it’s difficult to reach a satisfying conclusion.

In this case, I often see folks focus on the idea that smurfing is about high-level players dunking on lower-level players. And while that’s a fair take, I think the challenge with this framing is that occasional mismatches are a known tradeoff of automated matchmaking. They will happen, from time to time, no matter what, especially to players in sparsely populated leagues like Grandmaster. Because of that, smurfing discussions often devolve into a general discussion of what to do about mismatches, resulting in comments like “try to learn what you can” or “it’s just ladder points”. And I get why people land on these arguments, even though I don’t think they make a whole lot of sense in the context of smurfing specifically.

Let me rewind a bit. Back in the day, as it were, everything was a custom game. In Age of Empires II, for example, players on the Microsoft Zone played ranked by creating custom game lobbies and manually filtering out opponents until they got what they thought was a good match. Typically they’d put their ELO target in the lobby title, and when players joined, they’d check their rating and match history and latency and so forth, agree to some rules for the game (e.g. what map to play), and then get started.

This was, as you can imagine, a cumbersome process. Players had to create lobbies and patiently wait for opponents; then, painstakingly quality control said opponents; and finally, essentially invent parts of the competitive rule set like what map to play on, while hoping their opponent agreed to it.

Automated matchmaking emerged as a great solution to all this. I’ll use Age of Mythology as an example since it was the follow-up to Age of Empires II, although the general principles apply to all systems. Players started out with an ELO rating of 1600; when they queued up, the system tried to match them with similarly rated opponents, and added or deducted points based on the outcome in the range of [-20, +20]. The system took care of tedium like what maps to play on - introducing variety in the form of maps like Midgard, Anatolia, and Watering Hole - and what rule set to use - the game speed and so forth.

The upsides were significant. Reduced wait time and friction to get into a game created a virtuous cycle that encouraged more people to queue up and play, further reducing wait times. The adoption of a varied map pool and a consistent rule set meant that the proper competitive ranked experience was accessible to the entire audience. And the use of ELO, in combination with automated matching, meant that players were able to consistently get pretty good games against similarly matched opponents.

I mean, in general, I don’t think I need to sell anyone on this - there’s a reason every multiplayer game moved toward automated queueing and away from a custom games first experience. But a key downside is that that there are always going to be edge cases that are difficult for automation to handle. New accounts are a great example - how does the system differentiate between a new player’s first account and an existing player’s new secondary account? If you assume the former, you create a lot of mismatches when an experienced player climbs back to their normal rating; if you assume the latter, you ruin the onboarding experience of newbies. Either way, you’re making a trade-off, and somebody ends up getting some bad matches.

Fortunately, in the grand scheme of things, these trade-offs end up being pretty small. Modern games use rating systems like TrueSkill, which are remarkably good at accurately determining a player’s rating in only a handful of ranked matches. And even when you add up all the possible sources of inaccuracy - like players returning from long breaks and being rusty, or players maintaining a second barcode account, and so forth - the handful of mismatch cases are a drop in the bucket compared to the total number of games a player might accumulate across their entire ladder career. And the more players you have in your game, the less of a problem this becomes as ratings settle out across the population.

Bad Intentions

To sum up, while automated matchmaking is an imperfect system, its myriad advantages outweigh the handful of flaws that, anyway, get smoothed over as players play more and more ranked games.

Smurfing, then, is best understood as circumventing the matchmaker. This typically means exploiting weaknesses in the automation to intentionally create bad matches, like intentionally leaving games to artifically lower one’s skill rating, or routinely creating fresh accounts in order to beat up on newbies.

(Of course, you can smurf in custom games, too, but I’m focused on competitive ranked multiplayer at the moment, and custom game smurfing is a smaller concern in that context.)

Again - occasional bad matches are expected in any automated matchmaking system. Their existence, by itself, is not smurfing. Smurfing is when a player intentionally creates bad matches.

And that’s bad. Moreover, it invalidates most of the arguments you hear to defend smurfing - that it’s just ladder points, that it’s just one bad game, that you should “feel honored” to play against a top pro - and so forth. The question isn’t whether an imperfect system occasionally produces bad games in its mission to otherwise produce good games. The question is whether someone is intentionally creating bad matches that lack the trade-offs that normally make such games acceptable.

And I don’t think it’s a particularly complex moral argument to point out that intentionally doing bad things is… bad.

Do you need to intentionally circumvent the matchmaker to do what you’re trying to do? is a crystal clear, easy to use framework to evaluate whether and when smurfing is OK. It cuts through the BS and, as an added bonus, forces you to state upfront what you’re trying to do in the context of automated matchmaking.

For instance, when it comes to creating content, a player that wants to see how high they can climb while doing meme builds can maintain a separate account expressly for that purpose. This minimizes the number of mismatches to the first handful of games; afterward, by reusing the same account over and over, they enable the matchmaker to create good quality matches from there on out. Over time, the total number of mismatches ends up a very small proportion of the total number of games they contribute to the ladder.

Now - is this as much fun as creating a new account for each new video and dunking on Gold players? Well, no. But that approach creates unnecessary bad matches and bad matches are - wait for it - bad. Bad things are bad and if you can avoid them, you should.

Anytime someone tells you they “have” to smurf, really ask them - does this require you to intentionally circumvent the matchmaker? If you’re making a guide, can’t you just play against the AI (to showcase a clean build) or cast a replay at your own level? If you’re doing meme builds, can’t you use a meme account and simply maintain a GM MMR to prove you’ve done it? If you’re trying to help Platinum players specifically, can’t you play custom games against Platinum volunteers, or review replays from Platinum players, instead of smurfing them on the ladder?

If you’re such a famous celebrity that people would be honored to play against you - why not just host custom games and allow people to join if they want to?

Intentionally creating bad games with unwilling opponents is bad. Practically all of the reasons people cite for doing it can be achieved without creating bad matches.

Focusing on the Matchmaker

Look, to a certain extent, I get it. I get ladder anxiety from time to time, too, and the thought of jumping on a new account and grabbing some easy wins is tempting. It’s human nature! And part of the reason I like framing this as a matchmaking problem is that it clears out all the normal excuses one might offer for smurfing, and encourages folks to get back on the straight and narrow of playing on their main account(s).

But I honestly don’t think smurfing deserves a whole lot of mindshare. Bad-intentioned people selfishly creating trouble for others in order to benefit themselves is not an activity that you should try to derive upsides from. It’s not something that should give you pause, or make you think huh, let me bend over backwards to find the good in this intentionally bad behavior. It’s one thing to respond to smurfing in a mature and considered way, but it’s an entirely different thing to defend it.

Bad things are bad - just stop smurfing!*

And I think one thing we can do as members of real-time strategy communities is push back on this behavior when see it. No one should be run out of a community because they make mistakes. But bad behavior shouldn’t be defended, either. And we should rightfully ask professional players and content creators to be more respectful of their communities and avoid engaging in bad-faith activities like smurfing.

(Personally, I think prominent community members are sensitive to community sentiment and will adjust their behavior in response to good-faith feedback that something they’re doing is not good.)

Beyond that, I also think RTS would benefit from more general discussion of how to improve automated matchmaking. As much as I like to pick on folks with less than perfect intentions, I also think there’s a large amount of “unintentional smurfing”, too - mismatches that could have been avoided through improvements to the automated matchmaker.

One good example of this is skill ratings for team and solo play. StarCraft II, Age of Empires II, and Age of Empires IV all maintain independent, unrelated skill ratings across these games modes. This is not great, at least as far as placement matches go; while a 1v1 rating is not a perfect approximation of a team rating, it is substantially more useful than no information at all. Adjusting the skill rating algorithm to either use a different base initial rating, a different uncertainty score, or to try to arrive at the right rating faster could eliminate a lot of avoidable mismatches.

Another good example is regional ratings. Again, while not perfect, a Diamond player in North America is going to be somewhere in that ballpark in a separate region. Ideally, ratings in one region would inform a player’s initial rating in different regions, similar to the way they soft-seed a player’s initial off-race skill rating in StarCraft II’s separate-MMR-by-race implementation.

Basically, at a high level, I think there’s a large opportunity here. Maaaaany years ago, I wrote about the intricacies and nuances in both StarCraft II and CS:GO’s matchmaking systems. And one of the conclusions I drew from that is that there is a huge amount of complexity here. A lot of thought goes into how these systems are implemented, and there’s so much that could be made better.

(And so much that could be borrowed! I would be curious to see how a solid implementation of CS:GO’s trust score system would work in a sufficiently well-populated RTS.)

Automation is a powerful lever for improving engagement. It is, for many players, the primary way by which they find and play games. Making the system better has potential to pay enormous downstream dividends in improving player engagement and retention. We should invest more time and energy into making this better - and less time and energy justifying why bad behavior is good, actually.

Bad things are bad - you heard it here first.

Until next time,
brownbear

If you’d like, you can follow me on Twitter, Facebook, and Instagram, and check out my YouTube and Twitch channels.

* For anyone out there still stubbornly insisting on the classic “it’s just ladder points, bruh” angle on all this, I’d ask - how much time does a single instance of smurfing waste, and how many times would someone have to cut you in line at the grocery store to sum up to that? How many line cuts would you accept without getting just a teensy bit irked? What if the person turned to you and said “it’s just a couple of minutes, bruh”?