From Rivals to Teammates: Learning to Coordinate by Competing

"... performance improves through repetition [because] repeated collaboration improves coordination among teammates." [1] But is this also the case for repeated competition? Can a team of former enemies work better together than a group of strangers?

🥵 Competitive Familiarity in Elo Hell

All of us stuck in Elo hell must surely have made the experience of repeatedly playing against, and probably with, the same other people online. In an earlier episode, we have learned that playing together multiple times (cooperative familiarity) increases the team's performance. But does competitive familiarity—playing together on the same team after having previously played against each other—also enhance team performance? The researchers believe it could, because:

"The intensity and salience of competition provides teammates with unique insights into one another's behavior, leading to improved coordination among contemporaneous teammates" [1]

In other words, if you’ve seen firsthand how good another player is and what they typically do—and vice versa—it’s easier for both of you to coordinate and adapt.

To test their idea, the researchers analyzed about 6.4 million Dota 2 matches, including 4,272 professional and semi-professional players. The key variables were:

• Cooperative familiarity (how often a player has played with a now-teammate before)
• Competitive familiarity (how often a player has played against a now-teammate before)
• Team performance (team victory)

☯️ Compete and Cooperate

"The results show that competitive familiarity causes team performance to improve meaningfully." [1]

According to the researchers, this "suggest[s] that individuals are better able to coordinate with former competitors with whom they understand well." Depending on how often players have played against each other in the past, the probability of winning—when playing together in a team—changed (see the figure below). For reference, the mean of competitive familiarity was 1.92.

"Hence, a team with more than three competitive interactions is on average 9% more likely to win a match than a team without any competitive interactions." [1]

Another key finding of the study was that competitive and cooperative familiarity are complementary—meaning they work best together. For instance, if end up on a team with someone you have played with and against before, the odds of winning are higher than if you only had one of those experiences (competitive or cooperative). Similarly, if the opponent team has higher competitive and cooperative familiarity, your team's odds of winning decrease.

The last result I would like to touch on is the level of specialization team members have. "Players who typically play either carry or support are “specialists,” whereas players who often play both carry and support are “generalists”" [1] This, essentially, reflects the experience one have with a certain role. A specialist team has three or more members who typically stick to one role (carry or support). A generalist team has three or more players who frequently switch between roles.

"Specialist teams outperform generalist teams, and teams with high skill overlap underperform." [1]

Even when accounting for specialization, the positive effect of competitive and cooperative familiarity on team performance remains. For instance:

• If members of a team have competed against each other four times before, the team’s chance of winning increases by 5.03%.
• In comparison, if teammates have previously played together, their chance of winning increases by only 0.93%.

"The results speak to the duality of competitive experience... like cooperative experience... it builds familiarity, yet... competition facilitates a different, and perhaps deeper, kind of understanding of others." [1]

Thanks for reading. See you next Sunday. Best,

Christian 🙂

References

[1] Ching et al., 2024