-
Connection web after 10 matchmaking cycles with 50 users. Lines represent mutual trust relationships. S are ideal users and D least ideal
-
50 users after 50 passes of matchmaking. Color scheme represents percent matchmaking success so far. 0% matches found through 100%
-
A zoom in of the 50 user, 50 passes central web
-
50 users after 100 passes of matchmaking
-
A zoom in of the 50 user, 100 passes central web
-
50 users after 1000 passes of matchmaking
-
A zoom in of the 50 user, 1000 passes central web
Inspiration
Toxic multiplayer experiences have seemed to be ever present in my online gaming life. I wondered if utilizing the decentralized web of trust model from cryptography might inspire a different way for people to connect through matchmaking.
What it does
This project simulates a matchmaking environment. Instead of a centralized power handing out bans players simply note whether or not they would like to play with someone again after playing. This is the trust in the web of trust. When evaluating new players to matchmake with the trust of other players in your web can be taken into account, naturally phasing out bad actors and favoring players who cultivate good relationships with others.
How we built it
This project was built in python with the visualization using the networkx and matplotlib libraries.
Challenges we ran into
Turns out algorithmically combing an ever widening web of users is a bit of a nightmare. Had to scale the project somewhat to make it workable on a desktop. Had a feeling I might be stumbling on an NP-Hard kinda problem and I think I definitely did.
Accomplishments that we're proud of
I think the visualization turned out decently, and initial results were promising in showing good faith players being rewarded and bad faith players being phased out. Ultimately the project also left a lot of questions to be followed up on which is exciting in a way.
What we learned
Matchmaking is hard. I already knew this, but now I know know it. I think there's promise to a web like structure, but scaling it up to the 1000s and introducing latency/skill considerations certainly make it all the more daunting.
What's next for Just Trust Me
Probably more tinkering and graphs. I think there's still plenty to be learned by just tweaking parameters.
Log in or sign up for Devpost to join the conversation.