Referenced Paper: http://delivery.acm.org/10.1145/1090000/1080199/p116-jun.pdf?key1=1080199&key2=1552844021&coll=GUIDE&dl=GUIDE&CFID=18506295&CFTOKEN=73653288

BitTorrent is a protocol for downloading large files, and is often a faster alternative to regular client-server downloading. Essentially, BitTorrent is able to increase download speeds by utilizing a network of people downloading the same file at the same time. The network of people, in BitTorrent terminology, is referred to as a swarm. Every person in the swarm downloads different parts of a file from their peers in the network. Unlike other clients however, you simultaneously upload parts of the file you are downloading to other peers who need it the parts you have. An interesting caveat with the system is that BitTorrent requires users to upload to each other and also requires that there exist a few altruistic “seeders” who are willing to upload the entire file they possess.

As a user it may be tempting to refuse to upload to others and just download your file, but it turns out that you will be punished by the BitTorrent system. Your “leech” behavior and refusal to upload to others forces the BitTorrent system to “choke” you and reduce your download speed. Therefore, a dominant strategy is to share what you are downloading. In the paper referenced, this system is referred to as TFT (tit-for-tat) and is relevant to our prisoner’s dilemma problem when played multiple times. The TFT system always cooperates at first, and then copies the user’s previous move in order to end up at a Nash equilibrium all the time.

As we know from class, when the prisoner’s dilemma is played once, both players should “confess”, but with BitTorrent, the players are concerned with the future. Users are continually playing the prisoner’s dilemma game in order to download parts of the file they need. Therefore, in the long run, it is beneficial to coordinate. Essentially, if both people choose to coordinate they get better download speeds than if they do not coordinate. If they coordinate at not uploading, then they end up at a non-focal Nash equilibrium which is not the prediction of play because both players uploading would be more desirable. Additionally, the focal point Nash equilibrium (where both choose to upload) is a Pareto optimal solution because no change can be made where both players can achieve a higher payoff.

In the end, while the TFT solution has worked substantially better than other solutions, the referenced paper argues that the TFT system induces “free-riding”. Therefore, the authors of the paper propose a slightly better “incentive mechanism” for improving the results of the TFT method. For more information on this, click the link provided above.