The goal of a matchmaking system is to produce worthy matches. Whether the match in question is made by a dating site or a a ranked videogame (e.g. League of Legends), someone must design an algorithm that takes in data about the players and arranges the players according to the data. However, the results are often unsatisfying. In videogames like League of Legends, some players describe the unsatisfying results of matchmaking as “Elo Hell.” Some believe this Elo Hell does not exist, but others swear they are stuck there. I wonder: if it exists, can Elo Hell be described as a failure of data?
The data at issue is supposed to be “how good a player is.” However, we have a hard time evaluating that ourselves- so how can we represent a formula for something we cannot describe? Can we create systems whose basis we do not understand? Ultimately, companies rely on data- but more data than they can manage. So they create algorithms to manage the data: information to govern information. But our programmers just aren’t perfect. Even when the code is error-free, the programmers cannot anticipate every use by every user. Google Chrome sometimes asks to translate my news pages from a foreign language into English- even though they’re in English; Microsoft Word keeps thinking I’ve misspelled or misused words when (after 5 minutes of being completely psyched-out) I am quite certain I have done no such thing.
What these various examples demonstrate is the way that our use of data can be limited: code can mishandle information (trying to translate an already-translated webpage), or sometimes the data itself might be unknowable. In this way, there are at least two kinds of limits in data management: code malfunction and data (input) weakness (relative to the desired use).