ollipasanen
commented
3 years ago Hey Chris,
Thanks for the comments!
You wrote logistic when you meant logarithmic
Could you point out where this error was exactly? The mathematical language and terminology in this writeup is very sloppy at best. For example, by logistic correlation I've tried to say that some data points line up well (visually) with a logistic curve.
..., also it is my believe a conclusion should be a judgement or a decision reached by reasoning, you left it a little vague. You should define what your criteria is for a good ranking system and show if the data matches that hypothesis or not.
You are correct in that the conclusions are left slightly vague and that was done mostly on purpose. The goal of this project was to analyze the rating system and the matchmaking algorithm on FACEIT from a "data analysis" perspective, and not really to discuss whether the rating system is "good" or not. However, some predictive performance analysis was done in Chapter 5 and I noted some of my subjective thoughts about the fit of the rating system in Chapter 7.
As it was mentioned in the writeup, it's not trivial to conclude whether a rating system is definitively "good" or not. The accuracy etc. can be measured quite easily, but there are also lots and lots of subjective things involved too. For example, does the Elo rating system make sense in a multiplayer game like CS:GO? (Elo was meant to be used in 1v1 games like chess, not in a game where teams change and each player has their own rating.) Why does only winning/losing affect the rating and not the overall performance and other in-game statistics? The list goes on and on... I'd be happy to hear your hypotheses about a "good" rating system and maybe test them out.
Also AFAIK, FACEIT have recently upgraded their internal systems so the conclusions and analysis in this writeup might not even be valid anymore.
You wrote logistic when you meant logarithmic, also it is my believe a conclusion should be a judgement or a decision reached by reasoning, you left it a little vague. You should define what your criteria is for a good ranking system and show if the data matches that hypothesis or not.