Streaks in Overwatch, Simulation and Analysis and Data

This post is centered too heavily on averages. The reality is that it illustrates a possibility for a fair system to create these results, and nothing more. Wins/losses are not a coin flip system, if you assume skill to be accurately represented then each loss should increase your chance of winning the next game by a small amount(as it’s moving you into a lower bracket where your personal influence is inherently higher). By the time you’re at a streak of 8-10 losses, you should be playing with a subset of people that’s notably easier to play against.

The ‘data’ used in the post seems to be sloppily self recorded, parts of it are missing and it has comments like ‘undocumented loss streak’ and notes that are generally indicative of the person recording it not being very clever. I am sure it’s not Kaawumba’s own data.

While a 40-60 chance may be predicted, that does not mean the individual match in those circumstances is really 40-60. I imagine most stomps would play out the same way if the teams played several games in a row; there are identifiable players who are holding their teams back or carrying. Thus, any factor that can be used to make it more likely you get those poor players in the interest of maintaining a perceived 40-60 range, is rigging.

As ANTCRSTM continues to mention, Blizzard is a company with their greatest interest in profit. Promoting engagement at the expense of fairness is a logical decision for them to make, and they’ve been very tight-lipped about every aspect of matchmaking. Let’s not forget that they banned pursuit, added private profiles(with a default to on), and things of this nature.

When you get grouped with someone who has a 0% winrate over 20 games(has happened to me), it’s pretty obvious that person is doing something to force their team’s chance of success far below 40%. How can matchmaker be confident a game is in the 40-60% range while placing that person on a team? Surely the best they can do is give them 5 teammates that are seemingly excelling, and even then the match is likely not going to reach 40% win chance. Fortunately, in most cases you have absolutely no way to observe things like this, because profiles default to private and most people don’t even notice it’s a setting to begin with.

2 Likes