Dynamic Queue Statistics for 61,977 Matches

I found some really interesting data on dynamic queue matchmaking. I think you guys will be interested in it as well. Riot will not tell us the facts but that doesn't mean we can't figure it out ourselves. All credits go to richkcho for doing the statistical analysis. All the below is quoted from his report which is linked at the bottom of this post. > I amassed 61977 matches which included at least one challenger or master player. The data scraping attempted to obtain around 100 of the most recent ranked matches per player in challenger/master in this case starting from April 23rd. (This would vary based on players activity, inactive players would have older games in their past 100 games compared to a more active player) > > The matches were analyzed to find premades, premades being a group of players who have played at least 10 games together on the same team. There were a total of 1129 games which had at least one 5 man premade. (this might be why Riot doesn't care, matches that include premade 5's statistically speaking are a small minority of matches) The team facing the 5 man premade had the following distribution of premades: > > * 5 solo: 255 (23%) > * 1 duo, 3 solo: 313 (28%) > * 2 duo, 1 solo: 58 (4%) > * 1 trio, 2 solo: 220 (19%) > * 1 trio, 1 duo: 45 (4%) > * 1 quad, 1 solo: 168 (15%) > * another 5 man premade: 70 (6%) > > I also ran win-rate analysis. The win rates of the 5 man premade vs the other team based on the premade breakdown of the other team is as follows: > * vs 5 solo: 70% (179/255) > * vs 1 duo, 3 solo: 58% (183/313) > * vs 2 duo, 1 solo: 60% (35/58) > * vs 1 trio, 2 solo: 63% (138/220) > * vs 1 trio, 1 duo: 53% (24/45) > * vs 1 quad, 1 solo: 54% (91/168) > So Riot decided to limit the premade size to three as a band-aid approach to fix the problems elaborated above. The natural question is is this fair? > > I ran the same analysis on the dataset but this time focusing on 3-man premades. There were 8170 games with at least one 3-man premade (excluding any games with 4 or 5 man premade). The breakdown of the premade distributions of teams playing against a team with a 3-man premade were: > > * 5 solo: 4012 (49%) > * 1 duo, 3 solo: 2978 (36%) > * 2 duo, 1 solo: 548 (7%) > * a group with a 3-man premade: 632 (8%) > > Win rates were as follows: > * vs 5 solo: 61% (2464/4012) > * 1 duo, 3 solo: 54% (1604/2978) > * 2 duo, 1 solo: 52% (386/548) Tl;dr, Premades play teams of mostly solo players the majority of the time and destroy them. Full report: https://github.com/richkcho/league_of_tilt/blob/master/other/flaws_of_dynamic_queue.md
Share
Report as:
Offensive Spam Harassment Incorrect Board
Cancel