VALORANT Round Difference MMR Issues
When VALORANT ranked released one of the more interesting pieces was the MMR system. Riot threw in some new philosophies that were aimed at more accurately describing your skill level. The first of which is an additional component of considering the KDA. While that system already brings up some questions (Should more supportive players be punished in terms of rank? Is incentivizing kills over wins healthy for the game ecosystem?), the more worrying portion for me is the use of the round win rate totals in the MMR system SOURCE TODO. The reason for this worry is the sidedness of the maps. Map sidedness is the difference in round win rate between attacking and defending teams. CS:GO players will remember the old Nuke map where it was common for T(Attacking) teams to only win a few rounds on a side. This sidedness does not affect the win percentage of the overall game, because both teams get the same amount of rounds on each side of the map (if needed). However, the sidedness DOES affect the amount of rounds each team is expected to win.
What causes this difference? The problem arises because each game is not played to the full 24 rounds. The issue is easiest to see on a map that is heavily one-sided, for example 2 Attacker-10 Defender. Imagine that there are two evenly matched teams playing, Red and Blue. Red starts on the attacker side and wins the expected 2 rounds. Then after the side switch, Blue wins 3 rounds in a row. The game is over which is a problem. Red only played 3 rounds on the favored side, and if the rest of the rounds were played out it is expected on average that they would win 8 or 9. Unless Riot is accounting for the sidedness of maps, the MMR of the Red team would be unfairly degraded because they weren’t given the chance to earn those extra rounds.
Here is an example plot of the probability that each team wins on a certain round:
While both teams have a 50% overall chance of winning the graph shows that team A is more likely to have to play more rounds to win because of the side win rate difference. (You can play with the spreadsheet here.)
One answer Riot may have to this problem is that it does not matter because players will play on both sides of the map from match to match. However, that solution does not work well with placements because, while players will start both sides of the map eventually, placement games matter more. Another answer may be that the maps are balanced so that one side is never more likely to win than the other. Riot may have the maps in a close state, but maintaining a perfect 50% side win rate across all ranks on all maps would be a herculean task. The best answer is accounting for the sidedness of each map (perhaps on a rank by rank basis). With full access to match data Riot would be able to “undo” the effects of the map sidedness by figuring out how many rounds the teams finished from expected rather than just from the raw round values.