Quick & Dirty
- New rating system based on Glicko2
- 70% more accurate than Elo
- Is based on every game played by every player like Elo
- Has a hardened rating (exposing this soon!)
- Starts at 1500
- We are in Season 1. It began Feb 17 w/ the weapons patch
What is it?
Destiny MMR is based on Glicko 2. It was written as a successor to Elo. In Elo your rating went up by a maximum K-favor. The default K-Factor is 32. This is a slow way to get a player to an accurate rating, and Elo does not track any certainty data.
Glicko2 is the up and coming rating used in tournaments and competitive FPS games like CSGO because it tracks more variables and arrives at an accuracy rating more quickly.
Why do this?
Elo was designed for chess. It is based on 1v1 and was never made for team play. Glicko2 can be used for team play. It also tracks multiple variables so that as a player plays more, their rating will become "stiffer". To start, your rating will fluctuate up and down quickly and your rating "hardened" percentage will remain low. As you continue to perform as expected, your hardened percentage will increase and you will see much smaller movements up and down.
Why is it called MMR
We have implemented Glicko2 by the spec, perfectly. But when analyzing Destiny Trials games, we have implemented some logic to catch cheaters. Dis connectors, de-rankers, etc will be banned and their ratings zeroed out. Thusly, this has been adapted to fit Destiny.
Destiny MMR is better
Want proof? We took a random sampling of 2000 games played in March. We pulled the Elo and the MMR for those players. We then computed who Elo predicted to win, and who MMR predicted to win. MMR was 70% more accurate when compared to final results! When Both Elo and MMR predicted the correct winner, MMR prediction was better! When both systems failed to pick the correct winner, MMR was still a better prediction based on percent likelihood to win!.
More proof: 2000 games analyzed.