Browsing through the PDF, I’m getting the vibe that their way of measuring “skill” is weird. They claim to use multiple methods of measuring, they list a few obvious ones that they’ve found to be bad, but they don’t say which ones they are using because “we are constantly iterating on our performance metrics to optimize the player experience per game-mode”.
Elo-like systems tend to adjust skill based on the chance of winning current match X win/loss, but they’re not (just) doing that. I wonder if they have a few weird metrics that look good on paper/in the lab but don’t feel good in play.
Browsing through the PDF, I’m getting the vibe that their way of measuring “skill” is weird. They claim to use multiple methods of measuring, they list a few obvious ones that they’ve found to be bad, but they don’t say which ones they are using because “we are constantly iterating on our performance metrics to optimize the player experience per game-mode”.
Elo-like systems tend to adjust skill based on the chance of winning current match X win/loss, but they’re not (just) doing that. I wonder if they have a few weird metrics that look good on paper/in the lab but don’t feel good in play.
This might be a relevant starting point for you: https://www.microsoft.com/en-us/research/publication/trueskill-2-improved-bayesian-skill-rating-system/