Data is often too easy to misinterpret, and I think even here you misinterpreted it slightly.
This skews the data towards Fischer unfairly, and Byo-yomi still was close to winning for 19x19. I’m also not sure whether the data includes the effect of confusing beginners with two very different time settings. The first few times I used automatch this year, I didn’t even realize the left side time settings (30s + 5s) were Fischer until the game started, and I’ve been playing on OGS since 2022.
I’m not against using data, I’m against overly relying on data-driven development for design. It’s how engineers can compensate for limited artistic abilities as well as it’s a way to show progress in companies, but by definition it replaces intuition with empirical evidence, which can be counterproductive in some cases. Intuition is very powerful, especially compared to skewed, incomplete, or misused data sets.
Yes, this is the way. Disregarding preferences, this is better for faster matching.