### Better ranking system than A/B/C/D and W/kg?

Posted:

**Sun Mar 03, 2019 11:24 pm**I've been pondering how bike racing (and Zwift racing in particular) handle ranking and categorization. I realize that grouping into broad classes has made a lot of sense in the past. In real life, having Cat1 racers race with Cat5 racers can be dangerous!

But Zwift is different. Wouldn't it make more sense to do something like the ELO or Glicko rating systems, used in Chess and other zero-sum games? https://en.wikipedia.org/wiki/Elo_rating_system or https://en.wikipedia.org/wiki/Glicko_rating_system

The math is a bit complicated, but the way it works is intuitive and fair: You get points when you win, you lose points when you lose. If you beat a stronger player you win more points than if you beat a weaker player. If you lose to a stronger player you lose fewer points than if you lost to a weaker player. Over time the math makes it easy to compare competitors.

When a person completes a race on Zwift, you could calculate the pairwise scoring against every other rider, then integrate the results to get the new rating.

Many benefits from this:

1. Rating would be based on performance history.

2. The math has probabilities associated with it. If some magically performs far above/below their rating, the math can help in the detective work.

3. The ratings can be recalculated trivially based on age / gender / nationality / whatever buckets. For instance, older riders could know how they stand in the "female 60-65" group. There is really no way to know this if we only have A/B/C/D.

Sidebar: Is the ZwiftPower data public? I would love to take a crack at what I just described, and see what the result looks like. I'm a software engineer by trade and I have the skills to make it happen, but I don't have the data . (I'm sure I could scrape the data with a crawler, but I don't want to violate any terms of use )

Cheers,

-Mark Rebuck

http://www.markrebuck.com/

But Zwift is different. Wouldn't it make more sense to do something like the ELO or Glicko rating systems, used in Chess and other zero-sum games? https://en.wikipedia.org/wiki/Elo_rating_system or https://en.wikipedia.org/wiki/Glicko_rating_system

The math is a bit complicated, but the way it works is intuitive and fair: You get points when you win, you lose points when you lose. If you beat a stronger player you win more points than if you beat a weaker player. If you lose to a stronger player you lose fewer points than if you lost to a weaker player. Over time the math makes it easy to compare competitors.

When a person completes a race on Zwift, you could calculate the pairwise scoring against every other rider, then integrate the results to get the new rating.

Many benefits from this:

1. Rating would be based on performance history.

2. The math has probabilities associated with it. If some magically performs far above/below their rating, the math can help in the detective work.

3. The ratings can be recalculated trivially based on age / gender / nationality / whatever buckets. For instance, older riders could know how they stand in the "female 60-65" group. There is really no way to know this if we only have A/B/C/D.

Sidebar: Is the ZwiftPower data public? I would love to take a crack at what I just described, and see what the result looks like. I'm a software engineer by trade and I have the skills to make it happen, but I don't have the data . (I'm sure I could scrape the data with a crawler, but I don't want to violate any terms of use )

Cheers,

-Mark Rebuck

http://www.markrebuck.com/