We want to kick off this post by reiterating our appreciation for everyone that has participated in test events and provided feedback during this test period on Zwift Racing Score. With this data the team now has a clear path of tasks defined to get to the point where we’d consider Zwift Racing Score graduated from Zwift Labs. This post is intended to explain what changes you can expect to see as we prepare to graduate the feature.
Race results now include more transparency about hitting or establishing a minimum/floor score. We’ve received a lot of questions and confusion around why the resulting score from the most recent race result does not match the current score listed in the profile page. As of this week, the race results on zwift.com have been updated to accurately show when a racer encounters their floor with new indicators. The same change is coming to the Zwift Companion app next week.
: Your Racing Score increased after this race.
: Your Racing Score decreased after this race.
: You’ve set a new, higher, floor value for your Racing Score after this race.
: Your Racing Score has decreased to your floor value after this race.
We have begun slowly increasing score volatility, or how much a score can change from the result of a single race. It was useful in early days of testing to ensure scores didn’t vary too wildly from race to race, but the data (and your feedback) is showing we can release the handbrake a bit. This change won’t be immediately apparent, but as racing continues, we will start to see an increase in how quickly scores move.
Starting today, your score can drop 15% lower than where you were seeded. Before this, your initial seed also served as your floor or minimum score. Our original seed algorithm tried to undershoot the prediction slightly to account for this, but we’re seeing many cases where people finish in the bottom half of the field consistently and their scores are not allowed to drop any further. This creates a poor racing experience for those racers since they are never able to adjust to their ideal category. Now that the floor is slightly lower than seed, we hope that this provides the flexibility required to ensure the model works for everyone.
In addition to the changes that have already rolled out, you can expect to see a couple more updates over the next couple of weeks.
- A more accurate seeding and floor formula that increases the threshold of what we consider a “quality power curve”
- Support for filtering general classification standings on Zwift Power by racing score ranges
While these are the tasks we plan to tackle in the short term to get ZRS ready for primetime, we’ll of course always be improving and iterating. Keep the feedback coming!