@xflintx is there an update on what is happening with results based ranking and the revised Zwift score? I think I recall you guys saying there would be an update provided end of November.
While we are on that, is there any roadmap to bringing Zwiftpower completely in house and part of the Zwift client, or replacing it with something else for all racing administration and management? Zwiftpower has been an amazing community powered stop gap, but it’s been 7 years (or more?!) and it’s limitations such as still not seeing all riders are quite frustrating.
Also, if there are anymore exciting racing developments on the 2024 roadmap? I’d love to see live leaderboards for points races, with on screen prompts (like workouts) telling the riders when a Prime is coming up and then giving live results as the race progresses.
As mentioned in the linked post, we had some delays in getting it out, mostly around the overall implementation of some pretty robust backend systems to make the whole system work and which will make future changes more easy to do and accuracy of the scores more reliable.
What I was hoping would have been the announcement in late November would be the availability of public test events for everyone to participate in and give feedback on which would include two things: 1) the ability to sign up for specific scored races and 2) an entire new end-to-end flow for participating in those scored races (seeing your score, subgroup ranges, etc.).
For a bit more detail on where we are, here is what I can share with you:
Currently, we have scores for all Zwifters being calculated and showing on our backend. They’re still not to a place where we can yet say they’re good enough to test and then eventually release, though. This is for two reasons.
The first is that those overall backend system implementations I mentioned are also a necessary piece of the overall puzzle that need to be implemented before we are able to test publicly in a way that’s useful for both you as Zwifters and us as Zwift.
The second is that we’re currently working on making the tweaks to the various mathematics being used in calculating the Racing Score so that any Zwifter - whether they’re competing frequently or have never raced because they prefer workouts - can use and see their Racing Score and think “yeah, that feels about right.”
These two things are happening at the same time, with a common end goal and shared timeline; one is not holding up the other. So while those backend components get the final pieces settled into place, we are working on tweaking the scores to get them to feel right so we can hit the ground running with public tests.
I’m really excited to get this out to everyone because as you said, it’s been a long time coming, and I’m also of the mind that it’s better to have it delayed and released to a higher standard of quality than to rush it and have it feeling lackluster.
We aren’t going to impact existing series with the launch ideally. We have the ability to have races be scored or unscored, which will be easily identifiable on the front end for all Zwifters, and race organizers can choose to have their races scored or not depending on their needs. Our system is also not retroactive, so anything that was scheduled ahead of the launch to be a power-based race will stay that way until such time as the organizer wants to change, if they want to change to a scored system at all.
If we’re getting coffee as friends, I’m telling you I’m hopeful for January to begin testing, but I’m not betting the farm on it. This is just my personal feeling of when I hope it all rolls out, but with the holidays coming up and the ever-present potential for unforeseen complications to arise, that’s not a sure timeframe.
I would think there are a few learnings from zwiftracingapp that would be helpful to implement as part of the first go.
The first would be using either compound score, (or ideally the cool modified version zwiftracingapp uses to help normalize both light and heavy riders across the platform) for initial seeding. Then implement a compound score adjustment if someone well exceeds their compound score later through new racing results which would affect their ranking score.
Then I would expect that it would learn from what zwiftracing app did to help minimize sandbagging, and tanking - where they have limits on how far down you can go, and how one hugely sub-optimal effort won’t make it easy to game the system etc… There are a few methods they do to help here.
There are other things they do in terms of biases to flat/hilly/rolling routes, but those are secondary in my opinion.
well, i dont think it’ll look like zr. zr isn’t constrained by being mechanically intertwined with the zwift platform itself. i would assume zrs will place an emphasis on getting people seeded into a roughly appropriate level as fast as it can be feasibly done over pinpoint accuracy, and then keeping users there for at least a short to medium amount of time because it can’t be too volatile. zr is volatile - it corrects itself quickly too, but in it’s current design, ratings can still be manipulated quickly in a very short amount of time. you don’t necessarily see the immediate effect of that with vElo as zr isn’t enforced ingame, but the effects of that would become quickly apparent if it was a core zwift mechanic
This is why I “hope” that Zwift doesn’t spend too much time analyzing older data, and instead starts actually pushing a solution out. Once something is out people will try to break it, and Zwift will need to find a way to adapt it to make sure it’s a good system not just for people who are trying their best, but also a system that is good at minimizing the ability for people to tank their ratings and stay within a lower cat. Otherwise we’ll end up right back where we are.
