ah ok, now i get a clear view of the new feature. thanks for explanation martin👍
But ZP uses a 90-day result timeframe for categorization, does this scheme actually use even older results rather than just the most recent 30/60/90 days?
Anyway, I would suggest that an even longer timeframe (say at least 180 days) would be useful if more recent results are nonexistent or infrequent and not in line with past performance. If races catergorized with this system were to become the norm, you’d have to hope everybody taking the summer off Zwift does a conventionally categorized race or two first or else you’ll have returning racers trickling back via the lowest category all autumn (especially if you can’t choose a more demanding category yourself…).
Thank you, Martin, Steve, and Flint. Excited to see how this plays out. I’m a Cat B at ZwiftPower. I just signed up, using the instructions given above. I’ve been assigned to “WTRL AutoCat Test - C6 - Middle Earth”. Joining me in the C6 group is a ZwiftPower Category A+ racer currently ranked, among all racers at ZwiftPower, in the top 20 worldwide. Is this anticipated? Is this all part of the process, a mix of racers set-up to shake out discrepancies?
Thank you, Martin, for the email just now. I followed the instructions therein; the system recategorized me into C4. Thanks for the speedy work.
Thanks. It didn’t say this originally though! Anyway, all good!
Another change is that is doesn’t say Elo anymore, instead is says it references TrueSkill.
Its good to get away from Elo, and generally we are talking about a basic concept, rather than the Elo system itself which isn’t really suitable for Zwift racing. I don’t know a lot about TrueSkill, but there’s some reading here.
They haven’t updated the second paragraph of the OP bullet points so don’t ask for to much
I’m guessing that many people ride the Thursday TTT or the Tuesday ZRL so their times for certain courses would be way faster than they would set if they were racing them by themselves. I desperately hang onto the pack in races so probably have reasonably good times but it doesn’t really tell a story of my ability. Outwith these events I don’t ride a large variety of courses as I would do training sessions on 1 route whilst watching netflix.
Very curious to see how this all works and great to see you trying something new. Hope it works out and I’ve signed up for Wednesday. Really hoping C1 is the equivalent of one of the better C class groups and not Class 1.
ELO was developed for Chess and has been adapted for tennis and various other player v player sports or player v player within a team. It doesnt work very well when you have more than 2 sides in a single match and the complexities of unknown players mixed in. Many have tried to adapt ELO for this with some success but when you have over 200,000 players of varying abilities, ELO falls very short.
Agreed, there’s a big thread on here discussing it.
Trueskill is a better solution. It is owned and patented by Microsoft.
Making the assumption that you are willing to license it, Trueskill is the way to go.
Overall it is similar to Elo, but better developed WRT to multiple players competing in a single event. Which is what a bike race is.
What ZP does is up to ZP. If our Autocat system cannot find results in the last 30, 60 or 90 days for racing events only (not group rides) it will look even further back (which covers those that leave for the summer). It is also only the TEST events that do not allow you to select a higher class. While it isn’t a perfect system yet, we have been at this for many months and do keep up with the forums and feedback and are in daily contact with various devs - there wont be many things that have been missed and nothing has been mentioned so far that wasn’t in the original list of Zwifter gripes we are aiming to eradicate.
The Microsoft website is a little more informative than Wiki
No, this autocat I’ve been given is a total lolz. I am fairly certain it’s going to be an absolute cakewalk for me. See for yourself
Anyone here seriously think my race spot is correct? I’m in cat 6 for the aussie timeslot
Zwift Power - Login
Edit - I see there was a bug, check my post below
I agree. There are obviously a lot of bugs in the system as currently implemented. My C3 race looks reasonable but there are some absolute ringers in the C5 and C6 races across the time zones, including people who regularly beat me in races.
Ahhh I see now, there was a bug when I signed up. I’ll try again.
Well that’s more like it. C1 now.
Looks like a good option, especially with the focus on ranking newer players (riders) as quickly as possible.
If you are actually looking beyond 90 days if needed, that’s great. You might want to consider revising the wording “Last 30, 60 and 90 days Racing Stats” on the Zwift Classics - WTRL page accordingly, though…
For these test events it doesn’t go beyond the 90 days as there is little to gain for either the race, the racer or the test itself so what the website says is correct for the test events.
I’m not sure if anything changed since these comments, but it looks to me as if the ‘flyers’ (obviously unintentional) are listed in two cats/events on ZP. One that appears reasonable and one that clearly isn’t. Not sure why and if this causes problems later on, but I suspect WTRL is aware and on it.
As a general comment: I’m very pleased to see this development and hope it works out well. It doesn’t need to be perfect from the start. I’m sure these test events will help evolve and fine tune the cat boundaries and highlight obstacles so that ultimately every cat and any course has good fair racing with (more importantly IMO) a natural progression from one cat to the next. I have full confidence that WTRL in collaboration with Zwift will develop this further. Great work.