That is one way to think about it. I would think there will be races where you can ride a higher cat.
I prefer to Train hard to race smart! 
“I just made that up”
4 Likes
autocat by speed sounds interesting
but my speed is fast for 20 minutes and slow for 40 minute races.
I would need to be one level for a short race and another for longer events.
After an hour, I crawl lol.
2 Likes
Great to see some positive development in this area. A great pity you chose the Wednesday for the Down Under event as this clashes with our 5 year regular Aussie Hump Day Ride event which attracts over 300 riders each week. Can you change this to a Tuesday? Or add another event please?
Excellent to see development and testing!
I’m curious to hear the reasoning behind using this finish time per difficulty scheme — what are we gonna call it, anyway? — instead of pure race ranking, which benefits/tradeoffs does it involve? I assume there’s something that’s weighing the decision this way since race ranking already exists, and as such could be tested equally well.
My main concern is that it does not accurately gauge people against one another if they haven’t raced in the same race multiple times, which is what Elo systems attempt to solve, so I’m curious what the additional benefit is seen to be here.
I don’t think to aim is to accurately (100%) set people against each other I think the idea is to set up matched competitive groups. If you are a edge case you will somedays be at the top and other times at the bottom but always competitive.
2 Likes
One of the benefits is that it takes the difficulty into consideration. One single Elo rating will be based on all performances with the assumption that the difficulty is equal in all races.
It would be interesting to know what goes into the difficulty calculation.
As I understand this, our category is calculated based on previous races with the same difficulty. You are writing that we will be automatically be moved up or down if we perform outside our assigned class. Will that only apply for races with the same difficulty? Will there be one racer category per course difficulty?
Elo is a rating system that gives each person a rating such that you can compute the relative probability of who will win when comparing two (or more) people.
When two people compete, the winner and loser exchange ranking points based on the relative difference between them. Winning against a much lower ranked competitor gets you a small number of points, winning against a much higher ranked competitor gets you a large amount of points.
Generally for multi-person events you need to extend the 1 on 1 to be 1 on n exchanges.
For example, starting with the winner, he can exchange points with all other competitors that finished behind him. Then repeat for second place, exchange with everyone behind him. Do that for everyone.
Again, the points exchange between any two competitors is based on their relative rating. As you get a higher and higher rating, you rate of increase will slow down as you are mostly winning against people that you are better than. So the points you get will be small. Effectively the only way to increase to a higher rating is to win against people close to your rating or higher.
Elo used for this does not address performance differences based on different types of events. Which can be a problem. If I have attained a high rating based on my TT skills, do I want to take a hit by doing a hill climb?
4 Likes
What if it didn’t matter because you had one rating for TT and one for hill?
I would be happy to see a possibility to race above category especially if you plan to have an overall league and I might be at risk to be upgraded during this goes on. Same for woman being able to join men / mixed events.
I will have a try on wednesday. great to see that theres a little bit of improvement on zwift racing.
1 Like
Elo does account for difficulty — against the people in that same race (and thereby by definition as difficult both in terms of course and the racing itself), and then through the network effect against their opponents in other races of other difficulties.
But yes, I assume they’re trying to introduce some kind of a modifier to it, and I’m curious what the sought effect is, and which kinds of races it’s been projected on, if they see it work better for some situations than others, etc. The tradeoff here is at best a more complicated categorization algorithm (granted Elos are already something you don’t want to do in your head so might not matter), but at worst it could result in worse cats than pure race ranking. Or just become unmaintainable. This testing is part of due diligence, and I’m sure they’ve puzzled much of this out beforehand, so I’m curious about the conclusions 
1 Like
Yes, that is true, but the rating will only be correct if you are doing only one type of race.
If you are doing mostly crits, your rating will be of little use in an endurance race.
Correct, and if you only do endurances races then won’t be valid for crits. If you only do short Time Trials not valid for … (fill in here.)
IRL and even at the pro level this is true. Just watching TDF and we see two stage winner Cavendish barely making the cut off today on an Alpine stage. And we often see TDF GC changes based on Time Trial performance.
Effectively, ELO ranks you for the races you do. Presumably if you are doing those races those races are how you want to be ranked. And remember that Elo is (or attemps to be) a predictive tool on how well you’ll do in races. It is only as good as the data you feed it. And (effectively) mostly accurate for races of those types.
Since it is most reliable for the races you mostly do this is probably an OK result. By definition it is less accurate for the races you (typically) don’t do. But since those are the races you don’t (typically) do I’m again not sure that it is a problem.
6 Likes
This means that the speed/difficulty category system will do a better job than a single Elo rating of placing you in a category where you will be competitive in most races. Regardless of type of race.
The system should work equally well for all types of users. Putting limitations on what types of races one can expect to be competitive in, may feel natural for experienced racers, but not for users that are new to cycling and Zwift. Racing should be fun and users should feel competitive.
1 Like
Sure. As mentioned before I am concerned about the 30 days though. The more data you feed a system like this, the better. How many Zwift racers will have raced every course difficulty type once in the last 30 days, let alone multiple times? I’m guessing practically 0.
4 Likes
Yes, they must store the user’s last calculated “score” for each difficulty, and use that if there are no data for the past 30 days.
1 Like
Presumably they could use some sort of convex weighting system to give more weight to recent events, but to still consider older events.
1 Like
Has anyone been able to press the “autocat me” for this week’s tests??? I haven’t, so I checked zwiftpower to see who is in each cat, and see no riders listed.
Hey Zwift, do you have enough Women for this test? If not, I can try to join.
1 Like