This is true, but if the pens can be split by using one strategy, in a set of multiple strategies to choose from, this will work with ranking. Then the racers will meet different opposition in different races and we will avoid the ranking silos we have today in the current categories.
With only one ranking, the Zwift ranking will work very differently than e.g. chess. In chess and many other games where they use ranking, you will ideally have a 50/50 chance of beating an opponent with the same ranking as yourself. The game is the same every time. In Zwift it will depend on the course, and your chances of beating a rider with the same ranking as yourself will be skewed to one side.
1 Like
@OleKristian , agreed. Which is why we should broaden our horizons, leave the world of chess and look more towards systems in use in big field events and not the 1-on-1ās. For example, take a look at something like the FIS system they use in skiing (this is the umpteenth time I link to it over the last 2-3 years ever since someone else in here did the first time, but letās have another go).
Thereās the option of (something similar to) proprietary systems like TrueSkill etc of course, but itās often hard to tell exactly how they work. So if we stick to more āopen sourceā ranking principles for illustration, then I say a FIS-like system introduces an interesting angle, although it works best in a racing world where there is a World Cup and where you are a name, not just another avatar among thousands of avatars that come and go. Still, itās a fresh angle on ranking and itās not at all like ELO.
A problem with the USAC system the way it is implemented in ZP is it weighs heavily on beating riders of higher rank. Thereās nothing wrong with it really exceptā¦ isnāt it a bit backwards? The FIS systems takes the opposite approach. But like the USAC system it too combats rank inflation (I think) since it has an upper limit and since rank changes relate to in-race events, not to completely made up numbers. So itās very different from ELO. I mean, those ELO numbers are in a sense completely arbitrary and have absolutely nothing to do with the game of chess.
In a FIS world the ideal best racer of all has rank 0. So thereās an ultimate top spot you could theoretically hold and everyone would know you were the top dog. Letās say thatās me. You race me, I win again, and youāre behind by 10 sec. Then they use a formula like this to calculate the change in your ranking:
P = ((Tr - Tw) - 1) x F
P: points, i.e. a positive rank change (unlike ELO you want a low rank score here so any addition to your score is unwelcome)
Tr: Racer/your finish time
Tw: Winner finish time
F: Race dependent factor (explained below)
The F factor is a preset number that will vary between types of events. Slalom is slower than Downhill so time separation tends to be bigger. Hence F is smaller for slalom events than for Downhill to equate the two styles the way they affect ranking. You could compare this to races of varying length. Zwift has tons of data on typical finish times for different routes, so in theory you could give each course a unique factor. Either you go for one that will try to equate all routes, or if you would rather want to give extra weight to middle-of-the-road courses (they could be less exploitable, or more desirable for various reasons), then you could do that instead.
The idea behind the FIS formula is that they want to put time gap between you and the winner at the forefront rather than your placing. This makes a lot of sense in alpine skiing since you race solo and also canāt draft. In cycling, bar TTās, this wouldnāt work quite as well. We often have a bunched finish or at least people stay bunched until the sprint, and in the sprint the time separation is very small in relation to total race time even though there may be wide time gaps between groups of riders. So placing does matter to us. Maybe a weighted model that takes both into account? That would reward the sprinter as well as the [I wonāt spoil Strade Bianche 2022 - if you havenāt watched it, then you must].
If you want to go 100% placing priority instead, then you end up with a points system. You award different points to different placings. Like 1st is 10 pts, 2nd is 7 pts or something. And depending on field size the depth of this scoreboard could vary. But you could still tie these points to a ranking just like in FIS. Itās not a choice between points or ranking, you can have both.
Anyway, thereās more to the FIS model. It is a bit complex but thereās no esoteric maths in it or anything, it all hangs together and makes sense. And it seems to work pretty well for them at least. You can read up on it in this pdf, pages 26-33. The link is not FIS, itās the US system which is very similar (the FIS site seemed to be down while I wrote this).
I think there is a typo in that formula. It should be:
P = ((Tr / Tw) - 1) x F
I see that for cross-country skiing, they have two rankings. One for sprint and one for distance.
The factors they use are dependent on the race format and the elevation per km of the course.
Correct! Thanks. Thatās a typo. Itās supposed to be a quotient obviously. And yes, they use the same for both alpine and x-country.
- Pen Enforcement - greatly reducing racing in the wrong/inappropriate category - Almost there.
- All racers counted in any ranking system
- Non finishers all treated as finishing joint last.
Question Do these three steps together go some way towards making race results more legitimate by making them more trustworthy?
.
.
.
- A place to then view those scores, preferably in leaderboard format and somewhere that isnāt ZP
Question. Isnāt in Zwift the very best place for this to be? Surely the use of any other third party is just going to require further registration and acceptance of another privacy policy and all the trouble that brings with it?
.
.
.
Itās just a question of when in my mind. āResultsā as a full body of work comes soon, and the situation in particular that youāre presenting is housed under that body of work.
Weāre workinā on it!
Question Do you have a timeframe you are working towards and if so can you tell us target date?
1 Like
Category enforcement is a huge step forward.
But when you are running in the wrong direction every step forward is a step backwards.
Stop using categories. Start using an elo-rating. Like every other good video-game.
- I want to join a race and not a category.
- My elo-rating (initially based on my ācritical power/FTPā and changed by my race results) should put me automatically into a group of riders with similar rating.
- And the number of groups for a race should be based on the number of registrations.
The actual problems (for cat C,D) i noticed during my category enforcement races:
- Less motivation (less registrations) for riders just above the limit
- Heart rate sandbagger
- Smurf-accounts for racing
- Zpower or inaccurate exercise bikes
- heavy weight riders shouldnāt win mountain/hilly stages
- heavy sprinters should win sprints
- low weight riders should win mountain/hilly stages
Itās way to easy to abuse a system with fixed category limits.
For example:
1 Like
I assume youāve read all the previous discussion on this.
4 Likes
cat enforcement is a massive improvement but more needs to come.
What i would suggest is maybe do some of the series like TFC monday or SZR wednesday with split cats if you are looking for a the fairest level of racing. Organisers need the power next to set boundaries, block zpower users, block no hrm riders etc and get private events on the public calendar so its get more advertising
The weight problem has always been an issue the power floors for each cat i still believe are set too high and need to be on a variable scale as it massively skews when you get to the extremes
1 Like
**Cat enforcement isnāt an improvement. **
It only treats the symptom (sandbaggers) and not the cause (fixed limits).
Iām fine. Iām reaching top3 (zwiftpower) in every race as a heavy weight sprinter.
After a 4 month break my FTP dropped for 40 Watts to cat D. I will be a c-rider soon. But only if iām doing a new FTP-test or compete in an ITT-race.
Otherwise iām able to compete in D-races as long as i wantā¦
But this isnāt right 
I should be moved to C. There i will be kicked in the butt for the rest of my lifeā¦
Or after some races i will be demotivated and quit zwift 
This isnāt right too
1 Like
yep still not right but itās definitely and improvement than having cat A/B/C riders in a d race and getting dropped within 500meters. That does not happen now.
2 Likes
Maybe not, the new system uses all your activities (free rides, workouts, races, ect.) to calculate your category, even if you donāt save the activity or have a crash during a ride, it still counts.
1 Like
I read a lot. On zwift-insider too.
I read a lot of complains and ways to cure the symptoms.
I missed the way to change the system/cause.
My advice for the Zwift-product management:
Open your mind. And stop using categories for races.
Hereās some light reading for you. Please come back once youāve finished: Anti sandbagging and other areas that need development and communication
9 Likes
@ Mike
Theoretically you are right.
Practically free rides, normal workouts or d-races with d-riders wonāt improve my critical power, FTP, 5min or 8min output.
And you donāt need more watts to stay in a d-peloton. Improving means that my heartrate is dropping.
Iām able to show you a zwiftpower-profile of a d-rider (heart-rate sandbagger) who won dozen of d-races and even a c-race. And he is still categorized as D.
You donāt know what his CE category is though (at least I donāt think you do)
1 Like
I donāt need light reading.
I donāt know why you moved my topic too. It doesnāt fit here!
It wasnāt a complain. Neither a question!
It was just an advice/tip of a product manager / test manager with 20 years of experience in the same business. For free. And i also have much experience in elo-ratings, tournaments and professional video-gaming.
I know my english is terrible but i suppose much better than your german.
Much better than my German
I asked for it to be merged because you were starting a conversation that has already been ongoing for a long period of time and I didnāt want that to be duplicated.
The comments that you are making have been made already - which is of course fine for you to have an opinion, but Iād prefer it to be in the same place as all the others, regardless of where your experience may lie.
1 Like
All his races are category enforcement races. So i know his CE quite good.
1 Like
To finish this.
My conclusion: Stop using categories for races.
I donāt know how this fit to āHow is my category calculatedā.
But well. Iām only a customer and itās your playground.