I think the current system doesn’t work either way. I only did a few races so far, always in Cat D and I only once made the Top 10 (in a race with 12 starters…). Now I improved my ftp to ~2,5wkg and wanted to test that. So I joined a Cat D race. Right in the beginning it told me I might be flagged, which I already found odd. Then the race started (Crit City Dolphin) and like always everyone was going like crazy in the first round. I was able to barely hold on to the first group (7 riders). Then in the second round I got dropped and made a (for me) super big effort to catch up and just as I did catch up at the end of the hill I got flagged. So basically Zwift flagged me based on my 5m power, which is stupid. In the end I had a normalized power of 2,5wkg overall and last place of all finishers. So imo the flagging was clearly wrong. Also if the flagging would mean that they would just adjust my power accordingly this might make sense, but I found out that by doing ~3wkg I still got dropped by people doing 2wkg on a flat, that seems to be way too much penalty for someone who mathematically is <1% over the category limit.
On the other hand there were some people doing 3wkg+ more or less all the time who didn’t get flagged. And then some really odd rider doing 10s of 18(!) wkg and then 10s of 1wkg pretty much the whole race. I mean 18wkg… I didn’t see that even in the virtual TdF.
I have a feature request for an alternate approach to be used for race categories:
- it should be based on the full power-duration curve, not just 20 minute power. Limiting 20 minute power/mass only basically kills the advantage of those with more endurance, while preserving an advantage of those with punchy power (at least in categories lower than A).
- it should be based on a combination of speed on the flats and climbs, not just W/kg. Using W/kg hurts light riders, since they’re at a disadvantage at any speed where wind resistance is significant, which is the vast majority of the time.
Basically the present category system is based on how quickly riders can climb the Innsbruck climb. So those for whom that is a strength, except in A’s, have no advantage: they’ll be promoted to a category where it becomes the norm, and not the exception.
Dan
imo any kind of power based approach to this problem is pretty weird. It’s really hard to get a system that will work for all rider and race types. The current one obviously is especially bad since it uses power estimations and is both easy to cheat and very trigger happy for false positives.
In most games and sports I ever took part in categorization happens based on results, not based on power or “skill”. Just look at any competitive video game, they invest a lot of research into fair and fun league systems and it’s working pretty well.
Another big no go of the Zwift system is that they actually let people join any category and then apply penalties at some point in the race (which also seems to be flexible). Everyone who was in a race where everyone blew up in the first few minutes only to see that the leader gets flagged after the first two rounds knows that it’s not fun. Also taking part in a race and getting flagged after half the race sucks.
The whole idea of categories in Zwift is a bit strange. In the Discovery rides and the grand fondos, things naturally split up into groups of riders of similar abilities. You do the best you can (if you choose to “race” them). It seems to work fine. Being best in D’s typically just means you should be racing C’s (if not B’s).
But that aside, a good category system allows some riders to excel on some course, other riders on other courses. But a lightweight endurance-oriented climber excels on nothing. Because you’re matched up against riders of similar 20 minute W/kg, you’re the same on the long climbs. And you are better than them at nothing else.
In contrast, a sprinter may have A-level fitness, but with a personal weakness in 20 minute W/kg, gets put in the B’s or C’s. They can now keep up with the climbers in that lower category, then smash them in the sprints, or on the short intense hills.
A scheme which balances different types of fitness (W vs W/kg, short-term and long-term power) will allow each rider to excel in situations favoring their strength: sprinters in sprints, climbers on climbs, rolleurs on short punchy efforts.
That’s why I think categorizing riders based on their results would produce a much better system. It’s more in RL where if you are in the UCI pro tour you have to race against the other pros, but no one will stop you or replace your bike with a skateboard if you are producing too much power.
The same way that no game will try to put you in a different league because they measured that your reaction time is too fast or your player level is too high.
But while we are at it does anyone know how the algorithm actually works? Does it take your history into account, or just what you set as your current ftp? And then what do they calculate during the race? Because some people are getting flagged really fast while for others it seems to take a while. So they must be extrapolating your 20m power from 2-5m power, which is just not possible.
Basically, if you exceed the limit both for W/kg and for power, averaged over 3 consecutive races, you get upgraded to the appropriate category.
If you exceed the limits for a single event by more than 0.1 W/kg, you get excluded from the results for that race.
But it’s based on 20 minute power. For races less than 20 minutes, they adjust based on a model power curve.
Bu that’s specific to Zwiftpower. I think Zwift itself uses something else. For Zwiftpower it makes sense because they analyse the final data, so they have a proper 20m power to use. The Zwift algorithm must do something else to estimate 20m power. Or they also have other limitations besides the wkg limit that is shown next to the category.
Both are essentially based on 20min wkg output (ZP just has an additional condition based on absolute power to accommodate the fact that lighter riders have a hard time keeping up on the flats, even though they have a high wkg output)
Zwift itself doesn’t assign categories, it just hopes you 1) understand the concept of FTP and how it applies in general and not just to your expected power output in any given race, and 2) voluntarily categorise accordingly.
Not sure what you mean by algorithm, but the only instant where Zwift categorises riders (sort of), is during the Zwift HQ Crit City beta races (which take place 2-3 times per day). Even there you can still sort into the wrong category, but you will get coned (and artificially slowed down) if during that race you exceed some power threshold, the level of which is not disclosed (but from what I understand there is both a 5 min threshold and a 20 min threshold). But to my understanding, they don’t use historical power output… (strangely…)
Yes, I was talking about the beta feature. I wasn’t aware that this only applies to the Crit City beta races though. I just participated in one by chance it seems. 
So this feedback goes more to the beta test…
I wonder why they simply didn’t copy the ZP.com system, and imposed that, using historical performance available to them, to avoid riders would even be able to sign up for the wrong category…
In case you haven’t seen it, a very significant announcement was just made in the “Whats going on with Zwiftpower” thread:
shooj:
Our goal is to make ongoing improvements to make competition on Zwift fun and fair. In addition to the testing of live flagging in Zwift HQ Beta Races, we are also exploring messaging at event sign up to guide Zwifters to the appropriate race category. The long term vision is a categorization system based on the Zwifter’s historical performance.
To me, this means the w/kg categories will disappear and be replaced by a system of cats based on actual performance (presumably times in actual races). The sooner the better!
Zwift could and should do much more. If someone is racing on zwift without a powermeter, they should not be included in the results. Period. I’m a fairly strong C class cyclist and I get left for dead by cheaters all the time. I doesn’t bother me greatly as I’m interested in my performance, but it does leave me wondering what possible pleasure the cheaters get from ‘winning’ virtual bike races in this way. I can only assume there is a special class of nob participating in zwift who have no sense of self worth.
Unfortunately, I think it means they will incorporate the zwiftpowerdotcom system, which uses historical 20min power output to assign categories…
I wouldn’t hope for a result based classification any time soon.
anti sansbagging for child is reallly a mistake…
My child that have its account, get a big w/kg… but its ftp is poor… so does it need to race with B catagory ? No it can’t… but anti sandbaging ask it to do that…
They should also tak into account that poeple can have less kg than other… i really does not understand this classification based only on w/kg… why not ftp ?