zrCS
How many times does it have to be brought up
zrCS would not fix Race 2 although i agree it could be better.
zrCS also slightly overseeds heavy riders.
In practice, only counting 30s from races would have the same problem.
I set my max ever power sprinting for 51st place in a race of no consequence.
I could do this as the group was rolling around in z2/3 having long since been dropped from the front of the race.
My other biggest sprints would be in points races or ladder races, again where the pace has been low beforehand.
The only way to accurately combine 10m and 30s is from a single 10m time period. For practical purposes something like- what is your best 30s, as part of a 10m effort which exceeds 90% of max…or, you know, just use 5m…
1 Like
If we have to use power data to seed scores, something in the 4-8min of actual data, rather than prediction like zMAP.
Going back through existing power data to at least Jan 1st 2024, so almost 10 months, rather than 90 days.
Seeding score is being badly affected by both the formula and initial rubbish data from those who haven’t properly raced on Zwift since around March.
I don’t see the value of Zwift HQ posting the calculations on how racing score is calculated all this does is gives the gamers all the info they need.
“A” platform uses x method for calculating where a rider races as long as that metric is constantly monitored and improved if required over time that should be sufficient.
Having two or three different methods like CE, ZRS… this ultimately hinders development and riders choosing which system suits them best as in the least amount of effort required for the maximum gain.
4 Likes
No power-based seeding equation is going to be perfect. 10m/30s is awful. 5 minutes is better. For whatever reason ZHQ is married to the awful solution.
Firstly, I wanted to congratulate you for your work on ZRApp! Thank you as I think it was a very interesting and useful exercise 
Sorry to bother you, but I would like to know if it was up for debate (when you thought about the model for ZRApp) to use a type of CS that used some sort of combination of CP (raw & W/kg) & W´ ?
1 Like
You might be right but it’s where we are at and i don’t think they are about to overhaul it so i’m prepared to give it a go and see how rider rank moves them around if it’s only a seeding issue for a few races while riders find their level i don’t think that’s a huge problem.
and they have said they will continue enhance the seed algorithm so the only argument is really do we halt progress until we find the perfect solution that everyone approves or just shove it out there to the masses and they’ll soon see from the feedback what is wrong.
Right now we are in an odd state where riders are picking and choosing ZRS or CE depending what suits them better so it’s questionable how much valid data they are actually getting back from the zrs trial races.
1 Like
Just compared my pure Compound Score (3.043 x 281= 885.083) against my apparent zrCS (736) for the last 90 days, with me having been ~15Kg over my pre- long covid weight at ~95Kg over the summer.
See the comment earlier in this thread where a racer with a ZRS of 350ish finished around a bunch of ~500ers and gained 4 points.
1 Like
Can you give the event link for this?
Terrible idea. People who are not happy with their ZRS will almost certainly respond at a higher rate than those who are happy, severely biasing the results. He will learn nothing meaningful from such a survey.
4 Likes
1 Like
I’ll take this a step further. If the overarching goal of this analytic exercise is to produce races where more people are engaged in the fun, why not predict the “time delta” in races? If we all consider races where most riders are still together in the last km more fun (I certainly think it is), let’s set that as the goal post.
ZRS Seed = (CP600 Watts * CP600Wkg) / 2.5 * (65 / (65 - (0.5 * (65 - Rider kg)))) + (((CP30 Wkg) - (8)) * 20)
You can copy the gsheet and test yourself
I used CP600 and CP30 just because the original ZHQ formula is using…IMO 5min is far more superior than 10min
EDIT: This is formula proposal and not the original ZHQ Formula
what is ZRS with cm?
It’s using Height to adjust ZRS if we would go that far
are you saying the formula in the sheet is something close to zwift ZRS?
No it’s not…i’m saying this is how it should be for better SEED calculation.
1 Like
While this is a true statement, it also signals the point at which Zwift’s roomful of PhDs failed Statistics 101 by assuming that correlation implies causation. I’ll grant that in a pack finish the people with the highest 30s power are most likely to win. But in that scenario you’ve limited your analysis to the people in that lead pack. I’ll wager that pre-race analysis of 30s power yields no statistical ability to predict finishing position for the entire field. But doing that same pre-race analysis using 10m power is going to be much more predictive of general finishing positions. The 30s power metric is a SECONDARY predictor, not a PRIMARY predictor. As such it has no place in the seeding formula.
4 Likes