The GC is E only. By that I mean that everyone result is converted to E and there is an overall GC showing everyone that has taken part.
We’ll subsequently be adding a way to filter the result based on racing score.
2 Likes
It would be better if the longer ZWS races were set up with all categories visible. That way, I could ride with others even once I am dropped by my assigned category…which will probably not take very long 
Perhaps @James_Zwift could try a test run on the Zracing series with staggered starts on a non looped course to see what the reaction is to these?
Still run a chance people will ask stronger riders from the Cat above to sit up & then pull the group to a quicker time though is that such an issue with the power ranges in ZRS cats these days.
IMO - I believe Zwift is trying to tie up what your race score was at end of last race (if you had one, otherwise seed score) against what your score is at end of current race. Based on that belief:
Rider 1 is due to decay as they have not raced ZRS race since 13th July. So old race score of 342 less decay (unknown amount) plus race win points (unknown amount) to give new score of 337. Therefore a loss of 5 over 2 months so it has to be a down arrow.
Rider 12 has not done a ZRS race. They came 12th out of 25 racers so top half. I would expect them to gain at least a point so up arrow.
Probably nothing to do with the power either put out in this race.
No actual evidence or proof to support this belief just observation of many similar results.
Yeah, I kinda understand why their data would show that 30s power has a strong correlation to winning in addition to 10min power (because anyone who makes it to the line with a high 30s power is most likely to beat anyone without a high 30s power), but this is the value that will have the most variability in terms of who’s actually done a true rested max effort through normal zwifting.
I did my first race of the season yesterday, and I tried my best to push some power at the line, but all I had was 424W for 30sec. Looking at my power curve for the year, if I were to go do a singular max 30s effort without being tired from a race it would be well over 700W for 30s. So, if I go do that outside of a race as a true max effort I will then put myself in a much higher base score using a power that I’ll never come close to achieving in a race.
There’s a workout on Zwift that requires 30s max efforts I did a few times last year (The Swift), if I do that workout this year my seed will go up a lot I imagine, if I don’t it will stay low because I’ll never have good numbers at the end of a race compared to a true max effort.
I do like compound score instead of 30s+10min power because I believe there would be less variability in max 5min-ish efforts, and there’s a lot of opportunities in Zwift races to hit really hard 5mins fairly similar to max 5min efforts outside of races as well. I imagine it would be less variable than people’s 30s max efforts.
5 Likes
Thanks for that Ian. That makes a lot of sense, and explains the arrows pretty well.
It does still leave me with the feeling that although Rider 12 finished top half, all the riders above him look to have had a lower ZRS (they certainly did after the race), so I would still have expected his ZRS to drop, because he has, in ZRS terms, been beaten by a bunch of weaker riders. But maybe that’s a fault of my expectation, and I’m certainly not trying to argue with you.
I also would have thought that for Rider 1, the arrow direction would be based on his ZRS on race entry (or race start), not on previous ZRS race ridden. The system would ‘know’ it, surely, as it would determine his eligibility for entry, so could use it. But again, maybe that’s wrong thinking by me.
For race score, is the difference between seed score and actual score considered race craft ability?
Ce didn’t take under 5min into account after a certain time despite a big 3min effort lowering zftp etc the fact they are now using 30s and 10m to calculate when they have someone’s power curve is a bit strange but lower time values absolutely need to be taken into account but, and here’s the caveat, it needs to be race data not from pace partners or free rides which has been my bug bare all the time for ce data also
2 Likes
Unlike apparently most people in this conversation, I’ve quite enjoyed racing with Race Score so far.
I do wonder if something like a the Compound Score mentioned in this article is something that improves the racing score system.
In keeping with our hypothesis that both a high absolute power output as well as a high relative power output are important in determining performance, we have demonstrated that the product of these two variables has a greater correlation with and is able to predict a successful race outcome to a greater extent than either variable alone.
(PDF) The Compound Score in elite road cycling.
Available from: researchgate_net (Leo, Peter & Spragg, James & Wakefield, John & Swart, Jeroen. (2022). The Compound Score in elite road cycling. )
For example;
- Rider A: 60kg, 5 min max power 330W (5 w/kg) = compound score 1650
- Rider B: 80kg, 5 min max power 380w (4,75 w/kg) = compound score 1805
So although rider B has a lower w/kg then ride A, they have a higher compound score and is expected to perform better at one day races. Obviously this could be calculated for different time durations.
They are definitely aware of compound score because Tim who worked on ZwiftRacingApp vELO used compound score there (then later a modified version that was even better at predicting results in Zwift across various weight ranges). Tim is helping Zwift out with ZRS.
They landed on the 30s and 10min power after a lot of data analysis on Zwift races, so it was definitely not a random decision. That said I personally liked the model with modified compound score as I believe it’s less variable and less gameable than the combination of 10min and 30s power in and out of races.
I genuinely dont think they considered the impact of Zftp & Zmap on users power curves when they choose those time periods - Resulting in mixed abilities with different zMap & zFTP levels as A/B/C/D racers all mixed into a single pen.
But they missed that people were already sorted into otherwise-similar ability levels by zMAP/zFTP and so of course 30s (which is all that really counts in their equation) is going to be the dominant factor.
1 Like
Changes are all great and I appreciate the list of “next steps” as well.
One issue that’s worth thinking about is what “winning” means in Zwift races. I think the primary reason we have an issue of sandbagging in races is that “winning” means crossing the line first.
It we alter the incentive from crossing the line first to “crossing the line ahead of better riders”, there wouldn’t be any point in crossing the line first ahead of people with lower or same rank as you.
I would be curious to see how riders would behave if race results were displayed as a ranking of score gain instead of finish position. My hypothesis is that when people are rewarded for gaining score instead of finishing high, they will be more likely to put in an effort close to their best. This might address the sandbagging problem.
This can be set up as a simple A/B test for the ZRS test events. Something to think about for future testing.
I wonder if they also analyzed the 10min/30s data from mass start events, etc, or not. If they only analyzed races already categorized by zMAP/zFTP then yeah, this would be problematic.
2 Likes
Thanks for all of the discussion here! We have started a new thread, so I am going to close this one down. Please see Zwift Racing Score - Next Phase! for more information as we move Zwift Racing Score out of the Zwift Labs umbrella.