The score floor being 85% of seed score allows for high power readings producing the 5min number, due to a lack of calibration and/or turbo playing up. Using a single instance to produce seed score has the danger that it isn’t a true value.
What needs to change at the very least includes…
Minimum number of finishers 5+ to get results based score changes (not power) rather than current of 2+
Large fields giving larger score changes than the current issue of the opposite, really important change
Multiple runnings of the same TT processed for result score changes once at end of week, to compare efforts from for example all runnings of the TT club, often with small fields per run. I’d like TT events like this to run hourly
The ~10% top finishers of a race should have their result score change and podium bonus added to their seed score for a limited time (eg. 6 weeks)
Result changes need to be based on finishers, possibly DNFs that complete 50%+ of the course, but not on number of signups
Replace score floor 85% seed score with the same score value below seed score for all pens? Currently, an A pen racer can drop further than a pen D racer.
I’m not sure if the weight factor adjustment to raw Compound Score is working well with the physics used in Zwift.
Really, we need a complete revamp to a vELO system that rewards beating racers you aren’t expected to beat and vice versa.
Zwift is not replying to any comments. As such, there seems to be no point. Instead if anything changes its a change that won’t work and easily could have been known if they would read information embedded in this topic. It’s a skill.
If you have a bad reading one time you can contact Zwift support to have them remove that instance from your record. If it happens multiple times you need to sort out your equipment.
The 85% seed score floor also allows for some performance drop over the current 90 days (which I still think should be backdated at ~12 months for racers with history), due to things like an injury or recovering after an illness.
But this was one thing that was widely applauded when we went from CE to ZRS. Sandbagging would not be an issue anymore since people had to lose these races.
Well I’m sick of this poorly implemented non-ELO system and after ~8 months of pointing out flaws and issues, I’m not renewing my sub when it expires later this month, until I see some significant improvements to this bonkers version of Racing Score.
I find it so arbitrary and opaque, I am shaking my head after each race trying to conceptualize how the heck it calculated. I mean, come on - this could be simple algorithms to create something that makes sense and is published to us all - transparently!
From what I’ve read, the main chess ranking system run by the global governing body is a relatively straightforward vELO system, which has retained popularity despite theoretically better methods now existing. The popularity derives from the methodology being so well understood that players know what their ratings are going to be post-match without needing the official ratings to be published. The “random-number” aspect of some ZRS changes is a big downside re gaining acceptance. (The other one being there being no link between ZRS changes and the ZRS of who you beat / who beats you.)
Whereas, ZRS based on Trueskill which requires an understanding of Bayesian maths with tau, beta & sigma etc hits the spot just right for the knowledge level of the average zwift racer.
I think the vital words missing from that sentence are on the lines of “to code properly/quickly.”
It makes no sense at all that Racing Score simply gives you some position points and/or podium points, for finishing in the top ~55%, regardless of how strong/weak the racers were that you beat relative to yourself.
It makes no sense at all that simply finishing in the bottom ~45% reduces your Racing Score, regardless of how relatively strong/weak the racers were that finished ahead of you.
It’s confusing as heck that signups who don’t even turn a pedal in the pen reduce scores dished out after the race.
It’s confusing cubed that smaller fields give bigger score swings than larger fields, especially if it’s true that this holds true all the way down to one racer signing up and one racer finishing. We’ve certainly seen evidence of two finishers giving wild score changes.
Actually if it were as “simple” as understanding TrueSkill/OpenSkill that would be not as bad as it actually is.
As it is, it’s really hard to explain some of the movements we are seeing with a vanilla implementation of OpenSkill (such as more points for smaller fields, people losing points when they compete up a category, etc), so without them telling us how they modified OpenSkill it’s actually almost impossible to know what is expected.
What would be great is if they would engage in some of these conversations where people see counterintuitive movements and help explain what is, and is not expected.
In my experience this isn’t always true, but the difference between the scores has to be quite high.
My best guess on why it’s like this is that they’ve used one number for all races without any sort of terrain modifier.
If you take a heavyweight sprinter and a lightweight climber, if they have a similar rating then on a flat course the sprinter is practically guaranteed to win, and on a climb it’s going to go the other way.
You extrapolate that, there is some difference in rating where a sprinter will still likely beat the climber on a flat course, and go the other way the climber will drop the higher rated sprinter. So the parameters in the OpenSkill implementation have been set to account for this.
This means that the difference between scores before a result becomes “unexpected” is quite high. Because most results are within the expected range the movement becomes about where you finish rather than who you beat.
It’s clear that there’s a generalized disdain for the users over at ZHQ, the question is why they recently decided to completely avoid any and all interaction with ongoing changes. PD4 and the first ZRS announcement they actually engaged and it made for better outcomes. Now we have @Jon saying that he doesn’t want to talk to us because we don’t know the game and are going to hallucinate changes if there’s ZHQ/community engagement. Even though it’s clear that the engaged users here know Zwift far, far better than anyone at ZHQ because no one at ZHQ is a serious Zwift user.
Doesn’t take a genius to work out that the system is badly flawed when a rider who comes 3rd in a field of only 5 riders (time trials are notorious for this) gets score increase of 10 points , exactly the same as a rider who comes 3rd out of a field of 50 in a Tiny race.
I thought they were supposed to be sorting this out. My experience yesterday with this proves that they are not.
‘Minimum effort required’ update from November 2024 clearly inadequate too, as people are still tanking races left, right and centre to keep their scores low enough to stay in Cat. And good race results alone don’t seem to be increasing riders’ floor scores, only new 5 minute efforts do that .Zwift HQ clueless, of course.