It is recalculating using the updated method as we speak. The two fixes I detailed above will no doubt have fixed what you had seen to this point.
1 Like
āMeets expectationsā means if you are ranked 24th of the finishers and finished 24th, your ranking was pretty accurate so doesnāt need to change significantly.
If you are ranked first and finish first, you will still gain rank, but it will be minimal (depending, of course, on the ranks of the riders you have beaten)
1 Like
Iām not optimistic, but Iāll keep presenting them with the opportunity.
4 Likes
Thanks for the update. I was one of those that had a ridiculously high ranking for my ability (racing all those zwift monthly leagues) and I can see that I am now placed close to similarly performing riders. Great work!
1 Like
Just visited site and was about to report here something funky was going on with rank scores, then saw this thread!
One of the natural consequences of some riders having different CE, Zwiftpower and Zwiftranking pen allocation was some Zwiftranking scores were changing relatively wildly, will be interesting to see what happens once processing has finished in the score recalculation of past results.
How far back in time is being used to determine Compound Score?
Only using race data to get this figure?
But at end of the day, Compound Score is simply a starting point, any ELO methodology rank system will adjust riderās rank scores, the more frequent races they do, the quicker the adjustments will happen.
90 days
Yes, no access to other activities, but this would be a big benefit if Zwift took the same approach.
Exactly.
Iām pretty excited by these changes as I watch them process. It looks far more accurate, and it is now relatively simple to look at future races and see what āparā is.
2 Likes
If they donāt eventually find tougher fields, thatās true
The system is still working away to catch up from the reset last night. Thatās why you donāt see any rating changes for October/November just yet! 
2 Likes
Couple of small tweaks, so the reprocessing is starting againā¦
1 Like
Might it be worth pointing out that the scale of rating points has reduced so anyone seeing a big drop in their rating is aware?
Whopping drop for me; C-, must try harder 
Hahaha the preview 
4 Likes
That preview is quality. Bravo.
1 Like
Thatās freaking gold
1 Like
Just remembered something I spotted at https://www.zwiftracing.app/ and forgot to postā¦
When looking at any riderās profile, there is a symbol for each race denoting flat, rolling and hilly route.
However, it looks like the symbol is based on doing a complete lap of a route, the type of route actually experienced in short races like say the custom length approx sub 7Km Tiny Races can be very different.
At the mo it might simply be a symbol, but going forward we might get rank scores for different route types.
Just a symbol, but you have exposed 1 of the (many) issues with route profile based rankings though.
Pretty slick stuff. Nice work! Interesting to see how the Elo-MMR model works with Zwift.
Curious what your thoughts are what you like with this system versus something like the ZwiftPower ranking system. Anything you find lacking with it?
I admire your optimism that Zwift will take any notice whatsoever of the fine work that youāre doing here on this.
4 Likes
Iād like to think this was thoroughly covered in the Zwift Insider article, but the Zwift Power approach and the ZwiftRacing approach are fundamentally different.
ZwiftPower is an achievement based system. How high can you get in the ranking with your best 5 results, against the best racers possible. This means that the very (very) top of the rankings are pretty good to be fair. The best of the best work their way to the top. However this is a terrrrrrible approach for placing riders in to pens:
-
The vast majority of riders have inaccurate rankings, almost always too low
-
The distribution of rank is heavily stacked towards the low end, so any races towards this end of things would feature racers with vastly different capabilities. This wouldnāt resolve itself over time, so whilst the race experience may be decent for the top end, it would be terrible for the majority.
-
The entry point for new and returning riders is the bottom. What about those riders that are genuinely low ranked? Must they forever race against far stronger riders?
-
It only suits regular riders. 5 top performances is a lot to get even reasonable level of accuracy - and itās not that accurate for most, because the rankings of those you are racing is also unlikely to be robust.
Contrast this with the ZwiftRankings approach:
-
Industry standard algorithm already fundamentally tried and tested in many-v-many online games
-
Immediate suitable entry point based on Compound Score
-
K (confidence) factor to quickly rank riders accurately in their first few races
-
Always something to play for in any race
-
Ranking distribution that mirrors ability distribution
-
Suitable to define pens for enjoyable racing (already demonstrated, more trial races to come)
-
Clear and understandable progression system
-
ZwiftRacing rankings work and are accurate regardless of how pens are split (does not rely on pens split by ranking)
Thatās just a few off the top off my head post ZRL, but they are chalk and cheese.
You could split pens by ZwiftRacing rankings tomorrow and create some awesome race experiences. If you split pens by Zwift Power Ranking it would be an absolute horror show.
P.S. I really hope that comment does not mean you are still considering the Zwift Power approach.
Reach out to Tim, allow it to continue to develop for a period with all of the community testing that is ongoing then youāve got an oven-ready system thatās easy to implement.
8 Likes
Itās thoroughly covered, yes, and some time has also passed. So I was wondering how your opinion had changed over that time, especially given new improvements.
I like the ZwiftPower system for a good few reasons, and I like this system for a good few reasons as well. One thing I can definitely tell you is that I donāt think weād be successful at all making our own system. I also donāt think any one system is the best or more right than another.
Not looking for a silver bullet approach here, as I donāt think one exists. All I want is racing to feel fun and for there to be something at stake when you show up to the start line so you feel motivated to race, regardless of your ability.
Edit:
I do want to add Iām also not trying to sway opinion one way or another nor say that the ZwiftRacing site isnāt welcome in terms of something the community needs, is a bad method, etcā¦ Itās great to see the conversation around it and to compare that to our own parallel work of scoring and look for commonalities, differences, and overall opinion. Looking to have the honest chats here because Iām genuinely interested; when I say well done, I mean it.
8 Likes
Agreed there is no silver bullet.
What do you think the purpose of a ranking system should be?
I think this question is important and can help me frame the ZwiftRacing approach correctly.
Fun racing (for all abilities), but also addictive racing, where you feel there is always something to race for even if you are not the strongest rider in the pen.