Neither did I lol.
People “overwhelmingly” asked for results based ranking. Zwift started implementing it.
1 Like
Blockquote
This topic will automatically close in 13 days.
This topic will automatically close in 13 days.
Blockquote
What happens in 13 days?
Will we all get put in different cats? separating great B’s from lousy B’s? or is it great B’s race against lousy A’s?
There is a reason why no other games use a system that rank you and lock you in the rank. You are creating a situation where riders race till they can only fail. The better you perform with a one off good effort the worst you will do in the future.
Other games realize the players are the asset, no one want to play against bots. But if the highest volume racers are the ones that actually have the most realistic data, that have to compete with new racers with skewed data for not having enough races, those high volume riders will find new platforms.
Lots of stuff on points ranking etc etc and tried o look hundreds of posts but: 90 days? Let’s do 6 months or similar?
To get downgraded in IRL usac take some time same season? I mean if out of shape we still punch similar or close in shorter efforts and depending on the days etc compared to someone firmly in the next category down? A not really in shape X caegory rider is a very different animal even power wise than a well trained X-1 category rider.
I mean I’m mid B and vary between racing B’s (expecially if hilly) and A’s when racing, but at some pointwas a little injured and rode less building base back says I’m a C. I tried a C race ONCE to see (?) and confirmed between power and experience I could ride and then dominate…and then they cut me from the resulst saying I was a B! I almost didn’t win someone had a LOT of game tired himself attacking repeatedly but last one I waited a wee bit long for others to do some work before realizing oh right nada and launched…did blast past him but almost waited too long.
“It” says I’m a C again because guess what I’m not riding hard now and am 9 kg up (!) and it’s only 90 days rebuilding my base strained something a YEAR ago in 13 stage hisp tour then life got in the way BUT…I can do B numbers out of shape a goodly amount and have a LOT of experience I may not win dunno majority of C races if I was doing them (not) but expect I would win a bunch and never win B or A … I am a B power with wily dunno “A race head” and “power isn’t the only skill” (but ALSO useful!!). I don’t have an exact number (re 6 months, … 4 months?) I think USAC you have to apply to be downgraded? Not saying that (though himmm…) but mch longer window of results I strongly believe is warranted.
No matter what other variables others are discsussing up there this is a separate variable (how far back counts in ranking, results, categorization minimum etc)… in short, I have strong opinion should be LONGER. (There may (?) be exeptions but the onyl ones I can think of are someone miscategorized in first place. Broke your leg? Train once able and don’t mini surge take out the D’s if you’re an A/B doing rebuilding after hiatus. If it’s a long term problem in longer term your results will fade too. No judgements needed (which if think are miscategorized for long time for wrong reasons … am not sure I can imagine the case but THIS IS THE START (as per James etc) lots more to come. I’m just saying don’t fade those 45 min alpe races and be a B or C in 90 days lol!
1 Like
Sounds to me like this system needs two numbers.
Zwift Racing Rating that determines which pen you can race in, and can go down or up depending on how well you do in a given race compared to how well the algorithm thinks you should do in that race. There could be separate ratings for different types of race or course. This would persist from season to season, might have a built-in decay, and should rapidly adjust as people enter the system or after a long absence.
Zwift Racing Ranking that is the cumulative display of “hey, look how I’m doing at racing!” This could reset every season, enabling people to say they were the #10 Cat B racer in Q2 2023 etc.
Maybe that’s what ZR.app does already - I saw it had a season breakdown at least.
1 Like
As far as I can tell, it’s going to be the same, but even if it’s not, it creates more problems. I’m not coming to this fresh, I made a whole video on how the race rating work, I just wish I’d gone in to why it’s so rubbish for Zwift.
Everything you described is already on ZR.App
2 Likes
Doing badly is another way of saying that we don’t treat every race we do as an “A” race.
Sometimes we are just out to get a workout and racing is simply more fun than doing an actual “workout”.
So unless you want to come up with some additional rules so that people don’t get penalized when they are not actually racing up to their ability it is probably simplest just to look at events where they did race up to their ability.
I wondered if that might be the case, hence my last line.
On that basis, while I’m not a direct user of ZR.app, I’d stick my hat in the “just use that” ring. Reinventing an apparently very good wheel seems like a huge waste of effort.
6 Likes
Yep, that’s accounted for in zr.app too.
It’s a strange hill to die on. Some in here are going in to the weeds and not seeing the wood for the trees. The data doesn’t lie - see the great chart from @TrevenM above.
If you split riders in to pens using a points system, you will see huge disparities in ability. If you split using a rating system ala zr.app, you see closely matched riders by ability. As both systems move riders up with success (one more carefully than the other), isn’t keeping riders roughly closely matched (on actual capability for success) the key requirement?
4 Likes
Are there any committed Zwift racers (more than 30 races this year?) who think this system is better than zr.app?
Ive said this with CE and its the same with ZR Score. VISUALIZATION IS KEY!
Sure you can see what your current score is in many places but you cannot see when and where you scored the points. There is no leaderboards. Come on scores without leaderboards?!?!?!!?
I also think that ZwiftRacing.app nailed one thing they separated vELO rating for classification and Points for race score. I really like this. While Im no fan of private link races I really, really like how well the vELO works and the number of categories ZwiftRacing has broken things into.
I think its such a shame that ZHQ doesnt collaborate with ZwiftRacing to learn from all the invested learning and development that has gone into that app. ZHQ is at least a year or two behind ZwiftRacing on this and the catchup process up will be a huge pain for Zwifts customers.
7 Likes
If ZR.app got rid of their pheno types modifier that only serve to keep the status quo it would be better start.
There is still an issue with the whole basis of the ranking numbers are based on a foundation that is inaccurate - Start with inaccurate data, dont expect it to improve. That is, the current seeding of riders is not great as we all know (huge blocks of riders around Cat limits, only racing to stay top of Cat etc) so the actual results & ranking is not accurate (This is an issue with both systems).
That is not a statement on the vELO or the algorithm as all that works, I just think you have 2 different systems which are not in check which results in bad data going in to start with.
Ultimately for this whole change to work, you have to have accurate seeding of riders. Ive not seen any solution that currently overcomes that (Only having A-D is big problem with this)
If the seeding of riders is not accurate, dont expect the output to be.
more nuanced problems aside, of which there are probably about a billion, i don’t think it’s particularly well thought out. there was a B profile posted earlier in the thread with a score of 330 or so. i’ll just make up 4 pens based on a 0-1000 scale:
0-400
400-600
600-800
800-1000
regardless of who belongs where, beginner racers get shafted because the spread of ability is so wide, and are immediately put off
2 Likes
zrCS is an extremely effective seeding mechanism. Very high correlation with results.
zr.app is designed to work regardless of how riders are placed in pens… so it works with existing 20m A-D cats, and it works using ratings or zrCS as the determining factor.
It is true that we could clean it up even further without the legacy A-D cats, but high correlation with results is what we need, and high correlation with results is what we have.
1 Like
Exactly. That’s what I tried to show in the above chart.
There is some flaws with the system, and it currently runs off a bodge added to a bodge, so its not clean as it could be.
Take the initial elo, you had to add the podium bonus to the number (I know, as I recommended it) as it would become neigh on impossible to find a race with a strong enough field to get promoted as the score increase would just dwindle the nearer you got to the top of the ranking as there wasn’t the quality of rider for you to beat to score well – This bodge, then caused elo inflation amongst riders who race consistently and podium (sounds familiar), so there has been the introduction of modifiers at each stage to suppress this inflation (Pheno type, 200 point limit etc).
At no point is the root cause being fixed (that being the seeding of riders is not changing) as its out of your hands, but ZR.app is having to introduce modifiers to suppress numbers as they cant fix the root cause.
That being said, Its much better than ZP, its clever in what it does and those involved, Tim, you, others should be well-pleased of the work that has been done part time and it really does show up zwift lack of commitment to this.
If there were a simple way of doing it well, I think someone would have suggested it, if not actually done it. The bodges as you put it are optimisations!
1 Like
Most of the issues you talk about are due to zr.app not being adopted widely.
If riders are doing racers where entry criteria is flawed. CE, Zp scoring etc and they are are being put in pens against weaker opposition than they should be facing then unfortunately we need modifiers to stop their ranking inflating to inaccurate levels.
We would not need as many modifiers if it was universally adopted for racing.
And whether you call them bodges/modifiers or whatever look at the results do they correlate with the ranking which the simple answer is yes it’s pretty accurate compared to any other system that is being proposed
4 Likes
Yes this is completely correct (apart from the word bodge) - there are a number of fixes that have had to be applied to account for the A-D system. Those fixes work pretty well though, it would just be a much cleaner algorithm if the pens were more dynamic.
What I conclude from that is that if Zwift took on the zr.app approach, it would actually be even better than today, because at the moment zr.app has no control over which pen a rider enters.
1 Like