Plus you actually don’t learn much about a sprint finish if you’re coming of a 60% HR vs already being at 80%+
Also despite your variable way of using races and despite ZP being very crude, you’re ranked 597th in the world, so you’re only ever going to be in Cat A.
The only way to avoid a good ranking system ranking you well is to never do well, and what’s the point t in that?
I don’t see the conflict.
Like I said earlier, just like in most PvP online games:
There should be ranked events affecting your ranking. A way to climb, to prove yourself, to make Zwift actually look like a real sport. But those events need to be fair. And you earn your spot on the ladder, not by some outside measure but from your actual race results. You want it, you work it, then and there.
And then there should still be events that use either your current ranking or performance metrics to put you together with somewhat equal opposition but the results don’t affect your ranking, so you can ride/race without pressure on a bad day or just fool around and practice back pedaling or whatever to your heart’s content.
It’s the best of both worlds.
Edit: And this solves another problem too. You want a goofy race? Performance metrics on steel bikes riding Road to Sky Reversed? Great! Stage a race like that but you need to do it outside the ranked races. Make it an unranked race.
You want a ranked race? Then you need to conform to certain standards. The ranking system should be adaptable enough to handle races of varying length (within reason), with varying field depths etc. There are already ranking systems used in real-life sports doing this today and it works well enough. So there will be a subset of all available courses which are “approved” for racing, you need to field a minimum number of participants (or the race will default as unranked), etc. Reasonable standards for acceptable racing. And as long as you do, you can be Billy the First Time Race Organizer and still contribute to the sport and the community while providing fair racing. Or Zwift can go Riot Games (League of Legends) and monopolize all the ranked races. Hands off, community organizers. I don’t care!
And then Zwift replies “Great idea! You can do this yourselves in your own leagues. No need to involve us.” BUT THERE IS!
Zwift needs to be the UCI of Zwift racing. They need to guarantee and maintain the technical platform, the standards, the rules for the “official” races. We’ll help if we can, but don’t dump the responsibility on us. It’s like a dictator telling his subjects they can vote all they want and set up their own little village parliaments but it’s not going to change anything.
4 Likes
I don’t understand why you think your criticisms mean that the system wouldn’t work. I can only assume we have very different ideas of what it means for a ranking system to work. What would you like it to do? It may be that your expectations are too high, even impossible.
Any discussion of how well different approaches might work, must be grounded in a common understanding of what the goals are, and perhaps what the minimum acceptable level of performance is, if the goals prove to be too lofty.
For me, I’d consider it a success if I get put into some races where I’m one of the strongest, and accept that this also means I necessarily get put into some races where I’m one of the weakest. Actually I quite enjoy the latter on occasion, it gets a bit wearing when it’s all you get though. It would be nice to have the races reasonably close, but the range of outcomes is more important IMO than all riders being very similar standard.
As for anyone else who wants something else, something they consider better than this…what exactly do you want, and why?
4 Likes
Cheaters are always going to find a way to cheat no matter how well you design a system.
1 Like
As a suggestion & thinking out loud, could you not try to take the best from both…
Use both a race position ranking\score & physical performance to provide race rankings its probably overly complicating it, but the idea is this would allow you to seed by physical metric (wkg) and score by performance (finishing position against other riders). Pens could be split by either physical performance or ranking points, or division.
(This might be very close to the old CVR predecessor set up now I think about it)
For example, you end up splitting everyone into 10 divisions but your physical performance limits the lowest division you can go.
So Cat A riders would be able to be ranked in Div 1 - 4 and their race performance metric would place them across the 4 divisions.
Cat B rider would be able to provisionally Rank in Div 3-6 initially but with performance rankings allowing them to ascend to the top 2 divisions if they consistently keep winning with B grade physical performance
Cat C 5 - 8 same as above, but they can go above Div 5 if they continue to win.
Cat D 7-10 same as above, but they can go above Div 7 if they continue to win.
Feel free to pull it apart
Great post Graham! Congrats!
Within the performance based categories it is not possible to track your personal progression relative to others. This is where a ranking system comes in. You get the small wins by gaining a few ranking points.
I like to have a look at my ranking points in ZP after a race. I seldom win anything after I switched out my wheel-on trainer, but it feels good if I at least got a reduction in ZP ranking points.
I like your idea about having some high quality ranking races, but that would take away the motivation and joy of having ranking points for the other races. Maybe all races can count towards your ranking if some quality criteria are met? This must be known in advance, so that riders who wants to race as training can bail out. This would also have the benefit that race organizers would strive to set up races with good participation and quality.
Maybe Zwift could have weekly ranking races where the pens were split by ranking points. Think if you could have the ranking of the other riders around you in the riders nearby list. That would be motivating!
3 Likes
This is exactly what Zwift should have done by now or at least used as a logical next step after the cat-enforcement-with-questionable-added-benefits test. You could start small and try it out. But you’d need a test than ran for longer than a week of course… A recurring thing a couple of times per week over a season.
Later on, if it works well (of course it will… might need some tweaking but what system doesn’t?) then I don’t think you need very gruelling or limiting criteria for your race to pass as “approved for ranking”. You’d have to weigh generosity against room for exploits of course, but most races that are run on a weekly basis today would probably work nicely. It’s usually the same proven routes, good ones, over and over and most races run for like 40-60 min or thereabouts. So there’s already something of a standard to lean on.
Then if you want to run a 120 km race… maybe we can’t accomodate that into the ranking system. But you can still use the same categories or division you would use to split into pens and they won’t be any worse than the current ones for non-standard events, I don’t think. You probably can’t predict the winner based on relative ranking within a pen in an “odd” race like that, but it’s a half-decent starting point at least (20 min FTP isn’t necessarily a better division in a long race, quite the contrary).
What about short races then? Some like sprint races. I hate most of them because I learned to. As you know, they screw up your performance-based ranking and even a non-cruisers is forced to think like one from time to time or he’ll get hurt.
But… a course like Bologna is a nice one in my book. Well, you could demand two ascents to make it official. Or you could allow a range of short races by doing something similar to what they do in x-country skiing ranking. You adjust ranking by time separation between you and the winner in relation to total time. It’s a bit different with cycling, though, because draft is much more important than in skiing and in many races you’ll see very small time separations for that reason (riders often approach the finish in a group/blob). But you could juice up the, say, first 10 spots with a sort of bonus ladder, like the time deductions for 1st-3rd place in many WT races, to make rank score change more noticeable.
It would all need to be tuned of course. Sure, it might take a little time, trial and error. But it would be building a sound and fair base and working forward in a sound and fair direction, away from the performance measures.
Can it still be exploited? I bet someone will come up with something. You might need to tweak the system if so. Or you don’t. But it’s not totally obvious how to exploit such a system (except yoyo’ing /yawn).
With the system we have today it is obvious. It’s simple arithmetics to prove that a heavy rider will have some kind of advantage. It’s simple logics to figure out that someone underperforming at the top end of a cat will be at a huge advantage. It is self-evident that race length will play tricks on people’s categorization. It is all so obvious when you think of it that Zwift (and former ZP!) should have seen all this coming. Yes, yes, I know there were people within ZP who did, hence the nice little USAC ranking theft which, for various reasons, unfortunately never turned into more than a cryptic number at the far end of the results table.
I have mentioned before that the ranking will be influenced by what type of races you mostly do. We see this with the existing ZP ranking as well. With only one global ranking, we must realize that it will not perfectly reflect each rider’s chances in all races. Each rider will have to look at their ranking as a reflection of how well he does in the races he does the most.
I think the races that are classified as ranking races (races that count towards the ranking), should be the most challenging in the actual race format.
E.g., the flat endurance races should be long, the hilly races should have significant both total and single climb elevation, the crit races long enough to separate the cliff from the wheat at the end, and the punchy races have some length and smaller climbs.
If we assume that people mostly select ranking races that fit their abilities (because they don’t want to lose ranking) the ranking will end up reflecting the racer’s abilities at what he is best at and nothing more.
2 Likes
I very much agree with what you are saying in the whole of your post.
But on your point above I would ask do we really know why the majority of racers choose the race they are entering? Personally those races I have entered ( not many) are based on how long I am going to have to suffer, times of day, how many racers etc and not what might suit my limited ability best.
In your post you have referred to race ranking events several times. I’m sure there are leagues and races where rankings are important but I’m not sure what percentage of the overall Zwift racing scene they apply to. I would like to see race ranking become a much more meaningful metric.
As you imply two riders can have exactly same w/kg and race ranking and in isolation these two things wont tell you who is more likely to win on any particular route. Dig deeper and other metrics may start to point you towards the more likely of the two.
For others who have not seen a previous post of yours I very much like your thoughts on ‘TRENDING’. How can race ranking be used further than just providing a one off figure? - Exactly as you have written - TREND analysis.
Example of two racers exactly same wkg & race ranking. Say they are both almost exactly on the cut off line between category boundaries. With a trend element to the race ranking it might be one with race ranking ( winning/good performance streak) races group up and one with race ranking trending down ( losing streak) races down a group.
*** Probably not for a race ranking discussion thread but something I may suggest elsewhere. Does the power metric have to be a fixed line? Could it be a small band width spanning two categories where something like race ranking or trend of race ranking dictates whether you race in the higher or lower of the two categories ***
1 Like
I have spread my thoughts in some different threads and I’m maybe not the best to write clearly what I mean 
I’ll try to clarify.
I’m envisioning two types of races. One type with performance-based categories like W/kg, the MAP+CP/W’, or something else. These are the ad-hoc races you do. One performance metric is used for the stable categories like W/kg is today. You will know which category you are in, in the same way you can today. Most of these races do not count towards your ranking, so the result does not matter in that regard.
But some of these races can be ranked races. They are up to a certain standard and get this tag. You must then be aware that if you e.g. as a sprinter choose to enter a long race with lots of elevation, you will probably not get a good rank for that race and might end up losing some ranking from that race. This is dependent on how the ranking system is implemented. Ranked races of this type are also the races that can give you the biggest bump in ranking if you beat opposition who have gotten their good ranking from another type of race than you have (e.g. they: sprint vs you: endurance).
The other type of race is the “special” high-quality ranking races. These could be fewer and use the ranks to put you in a pen with other racers with about the same rank. You should here select the races that you think you have the best chances of doing well. The sprinter will e.g. choose e.g. a crit race.
The post you are referring to was meant as a simple/naive approach to show how to promote and relegate racers that have performed either at the top or at the bottom of their category in recent races. This was meant to be used for the performance-based categories. There are mathematical models to detect trends that would be better suited for the purpose than what I presented. See e.g.: Test for existence of a Trend in a Time Series | by Aayush Ostwal | Towards Data Science. If the trend is that a rider is placing higher near the top, or has stagnated at the top level, he is ready for the next category.
Are you suggesting that an increasing trend in ranking is what should be used to bump riders, who places repeatedly at the top of their category, up to the next category?
It is not given that the racers that win a lot get ranked higher and higher. It depends on the ranking of their opposition. Within a category, the ranking of these racers will stagnate because they have the highest ranking themselves and meet only lower-ranked racers. Magnus Carlsen, the World Champ in chess, has this problem. He has the highest ranking of all players, and can’t afford to lose if he is going to set a record of 2900. This is because the system is expecting him to win.
Having the highest ranking in a performance-based category does not necessarily mean you are better than all others in the category. It could be that you have a good sprint but suck at hilly races. We see that in the current B category, where there are many sprinters with very good ZP ranking. They usually only do certain types of races.
1 Like
With (genuinely) due respect, you need to apply more than a bit of practical thinking.
Start from the question what does any of that matter? If it does, you can show how by doing the math of gains. There’s no assumptions, just results.
1 Like
Agree.
The title of this thread started as:
Race Ranking - Will it work
Then moved to:
Race Ranking discussion
I wonder if a more positive move forward might be:
Race Ranking - What steps need to be taken to enable it to happen?
1 Like
Race Ranking - What steps need to be taken to enable it to happen?
What I’m about to say is me taking part in the discussion as much as it is to let you know we’re thinking of these things. It is not set in stone nor in order and as such is not indicative of what is coming next, is ready to go, etc.
Short answer: make racers on Zwift trust the results by making the races themselves more trustworthy.
Long-ish answer:
- Category enforcement (which we’re testing, as you know)
- Restriction of what hardware can/can’t be used in a race
- Potential addition of a necessity for having a 2nd power source paired (basically dual analysis in the Zwift client)
- A scoring system for race results that is accessible yet still meaningful, i.e., as a first step, do we use the current ZP individual ranking formula? Longer term, do we feel want to go on a simple cumulative points scale? Do we want to go more toward MMR or Elo? etc.
- A place to then view those scores, preferably in leaderboard format and somewhere that isn’t ZP
As a fun set of questions that I haven’t yet attempted to answer (but which we think about a lot):
- What do we do with those scores? Other than bragging rights, what incentivizes someone to pursue a top ranking?
- Should we try to make them like ladders in other games, meaning they reset every so often? Do we want strictly all time? Do we want a mix?
- What ways could we attribute scores to a race ranking that would be different and uniquely Zwift? How can we breathe life into racing by breaking from tradition? Should we?
So basically, category enforcement is the first step to making ranking by results - whether in conjunction with performance or standalone - because it’s a step toward making race results more legitimate by making them more trustworthy. We’re workin’ on it!
15 Likes
The immediate potential issue that comes to mind with using existing ZP ranking scores, is that won’t they be affected from people racing in the wrong category before enforcement bagan? Both those racing higher than they needed to under Zwiftpower, but also those racing in lower Zwiftpower categories than they should have.
Not that I’ve raced much over the last few years on Zwift, but I’ve always wondered just how representative of a rider’s talent are those Zwiftpower ranking scores. It’s our best five results over a 90 day period, it says nothing of our own poor results.
There’s also other issues like DNFs not being accounted for and we could be racing our hearts out in races against riders who aren’t registered on Zwiftpower and so we won’t get any ranking points off them even we finish ahead of them.
1 Like
Flint
Many thanks you for picking up on my indirect question to Zwift so quickly. Most grateful to know you are thinking about all the issues you have list.
1 Like
If y’all haven’t listened to the latest Zwiftcast podcast, Flint is featured and speaks to race ranking… basically what he writes here. I encourage you all to have a listen.
5 Likes
Don’t you think we need a way to handle the situation where riders win and win in one category without getting bumped up? To me, that is a big problem with the performance-based categories and should be high up on the list.
3 Likes
I do agree with that, yes. But what I believe we are better to focus on first is making sure that each performance is valid and that everyone within a race knows that the others they’re racing against are meant to be racing in that particular event.
Personally, I don’t feel that we need to focus on race results just yet because I don’t feel like the results are valid. As a racer, I wouldn’t believe that someone getting bumped up based on their race placement would hold much water because I would immediately ask “but yeah, did that person cheat? Is that a real result?”
Though the same can absolutely be said for performance. And that also doesn’t address the folks that cruise and just win over and over in the same group because they know how to avoid getting bumped up.
Again, that’s just personal, and I don’t disagree that we need to focus on race results. It’s just a question of when in my mind. “Results” as a full body of work comes soon, and the situation in particular that you’re presenting is housed under that body of work.
4 Likes