I’ve been spending a lot of time getting intimate with Zwift route data lately. Why? Two separate projects. One that predicts your finishing time on a given route… the other that tracks those times on a leaderboard against others!
Here are the details…
Introducing ZIMetrics
For years, I’ve wanted to add two things to Zwift Insider’s route detail pages:
- A route difficulty rating: something easy to understand, which can be used at a glance to rank Zwift’s entire universe of routes
- Route completion time estimates based on your power output: it’s not always easy to glance at a route’s length and elevation and figure out how long it will take you to complete, and Zwift’s in-game estimates aren’t always helpful. But most Zwifters know what sort of watts per kilo they can sustain for a given amount of time, so if I could calculate accurate time estimates for a given w/kg, that seemed like it would be helpful.
I’ve been tweaking and testing an algorithm for a few weeks, and I’m pretty happy with its results at this point – hence this announcement.
Time Estimate Notes
Let me start by saying this: there’s no way to estimate a perfectly accurate completion time on Zwift – not unless it’s a TT race and you’re perfectly following a particular “power plan” à la BestBikeSplit.
My goal with these time estimates is to predict how long it would take a semi-experienced rider, riding efficiently in a group, to finish one lap of the course, including any lead-in. The algorithm assumes you’re a 75kg, 183cm rider drafting on the flats, on a decently fast bike, and pushing harder on climbs than on flats and descents.
Basically, the algorithm is based on a typical Zwift group ride or race situation.
Of course, draft conditions vary widely, as do rider height/weight, choice of bike, and power variability. In order to keep it simple and avoid requiring people to enter height, bike choice, etc, I had to make certain assumptions, and I’m fine with that, because I don’t think we lost much accuracy in doing so. In my tests, I’m finding that the ZIMetrics estimates are almost always within 3% of the actual completion time. That’s certainly close enough for our purposes.
Route Rating Notes
Each route also has an objective “Rating” score between 0-100. This is simply a normalized version of the time estimates above. For example, if ZIMetrics estimates that a route will take 40 minutes to complete, that means it will have half the rating of a route that is estimated to take 80 minutes to complete.
Yes, it’s not a perfect metric. But I’m not sure such a metric exists. The ZIMetric route rating is a simple way to rank routes against each other.
I’ll be using this to compute the bidon score for the printable routes by difficulty page, once I’ve got all the routes updated. (That’s why this printable list hasn’t been updated for a few months, by the way…)
I should note that I’ve set it up so the current highest-rated (that is, the toughest) route – London’s “The PRL Full” – is rated at 100. If a tougher route gets released all scores would get automatically recalibrated based on the new maximum.
Not All Routes… Yet
ZIMetrics aren’t yet available for every route, because I have to edit each of the 236 routes in the database to include specific, accurate data before the ZIMetrics algorithm can spit out good results.
Here’s the complete list of routes that have ZIMetrics enabled (90+ routes currently). That list should continue to grow daily until all routes are finished.
[adinserter block=”4″]
VeloViewer Leaderboards Update
Our Route Hunter Leaderboards p/b VeloViewer, have been very popular since they were first launched in the middle of Covid craziness on August 18, 2020.
If you aren’t familiar with these leaderboards, they basically allow you to compete for overall fast time across a combined set of Zwift routes and KOMs. It’s a fun way to pit yourself against others, track your completion of routes, and see your own fitness progression.
These leaderboards are now updated with the latest routes and KOMs.
Questions or Comments?
Share below!
@Eric Schlange you wrote the PRL Full is the current baseline for rating, e.g. is rated at 100 unless another, harder route comes along. When I check the PRL Full route then it currently has a rating of 85.9. Does that mean that there is already a harder route?
Nah, just means the page was cached and I forgot to refresh it. Check it now! https://zwiftinsider.com/route/the-prl-full/
Thanks Eric, I appreciate you putting this together
Amazing. I was just asking if there was something like this today. The amount of routes over had to bin part way as I’ve ran out of time…. Great stuff.
Thanks Eric – this is shaping up nicely. How will this compare with the ZwifterBikes site info? Aside from adding a route difficulty metric, I find that the ZwifterBike site is pretty accurate for me (within a few seconds) when I compare my average wattage output over a given route to the estimated completion times given the bike and wheel set I choose.
I don’t know how this compares to ZwifterBikes’ estimates.
ZwiftBikes used data from ZwiftInsider to build an interactive website that can predict your times super accurately. My climb portal times are just a second off the actual times. The developer obviously has more free time since he does not have to keep posting Zwift press releases. 😀
Fantastic work-much appreciated!
This is really awesome, very much appreciate this work!
Impressive work and I think you may need some therapy! Well done though!
Amazing! I often ride on my own. Any idea what the average % time loss is if you aren’t riding in a group?
Not for a second suggesting you go through and re-do anything! Just wondering.
Only other thing I thought of which I try to work out is trying to tie in the distance travelled on a route when doing a workout.
First you would have to workout your achieved w/kg on a workout. Workouts will have a standard % of your FTP as the output over time which depending upon recovery blocks could be somewhat less than your standard non-workout rides. So, have a workout, input your FTP, gives you the average wattage for the ride. Divide weight by that wattage gives you the average w/kg.
The workout has a set time. Time at the average w/kg will give you the total work.
Then compare this with the routes you are thinking of doing.
This is only because I like trying to hit a route badge while riding a workout. I have missed the mark on a few, because although I think average km/hr on flat may be say 30km/hr, there is climbing involved in the route, and also the average w/kg on a workout could be quite a bit less than even your zone 2 endurance pace if there is a lot of recovery time.
It’s more a mathematical fascination than anything I guess, as not much point doing all the work to calculate that and very few would find helpful.
I guess at the least it would be good to know what the average power would be for a workout.
Aaaaanyway, awesome work Eric!
I’m also interested in what your best guess is on time % based on riding without much draft on the flats. I do ride some group rides and race but when I’m route badging it’s always alone. So many routes are would never be ridden if all a person did is ride in groups.
The problem is, you can’t just add a % to every route to figure in a lack of drafting on the flats… because some routes have a lot more flats than others!
Thx a lot Eric. Last weekend I had in mind to do the Pretzelfest and thanks to the time predictions… I could reserve a spot inside my comoplicated weekends 🙂
Great work Eric. Any chance the climb portals will make it to the VeloViewers leaderboards?
It should be fairly easy to combine France and Watopia just using the best pr.
They shouldn”t be included as they aren’t always available to ride so people won’t be able to reach 100% completion. This is the same as Event Only routes.
I disagree, normal routes are not always available to ride either because of rotating worlds. There is a world schedule and a climb portal schedule.
Always available is not a prerequisite.
Some climbs you will never see again and some rotate in so infrequently its not feasible for people to ride them all unless they have done so as and when they are released.
For instance, if you take the Laguna Negra it was last scheduled and available on the 7th March. Seems silly to include it on a leaderboard doesn’t it!
The other problem is that there are separate leaderboards depending which world you entered from.
I like the climb portal and if Zwift made it so you could ride any climb I’d be all over it to complete them all but I don’t bother as I understand that its impossible to ride them all.
With normal routes you can setup a club ride and ride most if not all of them when ever you like. Or you know that you only have to wait a week or so until the desired word is scheduled. You can’t do this with the climb portal.
You also have to consider that most people don’t like climbing which is why the flat routes are more popular and the longer harder climb portal climbs aren’t popular for this reason.
You have to remember that just because you think something is a must have feature doesn’t mean its practical or desired by others even if its important to yourself. Most Zwift feature suggestions unfortunately fall under!
Agreed you make a good point. I would not assume I am the onlyone requesting this feature, there are quite a lot of riders enjoying climbs after all.
As you say this may make more sense when the climbs are more accessible!
All worlds are available if you choose a workout first. For example, I choose the 2018 UCI worlds Short Lap today. With and FTP of 210, I completed the route about 15 minutes before the workout was completed.
Pick a short workout, say a VO2 max and a long route in a world and once the workout is completed, the ERG turn off returning to normal allowing you to Ride On!
We’ve got a separate Veloviewer Climb Portal leaderboard (or rather, two of them) here: https://veloviewer.com/segmentHunterSeries/1000007
Very cool, this makes the geeky spreadsheet I used redundant, mine tries to predict a time depending on group ride or individual or workout, I base it on my average power output from previous sessions of that kind. However in reality this feels like a feature that zwift should have. Especially for a workout. They know exactly how long it will take you. But great work thanks 👍
Hey Eric, thank you for this work. So will there be hope, that in the next time we will see an updatet “complet-rout” list with an duration/rating add on. Including the new routes? This would be awesom, if it is again a printable list. The inner monk will mark the executet routes with my textmarker 😀
Hey @Eric Schlange seemed a bit out on ZRL yesterday – predicted 57 mins averaging 3 w/kg, and I managed 59 at 2.5… (I’m 77kg, 174cm).
I think the Veloviewer leaderboards are great and I always try to complete all the new routes as quickly as possible so I can maintain my top 15 (sometimes top 10) position.
One tip for people is to keep track of when you rode the routes because from time to time you swich the Strava Segment used and Veloviewer doesn’t always find the ride and I need to modify\crop the Strava ride.
One observation though is that you are missing the Libby Hill After Party route which is a selectable route.
As a programmer I have been seriously thinking doing something like this myself because the zwift approach is fairly worthless and the BestBikeSplit was confusing at best. While you are on this I will put that idea in the fridge and see what you come up with!
@Eric Schlange Do you think you could put in a good word with the folks at ZHQ? In the companion app, where the route is mentioned for a certain event, there could be a little grey/green tick next to the route name for the route badge? Sometimes I pick events purely based on whether I’d get the route badge too (as per the excellent advice from ZwiftInsider 😉)
The w/kg estimations are great, been using these on the Climb Portals for a while and very helpful to see on routes too now
A comment on the ZIMetrics, with the scale being so great it gives the illusion that so challenging routes are easier. For example Peak Performance is 27, and while there’s a lot of flat that is still a serious climb (for me) and I would give it a greater than 27% difficultly. Although this will change as familiarity with the scores grows I guess
FWIW I like the ProCyclingStats ratings, but even the icons are useful to get a quick understanding of what a route entails – https://www.procyclingstats.com/info/profile-score-explained
A lovely addition, thanks Eric!
Thanks @eric Schlange this information is really helpful. Please could you also show the time for a lap without the lead in, that way we can calculate the time for multi lap races. Another really useful addition to the Sprint & KQOM Segments section would be the distance to the start of the sprint or KOM. I find this useful for points races where I want to increase my speed before the start of the segment.
You can do it yourself on Strava buy looking at other peoples rides on the Route if you’ve not ridden it yourself.
I agree we can look up all this data on Strava. What @Eric Schlange is trying to do is provide consistent data for a route in one location which saves users having to do their own research on Strava. I was just suggesting there are two times provided for each rider level as just providing the time for the lead in and a single lap does not allow people to calculate a time for say a 3 lap race. You could provide separate times for the lead in and the lap or display the lap time with the existing time shown.
Eric, just a thumbs up for the much appreciated things you do on Zwift!
Cheers Eric – I’m sure your algorithm is much better than the method I have used to keep my list up to date for a few years (copy and paste as many ZP results into Excel for a given course and work out the times by range!) – I was planning the next iteration and working out by a given w/kg and to try to make the results more accurate but now I don’t need to, particularly as I have a number of the new courses to do yet!
If you were able to replicate some of the calculations I have there then it would be great! (Time for x laps, how far can I get in y minutes)
My very amateur effort is here just in case anyone is interested!
https://docs.google.com/spreadsheets/d/18HvjSpq-2ooQfAoeVwX86GmehNUlJTYCQ68RzjA39d8/edit?usp=sharing
Hi Eric,
I suggest keeping the current 100 as gold 100 forever and not recalculate anything later (route difficulty rating). If any new route is more challenging the numbering could go simply above 100. In this way people could get use to a static scale and what was once x would be always x and everyone who already have some feeling/knowledge what a specific number means could rely on that knowledge without re-learning everything from scratch. Whereas, with an always changing scaling many people can get confused what means what. Just a suggestion.
Br Pal