UPDATE: Zwift greatly improved their banding algorithm after this post was published. To see current “keep together” test results, read the updated post.
How does Zwift’s “Keep Everyone Together” logic work? I’m not entirely sure, but after publishing speed test results for rubberbanded Club events many Zwifters said “Keep Everyone Together” mode behaves differently in Meetups than Club events!
That seemed goofy to me… why would Zwift have two different Keep Together algorithms? I knew I had to run some tests to see if the rumors were true.
In addition to seeing if rubberbanding behaves differently in Meetups than in Club events, these tests were also done to answer the question do rubberbanded speeds in Meetups make sense? Or are they too slow, too fast, or some strange mix of both?
The riders in these tests were 75kg in weight, 183cm tall, and rode Zwift Carbon bikes with 32mm carbon wheels. I timed the riders on the Fuego Flats Reverse segment since it’s flat and long enough to get a stable speed reading.
Test 1: A Pair of Bots
This first round of tests used just two riders in a Meetup event with “Keep Everyone Together” mode enabled. Here are each rider’s power outputs, segment times, and average speeds.
| Rider 1 | Rider 2 | Time | Speed (kph) |
| 300W | 300W | 10:19 | 41.06 |
| 300W | 225W | 10:55 | 38.8 |
| 300W | 150W | 10:56 | 38.74 |
| 300W | 75W | 10:56 | 38.74 |
| 300W | 10W | 10:56 | 38.74 |
| 225W | 225W | 11:42 | 36.2 |
| 225W | 150W | 12:09 | 34.86 |
| 225W | 75W | 12:09 | 34.86 |
| 150W | 150W | 12:37 | 33.57 |
| 150W | 75W | 14:10 | 29.9 |
Observations
The first thing that stands out here is how much faster the riders moved compared to a banded Club event. Here’s a comparison table:
| Solo Rider | Banded Meetup (2 Riders) | Banded Club Event (2 Riders) | Club Event (4 Riders) | |
| 300W | 40.23 kph | 41.06 kph | 32.58 kph | 40.19kph |
| 225W | 35.95 kph | 36.2 kph | 28.72 kph | |
| 150W | 30.81 kph | 33.57 kph | 26.12 kph |
Looking at these comparisons, it’s clear that the riders in the banded Club events were moving much too slowly, while the banded Meetup is pretty close to accurate. (Two riders holding 300W in a banded Meetup are just slightly faster than a standard event with 4 riders holding 300W.)
So something is clearly different with Zwift’s “Keep Everyone Together” algorithm for Meetups vs Club Events. Interestingly, though, both types of events share the same odd behavior when it comes to having a “strong” rider pulling a “weaker” rider. The results above, just like the Club event results, show that times don’t change as the weaker rider’s power decreases.
That is, when Rider 1 is holding 300W and Rider 2 is holding 225W, they’re traveling at the same speed as when Rider 1 holds 300W and Rider 2 holds 75W. That’s strange. Shouldn’t groups travel significantly faster if the group’s average power increases significantly?
Test 2: Solo Rider
Next, I wanted to test if a solo rider in a rubberbanded Meetup still traveled as slowly as the paired riders above. So I created a Meetup with “Keep Everyone Together” enabled, but only had one rider join. Here are the results:
| Rider 1 | Time | Speed (kph) |
| 1000W | 60 | |
| 300W | 10:56 | 38.74 |
| 225W | 12:09 | 34.86 |
| 150W | 14:09 | 29.94 |
Observations
First, notice that the times for the solo riders matched the times when we paired up two riders, but one was doing less power than the other (see Test 1 above). That’s certainly not a coincidence. And it actually matches what would happen outside! That is, you would travel at the same speed riding alone at 300W as you would if you held 300W but had a weaker rider sitting on your wheel.
Secondly, notice that the solo rider times in this test are slower than solo rider times when in free ride mode. They’re not nearly as far off as my solo rider tests in banded Club events, but they’re still off.
The 1000W test is also interesting since it’s sort of a “max speed” test that shows if banded Meetups have high speeds impacted. In rubberbanded Meetup mode the rider reached 60kph, while in free ride mode the same rider holding 1000W reaches around 62kph. Again, much closer than the banded Club ride (where the 1000W rider could only hit 47kph). But still not quite the same speed as a free ride.
[adinserter block=”4″]
Concluding Thoughts
These test results clearly show that “Keep Everyone Together” mode in Meetups behaves very differently from “Keep Everyone Together” mode in Clubs. Specifically, banded Meetups are markedly faster than banded Club events.
And that’s weird, right? I can’t think of any reason why Zwift would want to run two different versions of “Keep Everyone Together.”
Do banded Meetup speeds make sense? Or do they feel too fast, or too slow? My hunch is they may feel pretty accurate for medium and high-strength riders, since pack speeds in these limited tests are mostly aligned with the power being done by the stronger rider. For that reason, speeds may feel a bit fast for low-powered riders, since they’re effectively being pulled along at the strong rider’s speed, even if they’re putting out next to zero power.
Nobody will be breaking land speed records with the current algorithm like they did with the buggy one in years past, but for sure a rider who can only hold 100W is going to set a lot of Strava PRs if they jump into a banded Meetup with riders who can hold 300W!
Because of this, perhaps Zwift should consider tagging banded activities as “gamified” when posting to Strava, so the activities don’t show up on leaderboards or count for PRs. What do you think?
Coming Up Next
I’ll be testing speeds in a group workout to see if the algorithm that keeps riders together in these events matches up with either the Meetup or Clubs “Keep Everyone Together” algorithm. For science!
Also, we’re working on banded Meetup and Club event tests on Alpe du Zwift, to see how rubberbanding behaves on long climbs. Stay tuned!
Questions or Comments?
Share below!
Thank you. This is exactly why I stopped putting on the club events!
How do you feel about banded group rides on the Alpe that are marketed as a way to obtain the Tron bike without doing the work? Cheating or legit?
Not cheating, in the same way doing all your workouts on AdZ is not cheating. The game allows it, so it’s not cheating, imo. I wish that the results/time to complete a route was flagged so that it doesn’t artificially set PR times too low though.
Paul, I hear you. There are 2 bad sides to this, although because Zwift allows it, technically it’s not cheating. One thing is it undermines all the work that others do by slogging up the inclines like I did. But secondly, that ride will show as your PR on Strava. I was on one banded ride up there and never will again. There is no way I will ever make it up the Alpe in 1 hr 45ish minutes ever. So I’ll never get a PR.
I regularly ride on the Banditz Tronathon. I think it’s great that weaker riders can benefit from the band – the Alpe is a ride that many people would not do if not banded. My goal with Zwift is fitness, so more time on bike is good – it’s that simple for me. However the band is odd – on flat sections ( not the Alpe) the band seems to slow us down. On the Alpe the times are ridiculously fast – the whole group in regularly under 1 hour – which makes no sense when only a few ( maybe 10%) are riding at that kind of power level.
for me it would be much better if I did NOT have my PR set on a vended ride. I think I have a PR of about 52 mins – when I suspect my best without the band would be about 65 mins.
the Speed used to vary by world as well. A two person, non club, banded ride in NY being faster than Watopia.
Our club discontinued hosting events and instead post private meetups and try to invite everyone.
As for meetups, I’m not sure why it is so difficult to invite those followers tagged as favorites. There seems to be a lack of logic in the invitation screen. You would think that all of your favorites would be listed first, right? I frequently have to figure out who is missing and go find them using search.
I definitely want banded rides marked as gamified, and not to count for PRs. I do a banded meet-up once a week on Alpe du Zwift. It’s made me stronger and I enjoy the social nature of the ride. My PR set on one of those rides, however, is wildly out of my reach if I ride alone up the Alpe. This takes away one of the joys of attempting any climb, circuit or sprint where one can compare and contrast past efforts.
Not sure gamified would be the best way to do it I think that could put some people of doing banded rides entirely. Certainly agree there should be a way of discarding PR’s that you don’t agree with or are not realistic.
My Alpe time is 1h20m/1h30m normally; I had done it once in a rubber banded meetup, and I did it in less then 1h. I removed the activity from Strava, because I didn’t want it to count as my PR on the Alpe. I agree there should be a way to purge the PR from Strava without cancelling the whole activity; maybe just a simple button (for the entire ride, or even better for each segment/PR) that says “don’t consider this segment in any ranking”.
I would LOVE if banded rides didn’t record as PR’s. I enjoy occasional banded rides for the long badges that I consider out of my reach going solo, but would prefer they not count as my personal achievements.
In our meetups and during sprint segments, my actual power output looks correct on the heads up display, but my graph gets truncated to 600 watts? Plus savy riders can sit back and slingshot around the group using the pull of the band. Anyone else notice this during meetups?
Graph has always been truncated 600W. That’s just a UI thing… doesn’t affect your actual power.
I regularly do meetup workouts with a friend who has a lower ftp than I do and therefore works out at lower wattages.
My average speed difference is noticable between the meetup ride and what I would do on my own. Not a problem in a workout as it is the wattage targets you’re working to not speed, but in a recent meetup workout, the rubber banding snapped due to a technicality. I ended up 20 mins ahead on the road, showing that zwift is significantly reducing my speed in a meetup workout. And of course very little if any drafting when there is just 2 of you!
Another suggestion: try testing how the number of riders influence speed, both for club and meetup events.
I say this because I did notice the slower speeds for club events seems mitigated when there is a bigger crowd. But it will be nice to see the tests done with an incrementally larger number (ie, start with 2, then keep adding riders until perhaps 20) to see where is the “sweet spot”.
Thank you!
so many issues with the banded feature. sometimes it gets stuck at 48kph. glad someone finally did some tests with it. hope zwift works on this low hanging fruit
For your consideration — I have climbed the AdZ many times and now also the climb portal many times. Very few banded club/meetups on AdZ, none on Portal. Normally climb the portal on 100% but occasionally elect for 75%. On Strava, if using the 75% (and suspect true of 50% and 25%), they say the efforts is excluded (Excluded Effort). Perhaps Zwift —> Strava for banded events could do the same. Personally, I’m agnostic to numbers — it’s a game. Only IRL is real. If people worry about how others play the IZL game, they lose (sanity, perspective, fun). C’est la vie!
We stopped using “keep everyone together” for our meetups more than a year ago because of the silly 30 second window problem. Is that still a problem? Fyi the 30 second window problem is when someone stops pedaling, perhaps coasting down a long hill, connection issue or turning on fan — for any reason it seemed. Once they fall 30 seconds behind or thereabouts they won’t ever be brought back to the group. We all have to stop if we want them back. Our group is so good at riding together that we really only had problems keeping together when “keep everyone together” was turned on! We have a member or two that wants us to use it again, so I’d like to know if using it will mean we drop riders as before. We now do Club Events instead of meetups. Does the 30 second window problem still exist? Does this problem apply to club events?
“Because of this, perhaps Zwift should consider tagging banded activities as “gamified” when posting to Strava, so the activities don’t show up on leaderboards or count for PRs.”
YES! 100% agree.
I enjoy doing group workouts, I pay attention to cadence and if I am hitting the watts asked for, as well as time left till next change in power…. do not check speed at all. I do notice nearby riders and what their watts/kg is.
My group organised a group ride with keep everyone together on Lutscher CCW circuit. One of our group carried on after lap 2 and found the banded climbs quicker?
Lap 1) banded with all 3 riders doing 3-3.5 w/kg 21:59 15.6km/h
2) banded except last 1.5km, one rider at 3-3.5w/kg, one around 2w/kg plus short sprints 20:05 17.1km/h
3) solo rider 3.5w/kg 24:33 14km/h
No idea why the solo effort of the strongest rider was the slowest?
I see in the latest Zwift update 1.64 that this issue was “fixed”. It would be great to rerun the test to see if it helped.