You may recall that in March we ran tests on Zwift’s “Keep Everyone Together” (aka “rubberbanding”) mode for Club events and Meetups. Here are links to those posts:
- Speed Tests: Rubberbanding (Keep Everyone Together) Mode in Zwift Club Events
- Speed Tests: Rubberbanding (Keep Everyone Together) Mode in Zwift Meetups
- Speed Tests: Rubberbanding (Keep Everyone Together) Mode on Climbs in Zwift Club Events
- Speed Tests: Rubberbanding (Keep Everyone Together) Mode on Climbs in Zwift Meetups
In summary, our simple 2-rider tests showed that Club events used a completely different banding algorithm than Meetups, and that speeds didn’t fluctuate very realistically based on power variations. Our conclusion was that Zwift’s banding algorithm for Meetups was most accurate in terms of speed, and should be used for all banded events, perhaps with some tweaks to make speed fluctuations feel more realistic.
The good news is that the folks at Zwift HQ read our posts, and one of Zwift’s developers – Hugo Teixeira, the brain behind much of Zwift’s current pack dynamics – put in the work to improve Keep Together Mode, with an updated algorithm released in last week’s game update. Not only does the algorithm produce more realistic speeds, it also now works the same way for all banded events: Club events, Meetups, and even Group Workouts!
After running a few tests to confirm that the same algorithm was being used for Club Events, Meetups, and Group Workouts, we ran some tests to find out exactly how speeds changed.
Flat Test Results (using Fuego Flats Reverse)
| Rider 1 | Rider 2 | Time | Speed (kph) |
| 300W | none | 10:36.2 | 40.0 |
| 300W | 300W | 10:21.5 | 40.9 |
| 300W | 225W | 11:04.3 | 38.3 |
| 300W | 150W | 11:04.1 | 38.3 |
| 300W | 75W | 11:05.9 | 38.2 |
| 225W | none | 11:48 | 35.9 |
| 225W | 225W | 11:45.3 | 36.1 |
| 225W | 150W | 12:17.7 | 34.4 |
| 225W | 75W | 12:20.0 | 34.3 |
| 150W | none | 13:45 | 30.8 |
| 150W | 150W | 13:03 | 32.5 |
| 150W | 75W | 14:19 | 20.6 |
Compare these speeds to solo riders on a closed course:
| Rider 1 | Time | Speed (kph) |
| 300W | 10:35.0 | 40.0 |
| 225W | 11:46.11 | 36.0 |
| 150W | 13:43.56 | 30.8 |
Observations
Overall, these speeds are much faster than the old banded Club Rides. They’re close to the old banded Meetup speeds, but not the same.
The speed of a solo rider in a banded event now perfectly matches a solo rider free riding, which is great – that’s a definite improvement!
There appears to be a speed “floor”, based on 300W tests. There we see that a 225W drafting rider and a 150W drafting rider both return basically the same speed. Even a 75W drafting rider is just 1-2 seconds slower than the 225W rider.
Two riders holding the same wattage is always faster than one rider holding that wattage solo, which doesn’t match our TTT test results, which showed that a group of riders holding the same wattage travels at the same speed as a single rider holding that wattage.
That said, the difference isn’t massive. It’s interesting, though, that the speed for a solo 225W rider is so close to the speed of 225W riders (within 3 seconds) while the other two power levels (300W and 150W) show a much greater disparity.
[adinserter block=”4″]
Climb Test Results (using Alpe du Zwift)
| Rider 1 | Rider 2 | Time | Speed (kph) |
| 300W | none | 49:30 | 14.8 |
| 300W | 300W | 51:35 | 14.2 |
| 300W | 225W | 53:09 | 13.8 |
| 300W | 150W | 53:47 | 13.6 |
| 225W | none | 1:04:44 | 11.3 |
| 225W | 225W | 1:04:37 | 11.3 |
| 225W | 150W | 1:06:34 | 11.0 |
| 150W | none | 1:35:36 | 7.7 |
| 150W | 150W | 1:31:16 | 8.0 |
| 150W | 75W | 1:34:11 | 7.8 |
Compare these speeds to solo riders on a closed course:
| Rider 1 | Time | Speed (kph) |
| 300W | 49:28 | 14.8 |
| 225W | 1:04:44 | 11.3 |
| 150W | 1:35:36 | 7.7 |
Observations
Overall, these speeds are close to the old banded Meetup speeds, but not the same. They’re a bit slower for the higher power numbers, but a bit faster for the 150W tests.
The speed of a solo rider in a banded event now perfectly matches a solo rider free riding, just like in our flat tests. That’s a definite improvement.
Unlike the flat test, there doesn’t appear to be a speed “floor”, based on our 300W tests. You’ll see as the following rider’s power decreases, the climb time increases as well. This feels realistic to us, so no complaints here.
One rider holding a particular wattage vs two riders holding that same wattage delivers somewhat inconsistent results, with the single rider being faster in our 300W test, but the pair being faster in the 225W and 150W tests.
Conclusions
The keep together changes rolled out in Zwift’s latest update are a big step in the right direction, delivering (in our opinion) the best “banding” experience in game to date. Speeds vary more realistically based on rider power, they’re consistent across different types of banded events, and they’ll be overall faster for Club Events and Group Workouts, which is a change we think most Zwifters will endorse.
We have just two banding-related recommendations left. First, Zwift should mark banded activities as “gamified” in Strava the same way they mark Climb Portal efforts that are scaled to 50% or 75%. Because nobody is doing a banded event to set a segment PR on Strava (or if they are, they shouldn’t be allowed to). Having these efforts displayed on Strava leaderboards just confuses things.
Additionally, Zwift shouldn’t save in-game Leaderboard timings for banded events, although displaying them in the actual event would be fine.
In conclusion: nice work, Hugo. Ride on!
Your Thoughts
Have you tried Zwift’s updated keep-together algo yet? Share your thoughts below!
Definitely agree that they should be labeled gamified.
Still not sure why having a rider drafting you would slow you down.
It reflects reality….if we are riding with a slower rider, we have to slow down to accommodate….just like a “banded” ride.
Cuz it isn’t drafting it is “together”. Let us pretend the parens are the rubber band. (100+100)/2=100 then (100+20)/2=60; voila…in simple terms…
Should consider doing another series of tests based on the size of the group. Always wondered how the group size influences speed.
With these numbers, I’m OK with doing rubberbanded events again – especially the long distance rides.
I still LOVE when riding the Banded Rides, and you get the
Rider that still goes for the Sprint, and or has not figured out
what a Banded Ride is… And they are Hitting 10Wkg’s, and
can’t figure out why they’re not going any where 🤔😂🚴
He says sprint prs should not be allowed (agree) but speed DOES seem hard to generate anyway just as you say. I don’t see PRS being set on my banded rides with sprinting still being done.
I should say I don’t see prs set by the strong leaders perhaps weaker riders are made to look better, actually why I ride/spront much less hard
Can you run a speed test where one of the riders is on a MTB? Does this slow the group down even if they are producing the same power output?
I was in a 90 minute banded ride doing the Mountain Mash. We were all riding together nicely until the splits started to happen a little over the third of the way in. First two separate groups, then three, then four. Everyone seemed to stay together in their own separate group but it was kind of funny seeing your ride leader or sweep passing you in the opposite direction in certain places.
Kind of a shame that they didn’t do these comparison tests before the faulty releases and then leaving it out there for quite a while until a 3rd party intervened. You are providing a free QA service (thank you).
Sure it is banded ride so would mean little but curious what the positive and negative times of other riders means -if anything.