Speed Tests: Rubberbanding (Keep Everyone Together) Mode in Zwift Club Events

UPDATE: Zwift greatly improved their banding algorithm after this post was published. To see current “keep together” test results, read the updated post.

Have you ever done a “Keep Together” ride in Zwift? In this mode, all riders in the event are kept in the same group, regardless of the power you’re putting out.

“Rubberbanding” makes it easy to keep a group of cyclists together, and that’s a good thing, because any ride leader will tell you that keeping cyclists together without some artificial help can feel like herding cats!

A Short History of Rubberbanding

Rubberbanding or “Keep Everyone Together” mode has a colorful history on Zwift. It was first introduced when group workouts arrived in 2017, then quickly throttled to lower speeds after lower-power riders complained they were setting segment PRs in workouts.

For a couple of years, rubberbanding was only available in group workouts, and group workout packs moved notoriously slow. But then “Keep Everyone Together” was added as an option for Meetups in October 2019… and all hell broke loose.

The problem was that rubberbanded Meetups had a major speed bug. Groups would speed up over 45kph for no good reason, topping Strava leaderboards (see Box Hill as one example). And this went on for much too long (three years or more).

While rubberbanded Meetups were zipping around at silly speeds, Zwift rolled out Clubs, and eventually launched Keep Together Mode for Clubs in November 2023, along with updated “Keep Everyone Together” logic that replaced the speed-bugged logic used for Meetups.

Testing the Rubberbands

So how does Zwift’s “Keep Everyone Together” logic work? I’m not entirely sure, and Zwift isn’t revealing the secret sauce. So I decided to run some tests using rubberbanded Club events to find out what speeds rubberbanded riders reach at different power levels. For science!

The question I wanted to answer was: do rubberbanded speeds 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 Club event with “Keep Everyone Together” mode enabled. Here are each rider’s power outputs, segment times, and average speeds.

Observations

The first thing that stands out here is how slow the times are, particularly when the riders have matched power numbers. I know from other tests that solo free riders complete the segment at these approximate speeds:

• 300W: 40.23 kph
• 225W: 35.95 kph
• 150W: 30.81 kph

Why are rubberbanded riders moving so slowly, when they’re both holding the same power? Clearly, something in Zwift’s Keep Together logic is artificially slowing the riders.

The other odd thing is how the pair 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 Club event still traveled as slowly as the paired riders above. So I created an event with “Keep Together” enabled, but only had one rider join. Here are the results:

Observations

First, notice that when there was just one rider in the rubberbanded event, their speeds were still much slower than they would be when free riding. That’s goofy.

Also, notice how the 300W, 225W, and 150W solo tests all came in a bit slower than the paired rider tests, even when the slower paired rider was doing very low power. For example, a solo rider at 300W traveled at 30.58kph, while two riders holding 300W and 75W traveled at 31.73kph.

Shouldn’t the pair with a very slow second rider move slower, as the rider holding just 75W slows down the overall group?

The 1000W test is also interesting, since it’s sort of a “max speed” test that shows just how slow banded rides may be. In rubberbanded mode the rider only reached around 47kph, while in free ride mode the same rider holding 1000W reaches around 62kph. Or another way of looking at it: ~460W in free ride mode will get you the same speed as ~1000W in “Keep Everyone Together” mode. Sorry, sprinters.

Other Rubberbanding Discoveries

In the course of these tests, I also asked and answered a few Keep Together questions I’ve seen:

• When does the elastic snap? When a rider drops below 10W, they’ll drop from the group.
• Can you rejoin the banded group? Yup. In my tests with two riders, if a rider dropped back about 30 meters and then increased their power to 10W, they would rejoin the group. If they dropped back 40 meters before increasing power, they would not rejoin. (This “rejoin window” may differ based on group size, like it does for RoboPacers.)
• If a rider is lapped, can they rejoin the group even though they’re “behind?” Yes. Same behavior as noted above.

I also noticed a few things that surprised me:

• Coned riders disappear: one of my bots got the “cone of shame” for holding high power for too long. When that happened, the bot disappeared from the group ride. Well, sort of. It was more like a shadow ban: the shamed bot could still see the other rider, but the other rider didn’t see the shamed bot anymore. This is actually pretty cool, although perhaps Zwift should notify the coned rider that they’ve been hidden. I’m guessing this is how Zwift now treats coned riders in all events.
• Pairing screen doesn’t stop you: if you’re pedaling along in a Keep Together ride and enter the pairing screen, your avatar doesn’t slow. This is counter to pairing screen behavior everywhere else in game, where entering the screen is like hitting the brakes!
• Except when it does: interestingly, if both riders in the Keep Together ride enter the pairing screen, then they both stop.

Concluding Thoughts

Keep Together mode has always been popular in Zwift because riders want to be together, and the magic of Zwift allows riders even with very different abilities to go hard but stay together. That’s a big deal! So even when it was broken and zipping people around at silly speeds, people kept coming back because cycling is a social sport, cyclists are social people, and Zwift is a social platform.

Yes, rubberbanding is popular. And yes, Zwift fixed the algorithm so land speed records aren’t broken anymore in banded Meetups. But the results above show that Keep Together mode still has plenty of room for improvement. Specifically, banded Club rides are moving too slowly, and speeds aren’t fluctuating realistically.

I’ll admit these tests are very limited in scope, and it’s possible that the banded experience is better in larger groups. But I also don’t think that’s the case, as banded rides I’ve taken part in have felt much too slow, and other Zwifters have said the same. Plus, the experience needs to be good in groups of just two riders, because it’s often just pairs of Zwifters who want to ride together!

Coming Up Next

Soon I’ll perform similar tests on Alpe du Zwift, to see how rubberbanding behaves on long climbs. I’ll also be testing rubberbanded Meetups to see if they behave differently than Club events. Stay tuned!

Questions or Comments?

Share below!