Suggestions for Improving RoboPacers and Trainer Difficulty

RoboPacers
1

I would like to suggest three improvements to the RoboPacer system.

First, I think Zwift should add a 1.3 W/kg RoboPacer between the current 1.1 and 1.5 W/kg options.

This is not simply about giving existing riders another choice. I think a 1.3 W/kg category could play an important role in expanding the base of road cycling.

For example, it could provide a natural step for beginners moving up from 1.1 W/kg, while also being useful for riders returning from an injury or older riders who want to ride at a sustainable pace. Having a 1.1 → 1.3 → 1.5 W/kg progression would provide a much more gradual path for riders who are not yet ready for 1.5 W/kg.

Second, I think the actual W/kg of a RoboPacer should stay within a reasonable range around the selected category, especially on climbs.

I understand that a RoboPacer should not behave like a simple stationary exercise bike. Some automatic variation in pace is necessary to manage the group and to make the pacer feel more like a human pacemaker.

However, I experienced something different today. I selected the 1.5 W/kg RoboPacer, but on climbs I was already riding at around 1.7–2.0 W/kg and still found the RoboPacer moving away from me. This made me wonder whether the RoboPacer was temporarily riding at something like 1.8 or even 2.0+ W/kg.

I understand that the actual pace may move somewhat above or below the selected 1.5 W/kg. But if a rider selects 1.5 W/kg, I don’t think the RoboPacer should temporarily reach a level that effectively belongs to a different category, such as 1.8 or 2.0 W/kg.

In other words, I am not asking for the RoboPacer to maintain exactly the same W/kg at every moment. I am asking for the variation to remain within a reasonable range around the selected category.

Finally, Trainer Difficulty seems to be an effective way to adjust the riding experience. However, I cannot specify Trainer Difficulty as a numerical value.

I don’t necessarily need very fine adjustment. Being able to set it in 5% or 10% increments would be sufficient. The important point is simply being able to specify the value I want.

So my suggestions are:

1. Add a 1.3 W/kg RoboPacer between 1.1 and 1.5 W/kg, creating a more gradual progression for beginners, riders returning from injury, and older riders.

2. Keep RoboPacer W/kg variations within a reasonable range around the selected category, rather than temporarily reaching a level that effectively belongs to a different category.

3. Allow users to specify Trainer Difficulty as a numerical value, with 5% or 10% increments being sufficient.

I believe these changes would make RoboPacers more predictable and useful, while also helping Zwift provide a more accessible path into road cycling for a wider range of riders.

2

The RoboPacers put out 10% more power on hills of 5% gradient or higher. When the road is flat there is no variation in power.

3

I don’t understand why a 1.5W/kg pacer would ride away from you on a climb if you’re doing 2W/kg even with dynamic pacing. At that power you should be moving ahead of the pacer unless the climb is basically not a climb and the speed is high enough for drafting, because Bernie will never reach 2W/kg. I guess if it slingshots through a group you could see that at the bottom of a climb if it enters with a lot of speed from the draft, but it would not be sustained because it won’t reach that power.

4

Thanks for the clarification, James.

I was aware of the 10% increase on climbs of 5% or more. What I experienced today, however, still felt quite different from what I expected from a 1.5 W/kg RoboPacer.

It is certainly possible that what I experienced was a temporary issue on my side, perhaps related to my connection or communication environment, so I don’t want to assume that the RoboPacer was actually exceeding the intended limit.

That said, even if the 5%+ gradient rule is working exactly as designed, I think there is still an important question:

Is a 10% increase in power actually appropriate for the 1.5 W/kg category?

For example, a 1.5 W/kg RoboPacer becoming 1.65 W/kg on a climb may be reasonable from a technical perspective, but for a rider who specifically chose the 1.5 W/kg category because that is their sustainable level, even that increase can make a meaningful difference.

My concern is therefore not necessarily that the current rule is malfunctioning, but whether the rule itself is appropriate for the purpose of the W/kg categories.

I think this is worth considering when looking at how RoboPacers should work on hilly routes.

5

Thanks for your reply, Paul.

Yes, that is exactly why I found the situation so strange.

I was riding at around 1.7–2.0 W/kg on the climb, and the 1.5 W/kg RoboPacer was still pulling away from me. So I also don’t understand how that could happen if Bernie really cannot sustain anything close to 2 W/kg.

As you mentioned, perhaps there was some temporary effect from drafting, the group dynamics, or even a communication issue on my side. I can’t say for certain what happened.

But I definitely experienced it, which is why I thought it was worth bringing up here.

Thanks again for taking the time to explain it.

6

There is a really long and probably boring topic about static vs dynamic pacing. It doesn’t sound like they are considering it. My own view is that pacers should be offered with both static and dynamic pacing because they will each appeal to different people. Neither one is wrong. For me the difference between 1.5W/kg and 1.65W/kg or even 2W/kg is nothing and that will be true for many people riding with Bernie. For someone riding at the limit of their ability that will be very different.

7

I see what you mean.

My own view was that some variation in W/kg is actually a good thing, as long as it stays within a reasonable range and doesn’t effectively turn a 1.5 W/kg pacer into a different category. I thought that kind of variation made the pacer feel a little more human rather than purely mechanical.

But I hadn’t really considered giving riders the choice between static and dynamic pacing. I think that’s an interesting idea. It would allow people who want a steady, predictable pace to use a static pacer, while those who enjoy a more natural, dynamic pace could choose the dynamic one.

Thanks for the perspective!

8

I think so. I also think it is normal behaviour on a climb and often riders struggle to slow down enough to stay with the pacer.

As it is 10% of power output it is relative to the amount of power. The difference from 1.5 to 1.65 is not a big jump, especially when there is more resistance on the climb. Having to do 2.0 to keep up should not be expecting unless you are significantly below the weight of the 75kg pacer.

9

You may be experiencing a robopacer attack at the bottom of a hill. It happens all the time. I have a good video of it below where I deliberately hold higher power than the robopacer ever does but he just shoots off up the hill suddenly before slowing down. I think what is happening is the gradient logic being applied to the robopacer is running a couple of seconds behind realtime, so he keeps moving at the same speed as he would on the flat, then the gradient suddenly kicks in. It’s much more noticeable when there is nobody else around like in my video and on steeper kicks where you get slowed quickly.

10

The one thing to keep in mind is the power differences are magnified for people who are much heavier than the pacer due to the way Zwift dynamics works on flat vs. hilly sections. If you set your weight to 100kg for example and try to follow a pacer that centers around your tempo on flats you will notice a large swing in your power on the hills. So the experience of someone at 75kg vs. 100kgs is much different with a pacer as a result.

Edit: Example using Yumi.. 2.9w/kg (217W) on flat, but 3.2w/kg on hills. If you are 100kg you don’t need that much more than 217W to keep up on flat given zwift dynamics (definitely not 300+W), but on the climbs you approach the need to do 3.2w/kg which is 320W.

Edit 2: This is ignoring drafting of course, but draft benefit is much higher on flat vs. climbs, so this is again magnified assuming the 100kg rider is always riding in the draft on flat and hills.

11

As a 90+ kg zwifter, you can probably do less than 217 to keep up on the flats in the draft, then you get dropped on the climb as the pace increases.

12

We’ve also been told that Zwift is at its max for robopacers and adding more has caused all sorts of issues (they know this becuase certain events have used robopacers as ride leaders and when the extra are added problems arise). Adding a 1.3 w/kg pacer would require removing one of the other ones.

13

One of the world record attempt riders had his own custom red robopacer on Tempus Fugit telling people to maintain a steady pace in that group, so just one or two should theoretically be possible. Nobody complained about the steady pace messages either.

Otherwise they’d just say, no, we cannot do that, you are on your own with your record attempt.

14

If they were motivated to make this better they would figure it out. I accept that there are some changes they would need to make, but those are solvable problems. The problem is more about making the decision to work on it.

15

Brian, this looks very similar to what I experienced today.

The main difference is that the climb I experienced was much shorter.

After we entered the climb, the RoboPacer gradually moved through the group and came out ahead at the front, while I fell about 10 meters behind. Because the climb was short, we entered the downhill almost immediately afterward.

I’m not sure whether the RoboPacer’s “attack” was still continuing at that point, or whether it simply entered the downhill ahead of me and therefore carried more speed into it. Either way, by the time I reached the top of the climb, the gap had grown to around 20 meters.

About half of the riders in the group noticed what was happening and managed to follow the RoboPacer by increasing their effort, while the other half fell back just as I did.

So yes, the behavior in your video looks very similar to what I experienced.

16

One thing is that robopacers always have much smoother power than humans, we have to shift gears and always lose a little bit of momentum when we do that.

17

I didn’t know that adding another RoboPacer would require removing one of the existing ones.

Personally, if there is a limit to the number of RoboPacers, I would rather have one RoboPacer at each W/kg level with smaller gaps between the levels, even if that means having only one route for each intensity.

For example, if there are currently two 1.5 W/kg RoboPacers on different routes, I would prefer to have one 1.5 W/kg RoboPacer and use the other slot for a 1.3 W/kg RoboPacer.

For me, having more options in intensity is more valuable than having multiple routes at the same intensity.

18

I encountered another strange RoboPacer behavior today, which may be related to what we have been discussing here.

I encountered another strange RoboPacer behavior today.

During a long downhill section, the 1.5 W/kg RoboPacer gradually pulled away from me. By the time the descent ended, Zwift showed that I was about 50 m behind and told me to “Catch Up.”

I thought something was wrong, so I kept pedaling hard using the momentum from the descent. For a short period, I was producing around 2.4 W/kg.

After the descent, the gradient was 0%. The RoboPacer was supposed to be riding at 1.5 W/kg, while I was still riding at around 2.4 W/kg. However, even after riding another roughly 200 m, the gap barely changed — it remained around 48 m.

At that point I stopped because it simply didn’t make sense. If both the RoboPacer and I are riding on a flat section, and I am producing significantly more power, I should be closing the gap.

I can’t completely rule out an issue with my equipment, but my power reading appeared normal throughout the incident, including the 2.4 W/kg shown on screen.

This makes me wonder whether there could be another issue with the RoboPacer’s speed/position calculation or synchronization, especially during and after descents.

19

In my experience, these strange RoboPacer behaviors seem to be strongly associated with changes in gradient.

20

One more observation from my experience: these strange RoboPacer behaviors seem to be strongly associated with changes in gradient.

I don’t have detailed data from my early Zwift rides, so this is based mostly on memory. However, ever since I started using Zwift, I have occasionally experienced the “Wait… what? Oh…” moment when entering a climb, or just after a climb ends.

I don’t remember experiencing this kind of behavior on Tempus Fugit, where the course is essentially flat.

So I’m starting to wonder whether the gradient itself, or more specifically the transition between different gradients, may be a key factor in these RoboPacer anomalies.