Increased power but same speed

Anyone experience or know about a bug where you increase power output but your speed barely changes? It’s happened a couple of times now to me. The most recent was on the Tempest route just approaching the route end. Increasing from ~150W to 635W only resulted in a .1kph increase.
You can see it at about 89:23 into this activity. I increase power at 90:11 but speed stay almost identical. activity/944267528413134848 MacOS 12.0.1, Ant+, Tacx NEO 1.

I see here also inconsistent power/speed.
But there is something more fundamental what I don’t understand. NEO Tacx can give besides power and cadence also speed and more, all via Bluetooth at the same time. But with Zwift you have to choose between either power or speed (which is then taken via ant instead of bluetooth). IF power is chosen the Zwift calculates in some way the speed from power (and more parameters). Why? Speed is available directly!
I have made a simple Tacx controller, and I can extract power, speed and cadence, and the numbers all make sense without any calculations by myself.
With Zwift, if power is chosen the calculated speed while not peddling goes to 0 although the Neo is still spinning.
If speed sensor is chosen then, after stopping peddling, speed stays high until Neo nearly stops spinning, and also power goes slowly to 0.
Doing the same with my controller shows exactly what I expect: stopping peddling power goes to 0 very rapidly, and the shown speed reflects the spinning of Neo.
So: Why doesn’t take Zwift the actual power AND speed from Neo?

In-game speed is different from trainer speed. For example, you can be in super-tuck going 57 km/h on a virtual descent while your trainer is at rest.

Zwift physics are driven by your power output. The only reason that Zwift accepts speed as input is for those who have a speed sensor but no power meter. In that case, Zwift will estimate your power from your speed.

1 Like

I’m not sure what super-tuck is. But if I am training on my own, so not in a group, then it seems to me a lot of sense to take speed and power from tacx instead of trying to calculate one from the other. Going 57 km/hour in a virtual descent while trainer is at rest seems to me unrealistic, but maybe makes sense in some game-circumstances. Unless you mean by ‘trainer’ the person; tacx which will run by itself if track is going down. No need to play with speed numbers, I would say. At least it would be nice if the user can choose to take speed AND power from tacx.

Super-tuck in-game is when you stop pedaling and your avatar is going fast enough on a steep enough downhill grade, the avatar will go into an aerodynamic “super-tuck” position and will coast downhill at a high speed.

Go to the top of Alpe du Zwift, start a descent, and get off the bike. It will take around 10 minutes to get to the bottom of the hill. (12 km or so at about 65 km/h average.) This is realistic, but I guarantee that your Tacx will not report the correct speed for that descent. It doesn’t have enough information about the virtual world.

OK, I’ll try that, to see what happens. Tacx “knows” the slope, and you need to add the right rolling resistance etc, but there are probably Tacx limitations.

Zwift doesn’t change resistance with rolling or aero resistance, only slope. Your in-game speed will almost never match your flywheel speed.

Tacx only “knows” the slope if you set the in-game trainer difficulty slider to maximum. At lower levels, Zwift asks the trainer for less than the true virtual grade. (At the minimum setting, Zwift always asks for a 0% grade, regardless of the virtual terrain.) So, this is another reason not to take speed input from the trainer. It can’t work at less than maximum trainer difficulty.

I realize that descents are unfair to the Tacx, because Zwift asks the trainer for at most 50% of a downhill slope to reduce spin-out. Hence, your Tacx never “knows” what a downhill grade is in Zwift.

Still, I don’t see how the Tacx can possibly figure out your speed when you’re virtually freewheeling. If it knew your weight, the virtual bike’s weight, and the CdA, yes, it could do it. But it doesn’'t know these things.

Tacx indeed doesn’t have all parameters like weight.
I did try different slopes to see what Tacx is doing (using my own program); and it appears that Tacx has a limitation of 36.7 km/hour freewheeling, which it reaches at just 4% slope (with my resistance settings which may be not realistic). Making slope more negative doesn’t change that, only peddling fast increases speed.
OK, thanks, that makes things more clear.

There are other in-game mechanics affecting velocity that aren’t communicated with the trainer: aerodynamic resistance (which includes drafting, frame & wheelset selection, power-ups, rider height, etc.) and rolling resistance (tarmac vs gravel vs dirt, MTB tire vs gravel tire vs road tire), for instance.

One could reasonable argue that these things should affect how hard you have to pedal, and therefore should be communicated to the trainer, but there are no standard protocols for that.

Yes, you could offload all of the game physics to a smart-enough trainer. The Tacx isn’t smart enough.

Then the question is what Zwift tries to do: give the user as much as possible the feeling of a real ride, with slopes, aerodynamics, etc. etc.? In that case, Zwift should control the not-so-smart Tacx as good as is possible. And if the resistance for the user is about realistic then the speed will “by it self” also be OK, except for downhilling. But probably I am overlooking something. Would be nice to have more info about these calculations, but that is maybe proprietary…


I agree. The draft effect should reduce trainer resistance. Rolling over dirt should increase trainer resistance. Riding a mountain bike should increase trainer resistance. Etc. But, for some reason, Zwift has chosen to just change the virtual speed calculation instead.