Power data Kurt Kinetic vs. other supported trainers

(Nigel Van de Velde) #1


I’ve been testing Zwift for over a month now and have to admit that I really start to like it. I hope it keeps evolving into something even better!

I am riding on a Kurt Kinetic Road Machine and using a speed/cadence sensor for virtual power/z-power. Back when I started there were quite some topics about its power readings being rather low and irregular. I even posted some CP-curves comparing data from zwift to readings from other software like goldencheetah. I hoped that this would have been addressed before getting out of beta.

Unfortunately I didn’t experience any significant improvements on this. Power readings are still quite low (in comparison to virtual power in other programs) and very irregular.

Today I tested what power readings I would get when selecting the other supported turbo trainers on Zwift (still riding my Kurt Kinetic of course). I used 2 ANT+ devices, one for GoldenCheetah and one for Zwift. I tried to have a constant power output of +/-155W on GoldenCheetah and compared with Zwift afterwards. As I kept my power output constant, the power profile of each trainer is irrelevant and Zwift should give me a constant (yet different) power reading as well. You can find the data (both power distributions and power data) over here:


The data series in pink are virtual power from GoldenCheetah, while the data series in other colors are the respective trainers on Zwift. I didn’t care about accuracy and just looked at precision. To my big surprise the Kinetic is the only trainer which lacks precision. You can clearly see that its power readings fluctuate a lot more than for the Cycleops and Travel Trac trainers…

I really think that getting the power profiles sorted out, and that way create a level playing field, should be of primordial concern before leaving beta. Otherwise this will be a huge dealbreaker for (I guess) a lot of us.

Really hope to see some improvements soon, keep up the good work!


(Maurizio Gigliotti) #2

+1 more precision.

(Mat D.[X]) #3

I have used virtual power for kinetic road machine and trainer road app. Now I’m using a Quarq and can assure you that virtual power reading for the road machine in trainer road is 5% higher if not more than my quarq.

(Robert Lockhart) #4

Virtual power does have many variables such as how tight you turn the knob on the Kurt Kinetic and what tyre and pressure you use. Its fine for personal training if you keep it consistent but when riding against other power users it must be very difficult to make it a level playing field. Even power meters can be badly calibrated and inconsistent.
I matched the virtual power on Trainer Road with my Stages power meter. This way I had a consistent setup whether using my outdoor bike with a power meter or my indoor bike with virtual power. I adjusted the turns of the Kurt kinetic knob till the numbers matched.
I then used Zwift and the virtual power numbers are much lower than the equivalent on Trainer Road. The power readout also jumped up and down a lot on Zwift. Although there may have been a recent update that fixed this?

(Nigel Van de Velde) #5

@Mathieu: My post is not about accuracy, but about precision. The readings fluctuate a lot more for the kinetic than for the other trainers.

Concerning accuracy, I can’t say if virtual power is more accurate than z-power, as I don’t have a power meter. I did experience that virtual power is pretty consistent with speed and heartrate data from my outside (flat here in the northern part of belgium) rides. Anyway, a 5% error in comparison to real power as you state, would be wonderful actually. Combine this with improved precision and some z-power magic for accelerations, then I think a paid zwift subscription would be a no-brainer.

(Maurizio Gigliotti) #6

Example :

    time power cadence speed

51 0.36 159 83 29.0628
52 0.36819 161 83 29.1024
53 0.3763 173 84 29.1204
54 0.38453 174 85 29.376
55 0.39286 174 86 29.6172
56 0.40182 212 89 33.066
57 0.4112 163 88 33.5448
58 0.42071 164 87 33.9588
59 0.43056 164 86 35.334
60 0.44061 204 90 36.0252
61 0.45078 232 91 37.0116
62 0.4613 225 92 37.476
63 0.47188 225 94 37.44
64 0.48231 224 96 36.234
65 0.49214 251 96 34.344

(Robert Lockhart) #7

In terms of precision riding Z-power on my Kurt Kinetic I was getting about 75 watts of variance riding smoothly at the same cadence.

(Craig Martin) #8

Can I make a suggestion? Since you have 2 ant+ why not use the trainerrelay feature of trainer road to supply more accurate power data to zwift? I don’t say it’s a long term solution and hopefully zwift are as skilled as trainer road in mapping the power curve of various trainers but until then… In my experience (admittedly lemond revolution not kinetic) trainer road is a damn near perfect match for real world road power

(David Nixon) #9

For what it’s worth, I found that on my Cyclops Fluid2, Zwift was much closer to actual power than TrainerRoad. TR was 40 watts too high when I borrowed a bike with a Quarq that was showing 120w. The curves got closer together as you approach 320 in a fairly linear fashion.

(Eric C. (Zwift HQ)) #10

Regarding the number fluctuations:

Z-power fluctuates more than virtual power because we take acceleration into consideration. Virtual power tends to have a slower-reacting, smoother acceleration curve, hence the smoother numbers.

With each pedal stroke, z-power reads that acceleration and outputs the according number, so even if you feel you are riding with the same cadence, there are slight changes to your speed that we are taking into account and z-power reacts immediately to them.

We will continue to work on smoothing out that reading so it’s not so jarring.

Regarding the power readings:

We initially compared (in real time) our z-power readings on the Kinetic (as well as the CycleOps, etc…) with a properly calibrated power meter. That’s the basis for how we start to generate that algorithm. It will never be 100% accurate, but it’s within a tight percentage with the proper tire PSI and about 5 cranks of the Kinetic handle.

(Nigel Van de Velde) #11

@Eric: thanks for the response.
I understand the differences between z-power and virtual power from a previous discussion, but i don’t think that’s relevant here.

My tests were done at low power without accelerations. Furthermore I can’t understand the huge difference in fluctuation between the kinetic and the other fluid trainers. If the fluctuations were to be inherent to z-power, than you should see similar fluctuations for all of the trainers, right? I really think you should look closer into this once again.

@Craig: That’s what i’m doing right now. I display and record virtual power in goldencheetah simultaneously and even replace z-power with virtual power before uploading to strava. That way my data are still comparable to all my old indoor rides. Too much of a hassle to keep doing this however.

(Maurizio Gigliotti) #12

i try with simulant+

cadence 90rpm
speed 40km

zpower 420w - 428w every second