Power meter for races

So does that mean we all have to buy a Wahoo Kickr V5?

Wahoo have intentionally increased the power that trainer reads to compensate for drivetrain loses. This immediately makes this the only trainer a serious Zwift racer can use. Look it up.

This seems heroically idiotic. Are we then going to all have to buy the next trainer that universally overreads power?

Further, these trainers are a million miles from dead accurate. Many, do not have actual strain gauges. They use a rudimentary flywheel speed equations. They also have poor temperature compensation. Short story, they are far from accurate.

Now, some us own significantly more accurate power meters. Be that Favero pedals or spider based PMs like SRMs. It seems utterly illogical that we should be using the far less accurate devices to measure our power output.

Many serious cyclists that use Zwift for training, but also ride in the real world, prefer to use the same power meter for consistency.

Being forced to use the random number generator (insert most smart trainers here) to race is backwards logic. Particularly, if Zwift is actually intending to eventually be a fair racing platform.

If this isn’t addressed we are going to see a continual escalation of inflated power metrics by smart trainer manufactures. It’s already begun. Literally, any smart trainer manufacture releasing a new model now, has to calibrate it to match or further overread what the current Kickr V5 is reading. Otherwise, not a single racer could buy it.

So, I’m all for the dual recording, however all users should be using an actual power meter for power measurement. Not a cheap speed sensor masquerading as a measurement device.

Personally. I use my Favero pedals for all my riding. If I used my poorly designed Wahoo Kickr Core, I’d sometimes be World Tour or sometimes be weak, depending on it’s poor temperature compensation.

So, going forward @xflintx should we not be using the far more accurate devices many of us already own to control the gaming? Or, are we just going to accept that power, the one athlete generated metric that controls all of the racing, is never going to be calibrated for any kind of accuracy.

So, going forward @xflintx should we not be using the far more accurate devices many of us already own to control the gaming? Or, are we just going to accept that power , the one athlete generated metric that controls all of the racing, is never going to be calibrated for any kind of accuracy.

Hopefully if Zwift is improving the support for dual power they will both record both sources and use the most accurate.

I suspect that actual power meters generally will be more accurate than trainers (although there probably are exceptions.)

I’m assuming that the reason Zwift requires the use of trainer power and you record your Power Meter separately (e.g. via your head unit, aka Garmin 830 etc) is that the Zwift client is simply not currently capable of generating a second fit file for the trainer power when you use a power meter as your power source.

If the Zwift client can save two FIT files then there is no need to require that the power come from the trainer and not the power meter.

Kind of squinting sideways a bit at the last (second last?) post from Flint sort of implies that Zwift is looking at this.

I had a hard time finding the source for this, if someone has an easy link handy please share - is a recently calibrated kickr core not accurate? My drivetrain is pretty shot to begin with, but I’d be interested to know how much variability they use outside of their +/- 2% delta stated.


Why should they? There are many tests showing the best smart trainers are comparable with PMs (dcrainmaker, gplama).
I have a P2M NGeco (spider PM) and an “old” Tacx Neo - did much testing (double recording) and the difference was not more than 2.5%. Slightly better with freshly cleaned drivetrain, but I talk here about max. ±0.5%.

Edit - off topic here, sorry.

1 Like

There’s 1% between my Faveros and my Kickr and actually the pedals typically are up 1-2% on the Kickr above threshold, especially at sprint power, and down 1% below it. Many other dual recorders have the same or similar kit and there’s rarely much in it. So I think you’re exaggerating, a lot

Also, yes I want consistency with outdoor power, which I why my pedals pair with my Garmin where my training data goes (for Xert). Not the bit that drives the computer game.

You are never going to fix the power meter accuracy when you have such a diverse set of products and users. What you need & should strive for is consistency by users.

If a device is over\under reporting (strange how we think no power meters ever under report) by a few % in the grand scheme of things this isnt an issue - as long as its consistent.

Trainerroad discussed this a while back, and they said even premium brands of trainer \ power source will see a variation from device to device, never mind looking at multiple brands trying to gauge power from different parts of the drivetrain all using different methods.

They laughed at all the 1-2% accuracy claims and said this was marketing BS.

Until there is a single agreed way of calculating power with an ISO its a difficult place to go.


Source please. I tried looking it up with no success.

1 Like

If Zwift could run some kind of anomaly detection, e.g. by machine learning, they could tag users with inconsistent equipment and prevent them from entering certain races.

1 Like

I use the default scaling on my 4iiii Precision crank meter, because I’ve no idea if I have a leg power unbalance, but there’s nothing to stop users changing the setting to say they are heavily right leg dominant using a left crank meter.

No idea if this is unique to 4iiii

Pretty much the same for power averages and NP between my Kickr and Power2Max PM, although I’m not sure if its my imagination but the PM seems to respond faster to abrupt power changes, almost as if the Kickr is doing some power smoothing independently of the software,

I assume you’ve checked they are both on the same moving average interval? typically 3 secs?

Yes, the Zwift smoothing setting was the same, its the impression I get of how quickly the power figures respond to increases and decreases in power although its hard to verify as there can only be one device set as the power source in Zwift. The same is true for going into supertuck mode - the PM seems responds much quicker to 0 power.

using the default scaling on my 4iiii puts me about 10% or more at times lower than my elite direto (calibrated about 2 months ago to within 2 points of the offset) by adjusting it to 53/47 in favour of the left it is much closer in terms of numbers, my summer bike is using a 4iiii also on default scaling and the numbers are slightly higher than the turbo for certain times such as 5s and 15s, will need to test it on longer climbs to see about 2-5 minute efforts

How would they know which is the most accurate?


If mine, the higher number is more accurate, everyone else, the lower number. Its fairly logical principle.


Of course, what was I thinking??


I guess in theory Zwift could have a list of accuracy levels for each trainer and power meter and use the “most accurate”.

But that doesn’t account for calibration/zero offset, or for the fact that power at the pedals will always be slightly higher than that at the hub due to drivetrain losses, or for L/R imbalance in single-sider power meters, etc etc.

It’s a massive can of worms that I struggle to see the benefit of opening.

Outside of a big variation in a TT or a long hill climb like AdZ, 1-3% is largely going to be lost as a drop in the ocean in the 20-30% efficiency variation of the churn of blob’s draft anyway. I take racing somewhat seriously as I’ve disclosed and I dual record and occasionally glance at my competitors dual recordings. Even caring that much I don’t bat an eyelid at a small %. As said, it’s a can of worms to get too hung up in. I’m cool with it as it is enforced by most reputable race organisers today, if you want a decent placing in an A race to stand, be ready to upload dual recording with HR that’s close enough.

dual recorded a race Monday 5s power was 5w lower on the 4iiii difference, 15s was identical, 1min was 13w lower on the 4iiii, 5min was 19w lower on the 4iiii and 20min was 17w lower on the 4iiii

I’ll put some of the differences down to when the garmin was started with the 4iiii

Wouldn’t it be fairly easy to change one FIT file to match another? I see that there are a lot of FIT tools around. Maybe the file has a checksum for verification.

If two sources are going to be required, Zwift will have to connect to both and automatically store the data.

1 Like