PeteHski

I was expecting around 33-34 kph @ 210 W. There doesn’t seem to be many parameters in the app, but it does know my height and weight. On the Kickr bike fitting page it does give the options of race and endurance fit and I chose endurance, so maybe that adds some aero drag in the model. I will experiment over the weekend. I will also check if cadence matches the simulated gearing accurately for the wheel speed.

PeteHski

So I finally did a comparison of my Kickr Bike output vs Zwift on a 2 hour endurance ride on mixed terrain. Unfortunately I can't overlay the data directly at the moment, but there were some obvious differences:-

1. Zwift speed on the flat was around 6 kph faster than the bike was reporting. Roughly 36 vs 30 kph at a steady 200 W. I do think the Wahoo aero drag is a bit high, while Zwift is slippery low.

2. Zwift speed when climbing was much closer. Zwift was reporting around 1 kph faster at 16 kph and speeds were equal when climbing at 10 kph at a steady 225 W.

3. When descending, Zwift reports only half the slope gradient to the bike (with trainer difficulty set to 100%). So whenever descending, the speed reported by the bike is much lower than Zwift. I don't know why Zwift do this, but it's a well known thing.

4. The bike accelerates much faster and decelerates much slower in Zwift compared to what the bike reports. This applies on all gradients, not just descents due to the slope discrepancy above.

5. Due to the above effects average speed reported in Zwift was 10 kph higher (30.5 vs 20.5 kph).

6. The simulated bike gearing matches the Kickr bike speed as you would expect.

7. The average power reported by Zwift was much higher (183 W vs 145 W) which I expect is due to the "sticky Watt" issue whenever you stop pedalling for a few seconds. Power reported when constantly pedalling appears to be consistent. The power reported by the bike is in line with what I would typically see for a similar outdoor ride at the same perceived effort. Zwift average power has always looked high for me when riding in sim mode.

8. Average cadence was reported the same in both

RChung

Hmmm. I expected there to be some differences but those are huge. Faster to accelerate and slower to decelerate are consistent with the much lower aero drag component, but the magnitudes are surprising.

Have you ever looked at the implied virtual elevation profiles?

PeteHski

The aero models appear to be at opposite ends of the scale. The Wahoo bike seems to have a parachute, while the Zwift bike is super slippery. I was using a normal aero road bike in Zwift (Canyon Aeroad).

The elevation profile in Zwift looks as expected, but there is no elevation profile in the Wahoo app to compare.

choddo

That’s the quickest aero bike in Zwift isn’t it? Not that there’s THAT much difference. Or the manufacturers would pull their sponsorship.

PeteHski

Yes I think it is one of the quickest road bikes on Zwift. But I was really just clarifying that I wasn’t on one of the TT bikes, which I presume are even quicker.

RChung

I use speed and power to create a virtual elevation profile and compare that to the actual. Often that has turned out to work pretty well.

PeteHski

Unfortunately I don't have the tools to do that.

I did another test today with the Wahoo bike fit assigned as a "Race" fit. It made no difference to my Wahoo reported bike speed so I guess the Wahoo aero model is fixed. The only other parameter in Wahoo is my height, so I will see if that has any effect on the aero model.

This got me thinking about how trainer resistance is actually set when using Zwift and similar apps. I presume Zwift only broadcasts the slope value and the trainer then applies resistance based on its own physics model according to the broadcast slope value. That's how it seems anyway. I checked my simulated gear ratios too and they correspond perfectly to the Wahoo bike speed. Obviously Zwift speed is set by power and their own physics model, which in this case is much faster except when climbing at very low speed.

The Zwift ride I did today was the Ven Top climb. Average power reported by Wahoo was 201 W vs 206 W on Zwift. Much closer than my previous comparison because there was very little coasting. But Zwift did still manage to find me a free 5 W . Average speed on the climb was within 1 kph except for the flatter sections, where Zwift was as much as 4 kph faster at times.

I still think the Wahoo model is a bit slow compared to real life, but much closer than Zwift, which is silly fast on the flat and descents. But I always knew that was the case with Zwift and ultimately it doesn't matter how fast the Zwift avatar moves. It would be nice if the Wahoo physics model could be tweaked for aero and rolling resistance to match real life, but that doesn't appear to be an option. It does feel like I'm riding into a headwind on the flats!

Trakhak

In the old wind-up-watch days, they used to say that you always knew what time it was until you bought a second watch.

burnthesheep

Zwift is workout entertainment. I don’t take it seriously. It keeps me entertained indoors, so win.

Road bike riding on Zwift is super flattering for everyone under maybe a 6% grade climb. The aero of having sooooo many riders on a course is a semi infinite draft. I feel Zwift is super generous aero wise on the flatter terrain for high power high weight riders. As height and weight go up, the stack height or hip angle become rather non aero.

TT bike wise Zwift is sooo variable due to fixed bike fit. Real world I bet most folks aren’t that aero fit wise. Then again my TT bike in real life is nicer than anything usable in Zwift, including helmets and skinsuit.

Zwift cda well over .200. Real life it’s under it.

PeteHski

The generous avatar speed in Zwift doesn’t really bother me, but the “sticky” watts thing is pretty annoying, with a much higher reported average power on a rolling course profile. I didn’t realise how distorted this was until comparing average power reported directly from the bike vs Zwift data. But at least it’s fairly accurate on long sustained climbs.

In contrast, the resistance model on the Kickr Bike is a little harsh, which explains why I’m usually in a gear or two lower on the flat than I would be on my road bike, while my Zwift avatar is flying along! It’s not a big deal really and no different in terms of resistance from riding into a slight headwind or false flat.

It would be interesting to see how the Rouvy physics model compared to the Kickr Bike model. I expect they will be a closer match.

choddo

You almost never coast in zwift, compared to the real world, I think that’s why it has higher average power.

PeteHski

I wasn’t comparing average power against real world. I was comparing Zwift recorded average power against Wahoo recorded average power on the same indoor Zwift ride. But I do actually coast about as much as I do on my outdoor bike on similar course profiles. So on the rolling Zwift course I did plenty of coasting and soft pedalling and very little on the mountain climb. The Kickr Bike does a pretty reasonable job of replicating inertia on rolling terrain.

The only way Zwift can report a higher average power than what the bike records is if it manipulates the broadcast power data and I believe this is what some call “sticky Watts” where Zwift holds the power on for a couple of seconds whenever you stop pedalling or soft pedal. I was aware of this, but not the magnitude of it, which has the effect of boosting average Zwift power by 30-40 W on a rolling course.

When I compare the Wahoo recorded average power directly from the Kickr Bike with my outdoor power averages, they are very similar, which suggests a similar amount of coasting. I had always assumed that my Zwift average power was higher due to less coasting, but this is clearly not the case here.

choddo

Ah yes - that's indeed interesting. Sorry I think I was still quite heavily intoxicated after 2 Christmas parties yesterday when I wrote that reply.

RChung

Golden Cheetah will do it. (You can also roll your own in a spreadsheet, R, Python, Matlab, etc., but Golden Cheetah is easier).

PeteHski

Thanks, I was thinking of trying GC. When I get chance I will try to compare them properly.

chaadster

Is it? Could data sampling rate differences be having an effect? If the fast Race Mode option is enabled in Wahoo but you’re not connecting to Zwift app via wifi or DirectConnect, would the bike record natively at high sampling rate but Zwift still receiving at lower rate? Depending on which data packets made it to Zwift, it would be working with a different and incomplete set of data than Wahoo, right? Particularly when effort level is quickly variable, like a brief surge up a roller, there’s more opportunity for discrepancy between data sets. I dunno…but Race Mode is supposed to improve responsiveness in Zwift by minimizing data gaps from the trainer’s PM.

Even outside of Race Mode, wouldn’t a slow ANT+ connection to Zwift show data packet variance compared to the Wahoo recording (via Bluetooth to the phone)?

Wahoo also has the power smoothing feature for ERG mode, which if you have that on, may also be a factor.

PeteHski

I thought about sampling rates, but both Wahoo and Zwift are receiving data at 1 Hz via Bluetooth connections. I have Zwift connected through my laptop and Wahoo App connected from my phone. My Kickr Bike doesn't have DirectConnect for the Wahoo 10 Hz option.

I need to overlay the data sets (which I can't do directly at the moment) to see how the power data actually compares sample by sample. From what I can see looking at the Zwift data is that if I stop pedalling it holds onto the last recorded power and cadence for about 2 seconds before dropping to zero. But if I pedal continuously the data changes every second as you would expect. Zwift is definitely applying some sort of time lag whenever cadence suddenly drops or I soft pedal for a few seconds. It's not dropping data packets otherwise the drops would be completely random, which they are certainly not. If I pedal at a reasonably smooth cadence, I can see the power changing every second as expected.

The effect this has on average Zwift recorded power over rolling courses is a bit of a joke to be honest. The Wahoo recorded power data at the same frequency, over the same BT connection, gives a much more realistic power average in line with outdoor rides.

Edit: Also note that the average recorded cadence is identical in Wahoo and Zwift. It is just the power average that is significantly higher in Zwift.

chaadster

PeteHski Ok. I have Zwift on an older laptop that used BLE 4, but I use my BLE 5 equipped iPad because I thought the 2x faster data rate of BLE 5 reduced latency, but ai may be wrong about that, particularly since I have a ‘16 Kickr, which may be on BLE 4 and incapable of faster transmission anyway. Finding out that info an the Kickr seems hard.

Anyway, I’d like to upgrade to a new Kickr (or v2 Kickr Bike, really) and use the wifi connectivity for super low latency connection. I don’t see notable output wattage discrepancies between my old-school, “dumb” PowerTap bike numbers and my Kickr (almost exclusively on Zwift), perhaps because I never coast, so it’s not concern there driving me to upgrade, but rather the desire for crisper response in Zwift…well, and quieter operation.

Come to think of it, though, I have recently tried a few robopacer rides, really for the first time in earnest, and find it nearly impossible to stay with the group, either overcooking it despite really low power, or having wild power input demands that cause ridiculous yo-yoing off the front then off the back. There seems to be an art to it which I’ve far from mastered, or I’m just seeing latency more clearly since I’m in a bunch. Dunno. I tried races when they first came out too and had similar problems which I put down to a lack of fitness, but maybe it involved something else.

PeteHski

I don’t have a latency issue with my Zwift connection. It’s actually very responsive to sudden power surges. But it just seems to artificially hang on to power for a couple of seconds whenever I back off. I’ve read about this trait in Zwift with some trainers and power meters but a lot of the online discussions are a bit vague and inconclusive. The Wahoo recoded data shows that it isn’t the bike causing this issue.

Non of this causes any issues in erg mode or any continuous pedalling under power. Neither does it affect the ride feel in sim mode. But it does distort reported power averages on a Zwift ride or race, which is quite annoying as I feed this data into an adaptive training App.

chaadster

Okay…it’s just interesting that your issue sounds very much like the issue Wahoo has addressed with Race Mode and wifi connectivity:

“Slower data broadcasts can create gaps in power data transmission. Over the course of a race, data gaps can cause your avatar to surge and lag, making it hard to pace and draft effectively.”

https://support.wahoofitness.com/hc/...e-Mode-Feature

So yeah, it’s not the bike which would cause it per se, but rather data transmission rates.

Anyway, I hope you figure it out, because I know wonky data can be very frustrating. About 10 years ago, I tried CycleOps’ PowerCal HR style device which transmitted an ANT+ power proflle using an algorithmic power estimation based on HR. What a mess that was!

PeteHski

In my case it certainly isn't due to gaps in the data transmission. This latching of power only ever occurs when I soft pedal. It isn't random and it instantly picks up when I re-apply power. If I'm riding in a group I can soft pedal for several seconds without losing power (I have the Zwift power display set to instantaneous so I can monitor it in real time). There is no lag on slope resistance changes either. It's all very smooth and consistent with what I see on screen. Group riding in Zwift actually works well with my Kickr Bike.

I think I might try going back to Rouvy and see if it handles the power data differently. If the Wahoo app can read the data properly as broadcast then I don't see why it should be a problem for Zwift. I wonder if Zwift are applying this 2-3 second power latching filter deliberately to help users with poor internet connections?

79pmooney

All this brings me to the question -what are your goals? To be able to claim the highest sustained power? Win time trials? Post Stravas? Or win races? Mass start racing brings another element into the picture. Races aren't (usually) won by the rider with the most sustained power but they are always won by the rider who manages to get his front wheel across the line first. Coaches often train their riders to use lower gears and higher RPM than optimal for sustained power because that rider will have taxed his muscles less over the first 95% of the race. Burned more fuel, yes. But he still has muscle fibers that can power that big gear for the next 5 miles to the finish. The guy at the lower RPM who has been cranking that power more efficiently all race won't have that reserve. Yes, he still has fuel but not the legs to use it. Coaches have told their riders many times - if you can out spin your competitor all race, you can beat him at the end. (Now, we all have different optimum RPMS as others above have posted so that "rule" doesn't always work. A Jan Ullich outspinning a Lance Armstrong would NOT be a formula for Jan's success.)

And for us regular riders - if we do this (slightly lower gear, higher RPM) we can have that reserve when the big hill comes or we arrive at the town line sign. We will still have the muscle for that steep grade or the massive gear for a minute or two. Drawback - that slightly too fast RPM often isn't much fun. Slightly harder on crotch. Seats want to be better dialed in. The miles don't fly by as fast. (Our mental timepiece is altered by our leg rotation. Spin more and it takes more mental time blocks to ride that mile.)

rclouviere

good points. Goal is improved time trial.

PeteHski

In that case it is all about how many Watts you can average over your TT distance. So whatever cadence gives you the highest average power over the required distance.

Note that your optimum may not be a constant cadence over the entire distance. For example I do better overall (higher power average) by starting out at a higher cadence and then switching to a slightly lower cadence toward the end.