I did a some tests on Zwift a couple of days ago to check how accurately their speed calculations match what would be expected in real life.

The protocol was to use the London Classique route which has the advantage that you zone into a couple of short sprints. One is uphill the other downhill which adds another useful variable.

I did 2 runs on each of the sprints at a number of different weights ranging from 45kg to 90kg incrementing in 5kg steps. I attempted to keep power constant for each run and started each run in the same place before the sprint start so as to build up speed before entering the sector.

I kept rider height constant at 1.70m.

I captured 2 variables
Time was that displayed on the screen as I crossed the line
Power came from Strava for the relevant sector (since I was doing a flying start and wanted to pedal through the finish I figured that was more accurate and consistent between runs than using the lap button)

Gear was fastest TT setup Cervelo P5x with Zipp 858/super 9 wheels.

The file below contains the raw results. There are two sheets, the first for the uphill sprint, the second for the downhill sprint.

Segment data is from Strava
Uphill is 0.19km at 1.1%
Downhill is 0.18km at -1.4%
(sample run here https://www.strava.com/activities/5057250842

I have drawn my own conclusions from the data but do not wish to share them atm as Iâ€™d like to get some independent views just on the base data.

So hopefully there are some mathematicians and/or power geeks (hands up Iâ€™m one) who can take a look and say what they think.

The topics I am particularly interested in is how accurate is Zwift at estimating TT speed compared to real life and especially how does weight affect speed in Zwift TTs compared to real life?

Hope you can help and looking forward to hearing your conclusions

I have done no specific testing but in general I find Zwift speeds to be a bit fast on the flats. I think the coefficient they use for pavement is equivalent to what would be expected for pristine, new roads. There was also an article recently on ZwiftInsider about changing draft dynamics for more realism. Not sure these are enough to account for everything, though. I donâ€™t have any conclusions about dirt surfaces or climbing, or that it matters that much relative to the flats. I could just be slow IRL, too.

When you can do an Imperial Century in ~3:45 min with a group, I think it is safe to say that the speeds in Zwift (solo or in a pack) are not realistic.

And recently a whole bunch of people went well under sub-4 for the Uber Pretzel, which has a ton of climbing in it.

3 Likes

As an avid zwift racer I would have to ask, â€śWhy does it matter?â€ť

Out on the road and zwift are entirely different forms of riding and have unique circumstances applied to it.

1. No wind in Zwift - obvious impact

2. Height = cda - this is easily not representative of outdoors as people of significantly different heights can have the same cda based on position and also small variations due to inability to hold position perfectly all the time.

3. No change in o2 on the climbs - climbs donâ€™t have o2 level changes so not really realistic when comparing the impacts that that can/will have

Ultimately zwift physics care about Watts/cm on the flats and a mix of W/kg and W/cm on the climbs based on speed and acceleration. Could they be close in some edge cases sure but itâ€™s comparing 2 unrelated things

4 Likes

â€śUltimately zwift physics care about Watts/cm on the flatsâ€ť

I donâ€™t think it does. Hence the tests and hence why I asked for some folks to idependently check my data to see if they come to the same conclusions I did.

But you didnâ€™t test watts/cm? You just declared that it doesnâ€™t make a difference?

It is well known that height makes a huge difference in Zwift (5-10 watts equivalent for every 15cm according to Zwift Insider based on data from hundreds of automated tests). Iâ€™m not sure your tests add anything new to what Eric has already done.

Also your data (and Ericâ€™s) is very Zwift specific with no comparison to outside data, so Idonâ€™t see how you can draw any conclusions with respect to outside rides from the testing youâ€™ve done? What am I missing?

Note - Iâ€™m already very convinced that my times in Zwift wouldnâ€™t hold up against similar outside rides, but it also doesnâ€™t matter to me. They are 2 different things.

1 Like

Already tested and established by the standard bearer of Zwift testing:

Weâ€™ve done a lot of automated test laps with various height/weight/equipment setups, and confirmed this: all other things being equal, the taller you are the slower youâ€™ll go.

Also, they have some worthwhile info here:

3 Likes

CONCLUSION

While a Zwift ride will generally be faster than an outdoor ride of similar effort, the differences are minimal and not enough to invalidate including Zwift miles in your training metrics.

Hmmmâ€¦maybe for solo rides, but absolutely not for group rides. Not a lot of IRL groups knocking out 3:45 Centuries, I donâ€™t thinkâ€¦and absolutely not at sub 3 w/kg!!

1 Like

I agree, in disagreeing with the conclusion. He touches on real issues that will have notable impact on the different results, and I feel he disregards them incorrectly and washes his hands unnecessarily.

The simple fact that even when solo, the rider is in the drops once above a certain speed (skipping that search right now), when I would guess that most ride on the hoods in those same situations, will potentially have a HUGE impact on speed/effort.

Add in the drafting in Z with is super common unless you pick the TT bike (which disallows drafting, but is â€śfastâ€ť on itâ€™s own, along with the TT position nearly all the time) and you can see some big differences.

1 Like

But as TR users, we are all aware the miles is a invalide training metric, surely as long as the power figures and HR are the same as IRL, I just donâ€™t care

As a person living in the flat part of the UK, I donâ€™t take my outdoors â€śfree milesâ€ť to mean that I am fitter than somebody living in the Alps, or that including those miles invalidates my training

I wouldnâ€™t say that miles are an â€śinvalidâ€ť training metric, but they definitely rank (far) below hours and watts as a primary metric.

At the end of the day, IRL miles can be a reasonable substitute for hours.

But yeah, when I have done those sub-4 hour Zwift centuries, I certainly donâ€™t think I have gotten the training effect of riding an actual century. I have gotten the training effect of riding 3:45 at just under 3 w/kg (mid-Tempo pace for me).

I agree with your agreement re: the disagreement.

I think.

2 Likes

Yeah I agree with that, as long as you take climb, wind conditions and weather into account e.t.c, it can make a reasonable subsititute for hours, but I would make the argument that a far better substitue for hours, would be hours

1 Like

I knowâ€¦ what if we somehow combined this measure of distance and measure of time togetherâ€¦ into some form of distance over timeâ€¦ and then still realize it is not a full picture of the effort buried within the reality of the ride

3 Likes

With those data you might be able to calculate the underlying â€śZwiftâ€ť CdA and Crr. Youâ€™d need to know the exact distance, which you can get by looking at your data for both the uphill and downhill sectors.

Iâ€™ve done some estimates of Zwift CdA and Crr for a couple of bikes and wheels, though I havenâ€™t varied the total mass like you have. It works better if you test on a course when there are no other riders. There are a couple of ways to do that, but the simplest is to log in to Zwift, go to a particular course, and then disconnect from the network. Other riders will disappear and you can run your tests and collect data without worrying about drafting effects.

Thereâ€™s an important â€śnot real lifeâ€ť part of the Zwift model, which calculates a frontal area from weight and height. There is no â€śreal lifeâ€ť direct correlation between these parameters, although intuitively we can all agree that a 95kg person measuring 2,0m has a larger frontal area on a bike than a 55kg, 1,6m one. The Zwift model attempts to estimate these variations.

And taking that arguement one step further - two 95kg, 2.0m persons may have very different frontal areas dependent on the position they can hold on the bike.

If Iâ€™m honest, this all seems a little pointless.

1 Like

If you want to know the relationship between height, mass, and CdA you really have to measure it, but before that was easy (or, at least, easier) to do there were lots of attempts to estimate CdA from anthropometric parameters. The fit is great, and it depends very much on whether youâ€™re talking about road or TT positioning, but some of those old anthropometric formulas were discussed here.

Pointlessness is what we do best.

Sometimes, you want to understand how a virtual world like Zwift works and one way to do that is to vary weight and height to see what happens to the drag parameters. A side effect is that sometimes you can use those (semi-hidden, latent) rules to detect anomalous performances (i.e., suspicious performances, if you know what I mean and Iâ€™m sure that you do).

The OP actually had a different intent:

And this intent is not going to work, since it assumes a) there exists a reliable relationship between weight and frontal area in real life (hint: thereâ€™s no such thing), and b) Zwift accurately models this relationship (hint: you canâ€™t accurately model a relationship that doesnâ€™t exist).

Actually, we know pretty well how weight affects speed in real life: we have validated power-speed models that accurately describe this on real roads under real conditions. So the OPâ€™s intent can be reduced to knowing how weight affects speed in Zwift.

2 Likes