Accuracy of Speed & Distance


(Daniel Smith (AZ)) #1

Rode my first ZWIFT session yesterday and I noticed that the speed and distance displayed was different that what my Garmin Edge displayed. For example my average speed displayed on Garmin Edge was 21.1 mph but ZWIFT only displayed 17mph. As for distance my Garmin registered 21 miles while ZWIFT only logged 16 miles. Why the difference, if the data being displayed in ZWIFT is obtained from my Power, Speed, Cadence & HR sensors you would think the data displayed would be accurate


(simon orange) #2

if you have a power meter then zwift will use that for speed. The speed sensor will be ignored

obviously the power required uphill for a set speed will be higher than when going downhill.

the key point of zwift is that it translates your power into road speed. The only time that ant+ speed sensor data would be used is if you used virtual-power. But then the ant+ speed is just converted to power and then back to zwift road speed.


(anon18154799) #3

What kind of trainer are you using? If you are using a power meter on a regular trainer, then the difference can be accounted for by the power-to-weight algorithm based on the terrain. So, your wheel may have been spinning at a certain velocity on the trainer, but the lack of gradient effect isn’t replicated by your Garmin. If you are on a smart trainer, then I have no explanation.


(Eric C. (Zwift HQ)) #4

Speed sensor speed has no bearing on speed in Zwift unless it’s used primarily to help calculate zPower.

Speed sensors detect how often your wheel rotates (generally) - this is fine for the real world but doesn’t mean much with a virtual rider on a virtual bike riding on a virtual island with virtual hills :slight_smile:


(Carol DePietro SZR) #5

These are all good answers but did not really answer…is the speed and distance accurate, if not whats the point? Im just riding to see a pretend me on a tv screen?


(Paul Allen) #6

Carol, 

Distance is irrelevant on Zwift to be honest, it is more about watts and time. 

Yes you can see your avatar moving on screen and you get a virtual distance, but you are still stationary on your bike.

The more watts the faster you will go within Zwift.

The distance and speed that is calculated within Zwift are accurate, just remember that there is no wind, you don’t have to slow down for stop signs or corners and there are no pesky cars out to get you.

You can also read through some of these threads: https://support.zwift.com/hc/en-us/search?utf8=%E2%9C%93&query=Speed+and+distance&commit=Search

 


(Brett Deriso) #7

I have to agree with Daniel and Carol here… Maybe I’m just not getting my head around the core concept.

I use Garmin Connect and a Fenix 5X watch paired with ANT+ speed and cadence sensors to track my activities in meatspace.  When I log in to Garmin Connect, I expect that the distance, average speed, and moving time values are accurate, literal representations of physical distance, speed and time recorded (insofar as I have entered the correct wheel circumference).  Enter Zwift, where I routinely log 20% more virtual distance than the Fenix is measuring off the literal revolutions of my rear wheel at the same time.  This (inflated) Zwift data then gets pushed to Garmin Connect, where it contributes to (and erroneously inflates) my distance and training effect goals.  Not cool.

I get that Zwift is making certain assumptions based on stated weight, gradient, and pedal cadence to derive power, but regardless of the extent to which these extrapolations approximate reality, the rear wheel is still moving a specific number of revolutions on my trainer, which translates into actual physical distance, and Zwift completely ignores that.  If we take the virtual London route as an example, there is no excuse for substituting approximations when the map and topographical data are known quantities.

Whether Zwift derives speed from known distance over time, or distance from approximated speed (via power-to-weight and pedal cadence assumptions), I have no reason to believe either is accurate enough to count toward any of my training metrics.  This is also revealed in virtual world shenanigans like “coasting” down a Zwift descent long after I’ve stopped pedaling and the rear wheel is inert.  So, to Carol’s point, yes, it is basically just a pretty picture to stare at while you bang out calories without actually moving.  I gather that if I want to track the actual, physical distance I’ve moved my rear wheel (and by extension would have moved it in meatspace), I should record Zwift sessions with my wearable and ANT+ sensors, and disregard anything Zwift uploads to my tracking platforms.

As advanced as the tech is, it’s still a far cry from approximating reality.


(Paul Allen) #8

Brett, 

Your request is not going to happen. If you are using a speed sensor connected to Zwift here is how speed is calculated. Zwift takes your rear wheel speed and the power curve of the trainer you selected and converts that to watts, with those estimated watts, your weight entered, in-game bike used (yes, this can make a difference), in-game drafting and virtual elevation changes. The speeds and distance within Zwift are fairly accurate.

It is much closer to reality then what you are asking for. If what you are asking ever got implemented your in-game speed would be extremely unrealistic. For example, you would be flying up 15% grades at speed around 20mph! Does that sound realistic for a cyclist? 

I will again state that speed and distance within Zwift (any cycling for that matter) is useless data, and the only thing that you should be concerned with is watts over time.

Here is a good read for you: https://zwiftinsider.com/zwift-speeds/


(Brett Deriso) #9

Thanks for the reply Allen, but, respectfully, you merely confirmed what I already surmised, and that article does not let Zwift off the hook at all.  They begin by dismissing outright that their algorithm might be based on flawed assumptions, and you stated yourself that they begin with data from a connected speed sensor.  There is no way they’re interpreting the data from a connected speed sensor correctly if it produces a vastly different measurement in the game than it does in meatspace.

To wit, it should make no difference whether it’s 10 km of asphalt or 10 km of my trainer’s steel roller passing under my tire, but Zwift nevertheless introduces a difference artificially.

While route conditions (ie. Gradient, wind, and rolling resistance) do impact power and Zwift does a fine job of approximating these by altering the resistance in a smart trainer, they have no bearing at all on the distance traversed by an actual wheel spinning against an actual roller, unless the laws of physics have changed, and my tire somehow increased in physical circumference simply by attaching it to a Zwift-connected trainer.  I reiterate: a wheel with a known circumference, traveling a fixed number of revolutions in meatspace traverses an easily calculated distance (simple multiplication), assuming no slippage.  Divide it by the time spent moving, and you have speed.  If you fed the same circumference and revolutions and time to an algorithm and came up 20% faster, or further, your algorithm is NOT an accurate simulation of reality, insofar as distance and speed are concerned.

As for the 15% grade at 20mph, I don’t know where you get that. If anything it’s an admission that Zwift does not accurately translate gradient to resistance, and I’m sure that’s not what you meant.  When you are presented an uphill gradient in the game, the software responds by increasing your trainer’s resistance.  You then have to gear-down to maintain cadence at a lower wattage, or stay in a harder gear, lower your cadence, and exert more power.  Either way, your wheel is going to slow down on your trainer, and if Zwift WAS actually reading your speed sensor, the speed on screen would reflect that deceleration.  How do we know they’re NOT reading the speed sensor?  Because my avatar continues to fly down Zwift’s virtual descents long after my actual wheel is stationary.  And yes, I get that they have to do that, because trainer spin-down (whose tiny flywheel lacks the inertia of a 70 kg rider in meatspace) is vastly truncated.

You are correct in your conclusion that Zwift is (at best) a reasonable simulation of Watts over Time, and that may be how most people track iterative stationary biking activity now; but there are still some among us who make distance-based goals as well, and know that stationary biking can count toward those goals if the readings from a speed sensor are treated literally.  What can I say, “10,000 km per year” rolls off the tongue a lot better for some than “10,000 minutes of 2.5 W/kg per year.”

FYI, CycleOps’ own virtual training platform, Rouvy, DOES ingest and translate speed/cadence sensor data without embellishment.  You ride virtual courses (recorded or manufactured) of known distance, and the application uses the revolutions of your rear wheel to accurately reflect your speed and progress along the virtual route.  They also use the elevation data gathered from mapping to approximate resistance, same as Zwift.  Taking the literal speed of your rear wheel at face value does not in any way break or overly complicate the simulation.  It’s not rocket science.


(Paul Allen) #10

Brett,

Seems you are still not understanding the way that speed and distance works within Zwift. I will post this again on how it is calculated using a speed sensor (zPower/VirtualPower): 

Zwift takes your rear wheel speed and the power curve of the trainer you selected and converts that to watts, with those estimated watts, your weight entered, in-game bike used (yes, this can make a difference), in-game drafting and virtual elevation changes.

You think that if your rear wheel travels 10 miles on your trainer that it should also travel 10 miles within Zwift and that is just not the case. Lets say you are going up to the Radio Tower on Watopia and stopping once you hit the top. Using what you are asking for you should be traveling up the mountain at the same speed at which your rear wheel is spinning (that is the only way the distance would match), but that would make your speed during ascent of the mountain extremely unrealistic. 

Some people can wrap their head around the concept and others just can’t. Also remember that you are not moving at all in the real world, so those miles that are so precious to you don’t really exist. 

What you are asking for is to flatten out Zwift and to remove the virtual elevation changes and that would take away from the realism. Also, if you where just to apply your suggestion to just the users with speed sensors than they would be flying by the ones with power meters/smart trainers within Zwift.

Just ignore the speed/distance of your rear wheel and work on your watts and you will see an increase speed within Zwift and will have a much more enjoyable time. 

Speed and distance within Zwift are fairly accurate.


(Brett Deriso) #11

No, Paul, that is NOT what I’m asking for.  You don’t have to “flatten” your simulation to interpret it’s real-world mechanical inputs accurately.  I don’t know about your setup, but when Zwift increases the resistance on my CycleOps trainer to simulate an incline, my rear wheel slows dramatically.  Whether I’m grinding up a mountain in my lowest gear or flying along a straightaway in my highest, I’m producing a literal, physical action in mechanical space -an action that is very easily measured.  I don’t need to layer twenty different assumptions on top of the measurable physical motion of my tire to see that I’m decelerating.  And I guarantee you, that wheel will traverse that 10 km “incline” a lot slower than a 10 km straightaway.  There’s nothing unrealistic about it.  The fact that you guys can’t get your game to reflect that speed one for one tells me YOU can’t get YOUR head around it.  It’s also why your comment boards are littered with users who don’t believe the numbers they’re seeing.  It because the numbers are fabricated.  Badly.  These people aren’t fools -some are professional cyclists who know when the feedback they’re getting is out of whack.

After re-reading this thread, and combing through the paired device settings in the game, I fully understand that Zwift doesn’t use rear wheel circumference for anything, and is backing into a prediction about speed based on assumption s about power.  But no one who cares about accuracy should simply take your word for it and accept assumptions and convoluted math when concrete measurements are readily available.  Your method requires (at a minimum) two things:

  1. An accurate list of power curves for any connected trainer, and
  2. A trainer that conforms precisely to those power curves (i.e. Isn’t malfunctioning or defective)

What you’re doing is the equivalent of telling everyone their pot of water should boil at precisely 212 degrees Fahrenheit, because that’s what your lookup table tells you, while ignoring the differences in salinity or the barometric pressure under which each pot is being boiled.  Meanwhile, each user has their own thermometer and can see that your prediction is either spot on, or manifestly incorrect.

As for whether anyone should care about distance metrics on a stationary bike, I don’t think it’s your place to dictate to anyone how they should quantify the work they perform.  I could use the same dismissive attitude to point out that the computer-generated visual experience you care so much about is likewise irrelevant to the task of cranking out watts.

I suppose we’re just talking past each other at this point, but for what it’s worth, you have answered my question.  As you have repeatedly suggested in response to this kind of inquiry, Zwift is not a good fit for users who care about tracking distance or speed averages, and that’s a pity, because it could easily do both without sacrificing any of its power-tracking capabilities.

I say again, Rouvy figured out how to do it, and at half the monthly cost, I might add.  I know who’s getting MY subscription dollars.


(Paul Allen) #12

Just to let you know, I don’t work for Zwift.

Again, on an indoor trainer it’s about watts and w/kg (this goes for Rouvy also), judging your fitness in away using speed and distance in a virtual world is pointless.

I assumed you were using zPower, but reading your last post suggests you are using a smart trainer. You have to remember there is no wind, no cars, and you don’t have to slow down for turns on Zwift.

If Zwift matched the distance that your rear wheel traveled does that mean you did anymore or less work? Nope, it’s just an arbitrary number.

I guess all the pros on Zwift are wasting their time.

Have fun on Rouvy, I hope it works out for you and Ride On!