You’ve started zwifting… first of all, welcome! One of the first things you are going to notice as you speed around Watopia is that the name of the game is power. Everything in this virtual world runs off of those precious watts that you are able to generate as you turn the pedals. If you aren’t lucky enough to have your hands on a power meter, the kind engineers of Zwift have come up with a solution for all of us! That solution is zPower.
zPower is a calculated power value based on the speed of your bike and the resistance of the trainer you are riding on. Editor’s note: although the terms are used interchangeably by Zwifters, there is a slight difference between zPower and virtual power on Zwift. Jaden is technically using virtual power, but he uses the term zPower throughout this post. For more on this, read “Virtual Power Basics for Zwifters“.)
When I started on Zwift, I did all my training and racing using zPower, and one of the first questions that I asked myself was: “Is the power value anywhere near my true power?” The more I trained the more I wanted to be able to compare my numbers with my friends and see how I would stack up in group rides and races. I wanted to know whether the numbers I was getting for FTP, training zones, etc. were close enough to my actual power to be useful.
I am basing this article on a video I created, if you’re interested in getting the information in video format, check it out and don’t hesitate to subscribe to the channel:
Hope you enjoy the video, but in case you’re more interested in the text version of the story, here it is:
The Test
The first thing I needed to do was compare two different power data charts in the same time frame. To do that, I ran Zwift as I normally would, sending speed, cadence, and heart rate to the AppleTV to run my avatar. Separately, I used the new power meter on my bike (4iiii Precision) sending data directly to my Wahoo Bolt. After I finished the test, I was able to use ZwiftPower to compare two different power graphs based on data from Zwift (downloaded from Strava) and from my Wahoo Bolt head unit, downloaded from the Wahoo app.
When you add both of these data files into ZwiftPower’s analysis tool, you get some cool comparative information. The first thing you see is this graph (the purple line is zPower, and the light blue line is the 4iiii Power Meter):
We will get to the specific sections in the graph more closely soon, but before we get to that there are some other elements of the analysis that I wanted to highlight.
For an average power, I was very impressed with Zwift! A difference of only 6.1 watts (4.26%) averaged over the whole test was actually quite impressive, so kudos to the team for working hard and making something great. I won’t talk much about normalized power, but as you can tell in the graph it is also within about 4% as well, which I would consider very accurate for an estimation.
Maximum power is definitely a key point here, with an almost 10% discrepancy. The actual power meter read a significantly higher max power, which from what I can tell is due to the ability of a rider to push significant power into the pedals instantly, but the resulting wheel speed (what zPower is actually based on) does not increase quite as fast. If your sprint lasts for a long time, you may see the zPower number climb up to match that maximum, but if the power peaks and then drops, as mine does in this sprint test, your zPower numbers will not be able to catch up, and will not spike as high a power meter can.
To review the actual data, I divided the overall test into three segments that told three different stories about the difference between Zwift zPower and an actual power meter. Here are the three main test sections that I looked closely at:
Steady State
Most of the test was run as an (attempted) steady state ride, hovering around 200W. My intention here was to see what zPower could do if the speed was constant. As you can see on the graph, both of the values are tracking pretty closely together, with the zPower consistently lower, and the difference there was by about 10-25 watts. This may seem significant, especially on the higher end of that range, but the fact that it was tracking closely even though it was low helps me know that at least for workouts you will still feel the difference between hard and easy.
The other thing to notice on the steady state portion of the test was that my actual power data is quite jumpy. I’m fairly new to this power meter data, but from what I have learned from others with much more experience is that this is normal. When you first start measuring power, it’s reasonable that your natural pedal stroke will create slight variances in power even when riding at a steady state. The reason that the Zwift value is able to be so consistent, is because they are basing theirs off of wheel speed, which is more constant.
Intervals
I did some really quick interval work during the test, so I could see what the zPower readings would look like when you ramp up power, stay at that level for some period of time, and then bring it back down again. I think this result is interesting because it is the only meaningful time that the zPower value is higher than my actual power meter.
The difference in the graph partially represents a slight lag, so my power increases as soon as I put more pressure into the pedals, which makes sense, but my speed will only increase gradually, thus increasing the zPower. The interesting thing to note is that my continued high-power effort increased my speed fairly significantly, and even though I was holding a consistent power, I was still accelerating. That is why for a few seconds, my zPower will continue to rise past the top of my actual power numbers. The same small lag is noticed on the tail end of the interval, as I ramp the power back down again.
Sprinting
The most significant difference between the values that I saw was when I tried a sprint. I only ramped my power up to around 800 watts, but it was quick enough to see how zPower would react to the changes. As expected, since my speed increases slower than the effort I’m putting in, my real power increase is very sharp. The zPower is much more gradual.
Since I spent so little time at the power peak, it is reasonable that Zwift wouldn’t reach the same value, but the difference was about 70 watts, which represented an almost 10% difference. In a significant race-ending sprint, that wattage difference could be a big deal. Another interesting piece to note is that the decrease is much more gradual on the zPower side. When I stop putting in effort, my watts will instantly go to zero. However, and depending on your specific trainer, the flywheel you have will keep some of your momentum in the trainer and your wheel will continue to spin for some amount of time as your speed decreases.
Conclusion
After running this test, I was honestly very surprised at the accuracy that Zwift is able to achieve through a data-based estimation. It’s actually very impressive! I know the same accuracy results will likely not be the same for everyone (see the ‘Recommendations’ for tips on how to improve zPower accuracy) but from my testing I can say that I was impressed.
Knowing how power meters work, and how the calculations are made by zPower, can be a great tool to have when you are training on Zwift and are curious as to why you can’t put out huge wattage in a sprint as quickly as your fellow racers, or why you are able to keep a much smoother and consistent power curve than those with direct power meters are. I very much enjoyed diving into this data, and if you are interested in checking out my actual data set, I have made it public on ZwiftPower! Just keep in mind that I am still in training, so the power numbers are still those of an amateur cycling enthusiast.
Recommendations
There are likely a few people reading who are curious whether the results that I got are going to be similar to the results that you see. Since quite a few variables are involved, I’ll share three things that you can do to help your trainer give the most accurate zPower numbers possible (see Zwift Support for more information)
- Use a compatible trainer. The Zwift website has a list of the trainers (non-smart) that they have done extensive testing with. That data is the reason that Zwift is able to calculate an accurate power. I personally use the Giant Cyclotron Mag trainer. If you use a magnetic trainer (or any trainer with a variable resistance) Zwift will tell you which resistance value to use. Make sure to use it!
- Keep the trainer tight. When you tighten the trainer to the wheel, you want to be sure that it will not slip during the workout, because that will create inconsistent speed ratings. I do this by turning the wheel directly as a test to see if you can hear any slip. If turning the wheel manually causes that skid sound, it’s a good idea to give it another quarter-turn until it’s solid. If your wheel is slipping, Zwift will usually catch it and put a pop up in-game to alert you, and it could kill your power until you tighten it (it’s the worst if you’re in the middle of a race… so best to just make it’s tight from the beginning).
- Pump the tires to a normal PSI. This part isn’t tricky, but it is easily forgotten. Just as you would before an outdoor ride, check the PSI in your tires and make sure it is at a normal value, similar to what you would ride outside.
Ride on!
What About You?
Have you ever compared Zwift’s zPower estimates with a power meter? What did you conclude? Share below!
In before Z-power rangers start using this to justify their 12wkg FTPs
Exactly… I’ve seen those fliers as well, and the big justification for getting myself a power meter at all was to race without that worry for my own numbers! There are a lot of variables that need to go right for virtual power to be accurate.
So how does this explain the large number of zPower users who ride away from powerful groups? Is it all just about badly calibrated trainers or underinflated tyres?
The calibration definitely has something to do with it! Since a lot of your in-game speed is determined by your W/kg, the Watts your putting out only account for half of the picture though… if I can put out 800 Watts maximum, and my weight is 80kg, I’m doing 10W/kg. But if I (dishonestly) say that my weight is really 55kg, then magically my W/kg jumps up to 14.5kg! So ‘Weight Doping’ is probably more of an impact to the crazy speed of some users… Eric wrote a great post about that here: Weight Doping: An Introduction
My guess is it all comes down to setup. If you setup a smart trainer or dumb trainer incorrectly you’ll have these wild results. I saw a friend of mine who’s trainer broke so he used a set of rollers while keeping the trainer profile in game. Normally he is close 2.4w/kg but he was now cruising at 4.1 w/kg. People don’t understand the importance of setting up equipment correclty.
There are lots of variables that contribute to inaccurate zPower w/kg including weight, wheel size, trainer resistance. For example my son rides a 24″ wheel mountain bike on an old mag resistant trainer, I’ve done the best I can to set up the trainer but given his low weight, 24kg, inaccuracies are magnified and if he really goes for it he easily pushes 8-10w/kg, but at a normal steady state power for a 9 year boy he is at 2.0w/kg. I personally wouldn’t let him race with this set up even if I thought he was old enough because I… Read more »
Maybe this will stop some of the hate directed at those of us who use dumb trainers.
I can’t get my watts up as high in a sprint as I can on a real bike with a power meter, yet everyone assumes I’m cheating 🙄
I definitely feel the pain… especially when you’ve killed yourself for a high place in a race (even a C or D race) and see that your result was DQ’ed on ZwiftPower rankings for using virtual power! It’s a sad place to be! I hope Zwift can continue to find innovative ways to include that group yet discourage cheating, but it may be wishful thinking…
Cheating (whether thru weight doping or miscalibration) happens on smart trainers just as much.
Cheating is unavoidable when humans are involved, so I think it helps to remember we are always competing against ourselves at the end of the day.
This has happened to me several times. It is discouraging. Based on these test results I don’t see why zPower cannot be accepted in Zwift races as legit if you finish on the podium.
Problem is people will always find a way to cheat – it’s easy to say ignore it but when you see a lot of people in the top positions with unusually high (pro beating) power (or w/kg) and often very low HR (if at all). And have a look at: https://twitter.com/ulif/status/1270760278888779782?s=20
I agree. I honestly have a really hard time ‘ignoring it’ when I’m working hard for a good result in a race but then realize I never had a chance… but I’ll just keep training and improving myself!
Nice article and research, well presented.
One small thing to note is the assumption that the 4iii power meter is accurate, did you manage to compare its performance with say a smart trainer?
I’ve seen some pretty wild discrepancies with dingle sided power meters, due to the assumptions they make in translating the measurements say compared to PM pedals.
Wish I had a smart trainer!! Looks like the test will have to have a Version 2 based on all this great feedback! 🙂
i have 2 smart trainers. 1 is direct and 1 is wheel on; guess which one is not accurate!
True! I have a Tacx Neo 2. A friend of mine visited with his 4iiii Precision power meter, and it was constantly 30 of 40 watts higher than my neo 2 (riding at 200-250 watts ). After a calibration it was less, but still 10-20 watts higher. This makes it so hard to know which one to trust..
Any time you’re measuring power at the crank or pedals it should be higher than if you measure it at the cassette, since power is lost in the drivetrain. If they gave the same power that would be suspicious.
Presumably Zwift assumes power measured at the cassette, since this is what propels the bike, but if you want higher power especially at lower power levels where drivetrain losses are a higher fraction, pairing with a pedal or crank based power meter would make a much bigger difference than any in-game equipment upgrades.
one should only trust a dual reading device and even then it should be compared to say a neo; if not with 3% of each other there is a problem somewhere. racing on a neo is tough when everyone else gets free watts
Interesting article Jaden. I have very different results just based on FTP. Back in 2017 when I had a dumb trainer and first starting Zwift, my FTP was 330W. I was stunned that I had such great power right at the beginning, especially since I just started working out again. Later that year, I switched to a Kickr SNAP. My FTP was now 305. Still kicking butt and taking names. My rude awakening happened in 2018 when I got my 2017 Kickr. Despite working out all that time and dropping 20 lbs my FTP was 250W. It was night and… Read more »
Very interesting… I wonder if Zwift has made updates to their virtual power calculations too? Regardless it is a very hard balance to ‘include’ the racers without smart trainers but not elevate them above people with more accurate measurements… quite the challenge Zwift has to face
Three years ago I asked Zwift to cut back the power on Zpower. I went from 300 to 265 overnight when I bought a power meter. My wheel has 16 aero spokes instead of 32 round spokes so perhaps that is to account for a larger difference. But my trainer was the most accurate non-smart trainer Kurt Kinetic. I tested a different brand of speed sensor and got more accurate readings than my original brand name sensor. But, the scale should be re-evaluated. Non-included trainers should not read 400 watts for an hour, limit it to maybe 250. Thanks for… Read more »
Jaden good article, as we both know zPower is all about set-up and calibration. For the majority of new or junior riders who haven’t invested in a power meter or accurate smart trainer they just chuck their bike on and start riding. The majority believe they can produce the numbers they see, until something grounds them in reality. In the meantime everyone ride on and enjoy the game.
zPower is the reason Zwift racing has become a joke in recent months. And so long as Zwift keeps making money on them, they will do nothing about the havoc many create in a given race. I mean honestly, in one race today … 7.4w/kg for nearly an hour. Totally ridiculous! So many events have been ruined by these riders.
Derek you make a super point, zPower racers on dumb trainers produce crazy power numbers for sure. Last time I raced several B riders in the front of the pack were holding 6 w/kg flat out gapping the main pack. Anyway that said I laugh it off, there’s Zwift race results and then there’s the REAL race results on Zwift Power! I’ve finished some races in 20th place but get 9th in Zwift Power.com results so that’s where I go to see who wins, and results.
nearly all the fliers are ZP’s holding ridiculous numbers…. the article is a complete laugh trying to justify the bullsh!t
zPower has no place in racing events. At bare minimum, anyone wanting to race should have a direct-drive trainer for accurate power measuring. There are too many people that take this Zwift racing stuff very seriously, and it’s not fair to them to have riders with “estimated” power and improperly calibrated trainers in the mix. What need to happen is for Zwift to allow event organizers to NOT allow zpower in their races. But I don’t know that there is anything that can really be done about the inconstancy of wheel-on trainers.
Eh, I think it’s fine in the lower classes — because in those, if you’re averaging way above the limit, you’ll get pushed up. Does it really matter if a rider who’s actual power average is 2.3 w/kg is a little miscalibrated and comes out at 2.7 w/kg and ends up at the back of the pack in the Cs rather than the front of the Ds? There’s tons of riders in both ranges, and neither is a unrealistic range for a mediocre amateur. As long as it’s a steady over (or under — I’ve seen a few people mention… Read more »
Jaden nice article did you use a left only 4iii power meter?
If so that only doubles the left side. If your left/right balance is 48/52 say it would match the steady state power closer.
Yes! I used the left-side Precision meter, and I’m sure there is a balance impact to that… in general, what is the range of left-right balance that people typically see?
Even for an individual it’s very hard to state one figure as imbalance can vary somewhat depending on power output and fatigue. Most people seem to be between 48-52 one way or another in (NOTE) very general terms with a few I know going to 47-53 but that’s few and far between. I don’t know anyone beyond that BUT it’s impossible to tell and I’m sure there are outliers or those riding with specific injuries or imbalances. If you can ride a Watt Bike in a gym (preferably on a free try before you join pass…and when they all… Read more »
Just to add my own experience. Zwift got me hooked on riding with power. I had some months using virtual power then gladly invested in a 4iiii power meter that I have subsequently verified the accuracy of against multiple other power sources. My own experience and comparison was that virtual power was giving me a not insignificant 10% boost at race powers. Also the higher my power got the bigger the discrepancy, as my actual power curve deviated away from the virtual one. So for instance if I was sitting at 200W on the 4iiii I might be showing 210W… Read more »
Just to add, yes I selected exactly the correct supported dumb trainer and yes my tyre pressure was correct and regularly checked. In fact the picture was even more complicated because the power exaggeration increased as the fluid trainer itself warmed up too! Not a factor after a 20 minute warm up but nonetheless it shows how hard it is to cater for all these estimated variables!
Agreed! So many variables to account for, and based on quite a few comments I’m thinking I may have gotten lucky on how accurate it was for me, but very interesting data point!
Just to add further intrigue Zwift’s power curve for my particular supported trainer (and triple checked to ensure the correct one selected in Zwift etc) is obviously a little different to the published one I have dug out. The one via the power of the internet here http://www.powercurvesensor.com/cycling-trainer-power-curves/ is actually much closer to my power numbers for the relevant wheel speeds. Now it is a weight on trainer so perhaps Zwift assumes I’m even heavier than I am?! Or perhaps it’s algorithm is just wrong and inflated in my particular case…weird.
Just to add I write just to add far to much…
too*…
Zwift 3sec avg in the settings don’t work too well, that’s why it’s difficult to drop any one on ur wheel for 3 Sec delay 🤔
I realize it’s not in Zwift’s interests to point out how many ways there are to “cheat”. But it might help people realize that it’s not possible to have it be “fair”. You can lie about your weight, even by a lot You can miscalibrate your smart trainer, in one of several ways You can even use an app that takes your power reading and multiples it by whatever number you select. And guess what, no one will know, because there’s no difference in icons, like with the lightning bolt. Those who get worked up about z power need to… Read more »
Absolutely there are a million ways to cheat if you so desire. The difference with virtual power is that you aren’t necessarily trying to cheat, you are just using a system that can (I use the word can because I have read the article but see my own personal experience above in the comments where it gave me on average 10%) often dramatically exaggerate your power output.
Not to say genuine miscalibration mistakes can’t happen too on smart trainers and power meters but at that stage you often have more data to compare with and are more likely to be making an active choice to do it and cheat.
What would be a more accurate setup for determining power. A direct drive trainer or a power meter with a dumb trainer?
Depends on the power meter or the trainer! The power meter has the advantage of being able to be used outside too. If you want a well established, accurate smart trainer instead you probably won’t go wrong with any Kickr direct drive that’s recent (Core or up) or the Tacx Neo.
ZP is far from close. end
So you didn’t read the article then.
i read the article and agree with dan; comparing two inaccurate devices does not add any value to anything. if the op had a neo and dual reading device and compared those to each other and the dumb trainer then the article would be valid.
I’ve done one race this week and the winner was a zPower superhero who averages 500 watts in races (429 in this case, must have been holding back in this case so as not to stand out). He obviously either doesn’t care or thinks he is a superhero haha. Anyway maybe after a few complaints he has eventually disappeared from the results. The guy in second (now the winner) is a weight doper who in the couple of months he’s been on Zwift has gone from 79 to 58 and now settled on 70 kg as his preferred race weight!… Read more »
Definitely hard for Zwift to be both inclusive to non power-meter owners, while still being fair to those that are honestly trying to train and race with their true power! I hate watching 600W average riders drop the entire field… but I’m sure it’s something Zwift is very keen on improving!
Mr Elite Turbo Muin zPower consistently read +15% over a proper power meter. Flattering but untrue !
Only using a left sieed power meter? Still not accurate then, sorry to say.
Sided*
Agree with those taking the position that ‘cheaters’ and ‘zpower users’ are not terms for the same thing. I’ve used a kurt kinetic (dumb), a kickr snap, and now a kickr core. What happened to my FTP between each? Nothing, really. Setup your trainer consistently, check your tire pressure once in awhile, zpower will do an OK job with most setups. Do some structured workouts. Race and have fun. If you should be a C and incorrect power makes you a B, or the reverse, who cares? But if you are destroying everyone and wrecking races, stop—you aren’t that strong,… Read more »
I ride a Tacx dumb trainer and am under no illusions that if I’m hitting 5wkg up Alps du Zwift it doesn’t represent what I could actually achieve on the road. It’s set up in line with instructions on Zwift and Tacx and I’m honest with my weight, so much my wife ordered me a kickr core from Wahoo last week hopefully here for my birthday next week so I could get some proper output.
Thanks all
Update on new trainer is I am averaging 2.5wk g on fairly undulating courses, so I’d take that rather than the dumb trainer output
I wonder how much a difference pressure of a roller on a tire (which should substantially increase resistance) or tire pressure (also increasing resistance) should make. For example, I have a Rock and Road trainer I used to use, and it recommends 2 turns for normal use with I think 85 psi, but if I was using it for racing and I was trying to game the system, I might be tempted to get away with fewer turns and/or lower pressure.
Great test. It’s a shame though that there is also a lot of zpower riders who believe they are doing everything right (hard to know if they are), yet you don’t need to analyse their data to know the results are VERY unreal. Some of them (Achim Mueller who holds a lot of ridiculous zwift KOMs) will even call people calling them out ‘jealous’. It is simply undeniable that tons of zpower riders are impacting the racing tactics with superhuman power figures. I dont mind finishing behind them, but I do mind them impacting the tactics of the race. Zwift… Read more »
I would love to read a zwiftinsider.com interview with Mr. Mueller!
isnt your power meter (4iiii) a single sided unit; if your going to do something like this at least use a true dual reading device; your watts could be out by 30 watts at least using one of these.
apologies read the above and the obvious about single sided PM’s has been pointed out
really good article, thank you. I also have the Giant Cyclotron and always wondered if what I was seeing had anything to do with reality.
Anyone knows if source of power is visible on ZP, strava, etc.? Not that it would help to eliminate the problem with miscalibration or weight doping, it would be still a good indicator when comparing efforts