It’s known that being lighter at the same power output lets you go faster both in real life and in zwift. However I’ve noticed that on Zwift being lighter is disproportionality favorable than real-world physics. This was a hunch so I wanted to put some numbers to it. I wish I had the time to do deeper diving into this, but here’s what I’ve found.
Below is a table showing speeds both in real life and Zwift for a rider at various weights doing 4 w/kg on a 0% gradient. I equivalized the speeds for the 50kg rider as a baseline metric. In the graphic you can see the variables I used.
As you can see, as weight increases, the difference between estimated speeds on Zwift versus “real world” diverges.
Notes:
The point is not that these estimates are accurate by their lonesome, rather the impact weight has on speed is disproportionally higher on Zwift. I attempted to change other variables such as air density, rolling resistance, etc to see if any of those changes resulted in a pattern closer to the Zwift speed curve, but I couldn’t find any realistic adjustment to replicate it. This only confirms my initial observation.
It is also important to highlight that in both of these estimates, CDA or other variables are not adjusted. Meaning that the observed differences is purely stemming from weight. Think a more dense human rather than more weight around the belly. Obviously 50kg riders are typically shorter and able to get lower CDAs than larger 80+kg riders. However for the purpose of this exercise we are only adjusting weight in both situations.
I understand that the zwift estimates are not based on in-game zwift tests, but ZwifterBikes seems to be a reliable estimator tool. I would love to have the time to perform these tests myself in-game, but for now this is what I can do.
Discussion
What do you think of this? If the data above is true, is this something the game devs may have implemented on purpose? or a simple oversight? Should this get attention from the game devs?
I’ve noticed this as well. Maybe not to that extreme but I’ve noticed some performances where if the iTT was performed outside there’s little chance the 200w ~50kg rider would even come close to the 370w ~90kg rider on the same equipment, even despite height differences. However the numbers above isolate only weight as the variable of change here. So while height may also be another variable to look at, the above differences should thwart that interference
The base analysis is correct ish but limited in it’s conclusion. Recently you see most of the heavier riders being put in lower categories, which I guess balances it out for racing a bit.
ZRS does not however address height. The below riders are basically unbeatable in their cats.
How did you avoid adjusting CdA for the Zwift figures? Weight (and height) is part of the CdA calculation in Zwift so changing the weight will change the CdA. I assume zwifterbikes accounts for this but I am not familiar with the website.
Yes, the TT results are frankly ridiculous. See here:
According to zwifterbikes, for Tempus Fugit on a TT bike, those take about the same time (27:05):
190 cm, 90 kg, 360 W (4.00 W/kg)
160 cm, 55 kg, 256 W (4.65 W/kg)
Maybe the idea is to compensate for the fact that heavier riders would have a harder time re-accelerating after a turn? Wrong plus wrong equals “right”, I guess…
That is not what he is saying, its a simple comparison to real life
It did favour heavier riders but not in the software, it favoured them because of the watts per KG limits to each pen which meant the heavier riders just under the limit for each pen had far more power and a big advantage over lighter riders just under the limit resulting in lighter riders going over the pen limit in W/KG just trying to keep up with the heavier riders.
I wonder if the above stats if correct was a way to somewhat compensate for this!?
Just note that ZwifterBikes is using the Gribble formula under the hood, so using ZwifterBikes as proof of Zwift with Gribble as proof of Real Life might not be a valid test.
Also note that Gribble/ZwifterBikes only calculates steady state, and not dynamics (neither Momentum or Draft) which both will have huge effect on actual performance in Zwift.
And apology. Now its the evening and I’m on glorious acres of multiple monitors, I can see the extra details that my brain failed to spot when previously reading the thread on a small mobile phone - chiefly a variety of weights. Sorry.
That said, I will still stand by the “IRL comparison” being 100% not IRL at all - its just a different theoretical cyclist algorithm from Zwift’s. And doing so with a single assumed Cda for riders of different weight.
How did you avoid adjusting CdA for the Zwift figures? Weight (and height) is part of the CdA calculation in Zwift so changing the weight will change the CdA. I assume zwifterbikes accounts for this but I am not familiar with the website.
Adjusting weight in Zwift doesn’t impact CDA to my knowledge. CDA is calculated using zwift’s baseline player CDA, equipment chosen, and player height. The estimator tool I used has options to adjust variables and the only one I touched was weight, keeping height at 175cm and riders on the tron bike.
If adjusting weight in Zwift does in fact impact CDA, that’s another conversation to have in my opinion. But I don’t believe this is the case.
That said, I will still stand by the “IRL comparison” being 100% not IRL at all - its just a different theoretical cyclist algorithm from Zwift’s. And doing so with a single assumed Cda for riders of different weight.
I agree this isn’t an IRL comparison and its not a perfect analysis for sure. However like I said in my post I did experiment with other rolling resistances, CDAs, etc and could not find any pattern which matched the Zwift’s speed curve based on weight adjustment. If there is one that I could plug into the Gribble calculator, I either am missing something or it would be unrealistic inputs.
My post just includes one of those scenarios where I assumed a .272 cda, .003 rr etc
While other inputs led to different results of speed, they were closer to the results from above rather than Zwift’s speeds. Especially when looking at the 85kg rider if matching the 50kg rider speed.
Just note that ZwifterBikes is using the Gribble formula under the hood, so using ZwifterBikes as proof of Zwift with Gribble as proof of Real Life might not be a valid test.
That’s a very important point to consider. I hope that someone with the time and know-how may take this idea and perform in-game tests and a deeper analysis of how weight impacts player speeds and how that compares to real-world physics.
Heavier riders at the same w/kg of course have a benefit, but what I’m suggesting is that benefit is reduced on Zwift compared to real life.
But how are you factoring in the pace dynamics 5 adjustments?
You will remember light weight riders had a very difficult time with those and being pushed out of groups and then being blocked from moving forward or having to make a big effort to do so.
That’s the main reason why lighter riders go faster than heavy riders on flat terrain if they are doing the same watts.
" While some Zwifters like to say “it’s all about pure watts on flat roads,” it’s not that simple. Weight does affect your speed on the flats in Zwift, since it changes your computed CdA. Or to put it another way: two riders of the same height but different weights will have different CdA values in Zwift, meaning the heavier rider will need to push more watts to keep up with the lighter rider, who is encountering less (virtual) air resistance."
“Zwift calculates your avatar’s frontal area based on your weight and height (your BMI, essentially). That frontal area, in turn, is the key factor in Zwift’s CdA calculation, which accounts for the majority of the virtual air resistance your avatar encounters.”
But how are you factoring in the pace dynamics 5 adjustments?
You will remember light weight riders had a very difficult time with those and being pushed out of groups and then being blocked from moving forward or having to make a big effort to do so.
Can you put that in your modelling?
Chris, this is about the speed calculation of Zwift for a rider, nothing to do with multiple rider dynamics regarding draft calculation.
Very interesting. If this is true, which coming from a Zwift employee sounds like it is, then I have many thoughts about this. I find a penalty for higher weight into CDA, while in concept makes sense, shouldn’t be applied evenly. But I suppose that’s for another thread.
I see now, weight does in fact play a role into CDA for Zwift. Someone else also linked a useful thread.
My main point still stands though, I believe that this weight CDA penalty is contributing to the “slow down” too strongly. For example while in the last 3 years I haven’t grown any (im 25) I have gained nearly 10kg of muscle. To real life my overall stature is hardly different aside from beefier thighs. Moreover, this muscle lets me hold a more aggressive position for longer outside. I know this isn’t always the case and sometimes 10kg of weight does impact IRL CDA poorly, but that’s the thing I suppose- I don’t think weight should impact CDA on zwift flat out. That said, currently in Zwift I receive a CDA penalty for this which I believe is too strong. I would need to re-do my analysis with this CDA penalty in mind, but my hunch is that this penalty is too strong.
I can use myself as anecdotal evidence I suppose until I can get around to doing an analysis.
When I started riding I was around 69kg and now I’m floating 79 and even 80 some days. It’s been a super gradual increase of just a couple hundred grams over weeks/months at a time. However I’ve noticed that while my total watts have increased, my riding speed hasn’t- at least in Zwift. My outdoor speed though did increase- a lot. Even on the same bike/equipment/power meter over time I’ve noticed an increase in 4+ kph at my zone 2 pace (which in terms of w/kg hasnt changed much, but raw watts has). However on Zwift I’ve hardly seen any improvement in my zone 2 pace overall.
So I guess it comes down to the fact that Zwift penalized me for my weight gain in terms of CDA while in real life I haven’t seen that impact.
