It all started with a simple tweet from SRAM:
RED eTap AXS drivetrains look good in person, but how about in a virtual world? @GoZwift now offers select frames featuring the groupset you know and love. Who’s on their way to upgrade their avatar’s ride? 🙋♀️ #eTapAXS pic.twitter.com/pQfJtf85VB
— SRAM Road (@SRAMroad) December 17, 2021
SRAM RED eTap AXS in Zwift, you say? This was the first I’d heard of it. Cool – a shiny new groupset in Zwift!
But then I remembered what happened the last time a new groupset was introduced to Zwift. That would be Shimano’s Dura-Ace 9200, which showed up in game in September’s update, slowing every frame that had it installed.
(Putting the new Dura-Ace on Zwift framesets, it seemed, added a bit of extra weight to your setup, which slowed frames by 1.5 seconds over our ~50 minute flat test, and 19 seconds on our Alpe du Zwift climb test.)
Would the new SRAM RED eTap AXS have the same problem?
I scrolled through all the frames on Zwift, and only the Cervelo S5 2020 was using the new groupset. I sure hope they didn’t nerf the Cervelo… that’s a favorite of so many racers...
So out to Tempus Fugit and up the Alpe went the Cervelo S5 2020 frame with its shiny new groupset. And the numbers came back. The Cervelo S5 2020 had indeed been slowed – by 2.5 seconds on our flat test, and 21 seconds up the Alpe. Ouch!
You read it right, racers. The fastest frame has fallen.
Watopian Irony
I reached out to Zwift, and they said the groupset was upgraded at the specific request of Cervelo.
Do you suppose Cervelo realized that the “upgrade’s” unintended consequence would be hundreds (thousands?) of racers placing the S5 2020 back in their virtual garages? Doubtful.
Groupsets in Zwift
One of the many fun ideas rattling around in the heads of Zwift’s game developers is this: give Zwifters the ability to customize more of their bikes in-game. Right now all you can do is choose your frame (and sometimes its color) and wheels. But what if you could change your tires, groupset, pedals(!), brakes, seat, handlebars, even your bar tape?
Some of it would just be cosmetic, while other parts might actually affect your speed, like the oversized CeramicSpeed pulley wheel featured in a mission some time ago.
With that in mind, Zwift has (I’ve been told) coded the game so your bike’s total weight and CdA is based on the sum of its parts. We can’t customize those parts yet, but if the customizer is ever built, the physics are in place.
Without a customizer in the game, frames come with their associated parts already attached. So each frame has a groupset associated with it. Here are the groupsets currently in game:
- Campagnolo SuperRecord
- Shimano Dura-Ace
- Shimano Dura-Ace 9200
- Specialized S-Works
- SRAM Red
- SRAM Red eTap AXS
- SRAM XX1 (MTB)
All this is good news. Further bike customization is a fun idea that many Zwifters would enjoy. Zwift hasn’t built it out completely yet, but some essential parts are there, with the rest to be built sometime zoon.
A Bug Revealed
Until Dura-Ace 9200 arrived in September, all the frames that had been released on Zwift had kept their original release groupsets (as far as I know). That is: groupsets were never upgraded.
But when Dura-Ace 9200 came out, Zwift put that new groupset on a few frames which had previously been released with different groupsets. We had tested them with the old groupsets. Then we tested them with the new. And this revealed a bug or misconfiguration on Zwift’s side, because the “upgraded” frames were now slower. Heavier.
What’s actually going on here? Clearly Zwift didn’t set up Shimano and SRAM’s newest flagship groupsets to be heavier than the old ones. The only logical explanation is that, for some frames in game, some portion of the groupset weight isn’t being taken into account.
So it’s not that Dura-Ace 9200 (and now, SRAM Red eTap AXS) are overweight in Zwift. It’s that some older groupsets are underweight, or their weights aren’t being calculated correctly.
What’s the Fix?
If I’m right about the bug above, a fix would involve, at a minimum:
- Assigning all seven groupsets in game a proper weight and CdA value
- Assigning all frames a specific groupset
Once Zwift made this change, all hell would break loose here at the Zwift Insider lab. Because we would need to re-test each and every frame to determine its performance, since we wouldn’t know if that frame’s old data was valid anymore.
It’s quite possible that we’ll see a major shakeup in frame rankings once this is fixed, because there could be many frames out there currently using “weightless” groupsets. Only time (and fixes)(and testing) will tell. Until then, all we can do as racers is ride whatever performs best in game, based on tests.
What Now?
After the Dura-Ace bug was revealed back in September, I had hoped Zwift would fix things quickly. But here we are three months later, with the much-loved Cervelo S5 2020 being nerfed thanks to a groupset upgrade.
So today I updated the Frame Performance Charts and Tron vs Top Performers chart. I’ll be updating other pages/lists on the site soon, but for now I would recommend using those two pages if you’re looking for current frame performance data.
Questions or Comments?
Post below!
Comments? “LOL”
The S5 was getting boring, but it’ll be even worse if everyone is back to riding on Tron bikes
Damn right John ! … I really hope that THIS IS NOT THE FUTURE, that we ONLY see a “GAME OF TRONS” … let’s pray for different brands which can continue to compete on a similar level with the Tron … to be able to chose different frames & wheels (and outfits !) is INDIVIDUAL ENOUGH in my opinion … in a fast race, I wouldn’t be able to see any other groupset anyway … please, let it be ZHQ … use your time more sensibly
Maybe the tron could be disabled by year’s end, like, turned to grey without the glowing meaning it becomes quite slow, and you need to achieve current year Tron by doing another big quest.
I hope not, been working a year now to get that Tron bike. Must admit though, the novelty of seeing so many sort of kills the appeal to bother getting it so much
I got my Tron and don’t use it.
I think they are deliberately increasing the number of tron bikes in the game so when they release a physical training bike that looks like the tron bike, people are attracted to it
Best conspiracy theory yet!
Talking about the ceramic speed pulley, it made me realize Zwift is not doing monthly missions as they were doing in the past.
Any ideas Eric why Zwift is doing less monthly missions?
In a dense Zwift press release from about a week ago, they mention a climbing mission that will launch early 2022, which will see the “return of missions in Zwift”. Sounds promising.
It’s at about a page in, here: https://news.zwift.com/en-WW/208041-what-s-on-zwift-your-season-guide
Missions should be back soon!
Love the idea of the full bike customtizer, being to make your bike different to the guy/lass next to you is great.
Soon the Buffalo 🐃 bike will be the go to bike. 😉
Coincidentally i was struggling in a flat race today using the S5. Now it all makes sense. Definitely not because of chocolate cake and glühwein.
Definitely not!
Eric, you have the patience of a saint, dealing with all these shenanigans! Thank you for keeping us abreast of what’s going on.
Regarding the pages you’ve updated — “So today I updated the Frame Performance Charts and Tron vs Top Performers chart.” — you might need to look over the conclusions texts, at least for the latter of these two cases, as there are references to the previous results and the Cervélo being fastest, etc.. (But please enjoy Christmas first! Warmest wishes to you and yours.)
Fixed! Thanks.
Cervélo2020’s climb time (and percentile) in your frame ranking chart needs a little fix (+21s)
Damn right Roule, we are all blessed to have Eric as our hardworking & patient scientist to bring some light into the drop shop darkness. But if ZHQ would really change the groupsets (so the CdA) for all the frames, we wouldn’t see, hear and read Eric for weeks 🙁 … let’s hope that this will not happen
Lightly used Cervelo S5 2020, in green, for sale. Shucks. I’ll just GIVE it to whoever wants it!
I’d keep it. My guess is, once Zwift fixes the groupset issues, the Cervelo S5 2020 will be the fastest again.
or the Uranium would be the fastest on flats then, because it already has the Dura-Ace 9200 in its test numbers 😉
True, Carsten. Forgot about the Uranium!
Interesting, Eric. Have you tested the Specialized Aethos again as a climber as I started to doubt it in a hill climb race last week!
Haven’t tested it in a few months. Maybe I’ll that next!
I just tested Tron vs Aethos+rovals up reverse KoM on Innsbruck (220W @ 54kg) and Aethos was 18 secs quicker so it is still faster 🙂
Back to the Tron bike. That one doesn’t have a groupset, so it cannot be more aero or lightweight than that
I really hope that THIS IS NOT THE FUTURE, that we ONLY see a “GAME OF TRONS” … grrrrr … let’s pray for different brands which can continue to compete on a similar level with the Tron … being able to chose different frames & wheels (and outfits !) is INDIVIDUAL ENOUGH in my opinion … in a fast race, I wouldn’t be able to see any other groupset anyway … please, let it be ZHQ … use your time more sensibly
says guy who doesnt have a TRON X)
Hi Eric, I ride the Tarmac SL7. The rating in Zwift was 3/4 stars in the past and now Zwift updated it to 3/2 stars. I think, Zwift will update every frame in the future.
Yes, when they changed the SL7 to the new Dura Ace it’s ratings dropped.
It might be my last chance to set a PR on Alpe du zwift before all the frames get slower
Perhaps!
Thanks a lot Zwift. I just bought the S5 last night to use for the Rapha 500. I want my drops back!
Didn’t even realize I was riding with a different group set! Back to the Venge as the fastest bike now?
Yes Oliver, for today we can say : “welcome back Venge” as the top dog on flats.
But let’s wait and see tomorrow and then talk again 😉
The “ReVenge”
There is no link at all to Specialized being a Zwift Investor and the best Flat & Climbing bikes being Specialized…
Nothing beats a Tron around Neokyo though
So Specialized are back to having the fastest bikes again, not a coincidence me thinks.
Zwift should substantially simplify its equipment rules. Have 3 road frame types: aero, all-around, and climbing. Within each group and level, each is the same. Then have upgrades for each category with unlock level. These higher levels could represent higher end frames or higher-end components. Then do the same with wheels. Maybe frame levels are 10-22-34 and wheel levels are 5-16-28. But it should be transparent what the performance is.
As it is now — the safe bet is the Z1 so if you stick with that for most non-pure-climbing races on pavement you’re good.
“One of the many fun ideas rattling around in the heads of Zwift’s game developers is this: give Zwifters the ability to customize more of their bikes in-game. Right now all you can do is choose your frame (and sometimes its color) and wheels. But what if you could change your tires, groupset, pedals(!), brakes, seat, handlebars, even your bar tape?” Honestly, I think in addition to this they should consider making the bike rankings slightly less realistic and slightly more videogame-like. You constantly see the same bikes being used all the time in Zwift, and it’s because everyone wants… Read more »
Nobody I know gives a hoot what bike u ride. How about spend some time adding features instead of this pointless rubbish
The fact that they are _not_ spending time on this is why this bug continues to exist.
Agreed. Im not so sure why there is such a concern over virtual bikes. I’d also love for Zwift to focus their efforts on features and adding more routes and worlds to ride.
Who can see these damn groupsets in game anyway? So why bother? Like I said a few posts back, it’s all too many moving parts from Zwift to handle. Keep it simple/stupid and they’ll have less bugs to figure out.
I was just about to celebrate level 27 with the S5 frame. My virtual Christmas
present will have to wait…
Oh yes, seasons best to all and thanks for the hard work in keeping us all informed
The Pinarello Dogma F has no cranks since the last update, it might be the fastest frame ingame right now.
Maybe it’s time to drop all of this nonsense of bike x being fractionally faster than bike y given that it’s a virtual platform anyway and everyone could ride whatever they wanted. At most have a few categories such as TT bike, lightweight bike, all-rounder, MTB, Gravel and similar for the wheels. Less scope for bugs too
Agreed. But the problem arises when Zwift keeps trying to bite off more than it can chew. It seems they can’t manage to do much of anything without unleashing an infestation of bugs. Leave the damn bikes alone! I can’t see the damn groupsets on them when I’m riding anyway. So who cares? Start screwing around with bike and wheelset and groupsets and tires and tube weights and it’s just too many moving parts, apparently, for Zwift to pull off without problems.
Surely they get some money from these companies to get the bikes in game, no? Otherwise, why wouldn’t I be able to get my Otso in game?
What now??? Back to the Tron for me. Things are getting too confusing with frames and wheels and groupset bugs for my liking.
Well… all the new component groups are heavy compared to what came previously so technically it isn’t really a bug. Better shifting and braking and 12 instead of 11 doesn’t make a bike faster, and 1/2 to 1 full lb extra weight isn’t going to make it faster either.
Cool, Eric, thanks for updating the frame timings, much appreciated. On the one hand, it’s a pity to lose the performance on the S5, on the other hand, I like it that there is no clear favorite frame any more. Strong climber are poor in the flats and vice versa. It’s hard to find the perfect combination and that’s the charme, at least in my point of view. At least the top frames for climbing and the flats respectively are very cheap (Cannodale Evo and Uranium).
The entire concept is kind of silly — bike performance is essentially completely uncharted IRL, and there’s really no reason to start adding completely arbitrary values to frames and parts and throw them against an algorithm that doesn’t represent the real world anyway. Maybe the idea was that people could pick bikes on characteristics… but the thing is when there’s always a best option and unlike IRL people can have 25 bikes in the garage (or afford them to begin with) the alternatives are not going to get used. It’s also terrible marketing for a bunch of bikes out there.… Read more »
I believe the idea is to encourage users to level up in the game, in order to provide better equipment with this level advancement, and thus provide a small “advantage” in terms of frames in the events for those who spent more time and effort within the platform…..that old idea that better equipment will make you pedal better. Sweet illusion, even on IRL
With these changes in the performance of certain bikes on Zwift, I think the time has come when the option to return the bike and receive the amount or part of the amount (Drops) invested in the acquisition should be implemented
Different helmets have no effect other than to change the appearance. Zwift didn’t think through the logic of virtual props from the start
Eric, In the Excel/CSV download in the frame chart the climbing time for the Cervelo S5 2020 is listed as 2948 seconds. I think this is incorrect as it’s the original time and not the new time which from the above article should be 2967 (original time + 19 seconds). The flat time is correct. I’m not sure if this also affects the % numbers.
+21 (2969) … still written above 🙂
Hi, Erik!
Please take a look at Canyon Aerod 21 @ https://zwiftinsider.com/charts-frames/. After update you mention in this post its climbing ranking has become significantly worse and it is in contradiction with what is writen here https://zwiftinsider.com/fastest-frames/ and elswhere. I remember it is being fastest climbers among aero frames. Please check if it is right. Or did they change something with this frame as well?
Thanks,
K.
Zwift doesn’t get the algorithm for the in-game speed and the draft right, but they invest energy in the weight effect of a groupset? Eric, couldn’t you ask Zwift, if they will ever work on making the riding more realistic? You are my only hope.
My Dogma 12 doesn’t have a groupset or pedals at all. Is this a fault or is Pinarello looking to have the fastest frame in game. 🙂
Can you make a list of all the bikes affected by this bug?
I know that the CeramicSpeed oversized pulley thing is based upon a paid promotion, but there is still no independent confirmation that they are actually beneficial.
I don’t know what groupset is on the picture in this article but it doesn’t look like SRAM Red eTap to me. I’m pretty sure eTap has a battery sticking out of the rear mech and no derailleur cable on the front mech…
I have the suspicion that anything with these new groupsets, are also carrying the weight of a stock groupset or at least part of it in the frame. Since they were all one component before I doubt the subtracting the original out is working properly.
I am nearly at level 20 and currently ride the Aeroad 2021 for everything except the Alpe (when i switch to the Bridgestone). Looks like I should now get the Madone when I get to level 20 as it is now a better all round machine. But who’s to say that Zwift won’t then change the groupset (or its weight) on the Madone and screw up its climbing performance just as they have with the Aeroad??!?
Our of curiosity, how do you do these tests? I’m assuming it’s an ANT simulator or similar, but just wondering how you can be sure that a 1.5s time difference over 50 mins isn’t due to some other factor… How much do the results vary when you repeat the tests?
ANT simulator, yes. And we do each test at least twice – so the flat test twice, Alpe test twice. The half-seconds happen if multiple tests show the bike turning in a time of, say, 51:30 the first time, then 51:31 then next time, then back to 51:30. This tells us that the actual time is closer to 51:30.5 (Strava doesn’t do fractional seconds).
The results don’t really vary between tests, because we do them in an isolated environment (firewalled) so the test rider isn’t interacting with others. So they take a consistent line in the road, don’t draft, etc.
said they added Etap though.. not Dura Ace
Dura Ace was a while back. He’s just using it for comparison
I can see the problem…they upgraded the cranks and shifters to AXS which is wireless and combined them with….old gen shimano mechanical derailleurs, no wonder it doesn’t work well…
Specialized is evil… the Aethos has the “old” Dura Ace. the Venge S-Works has new Groupset. = Specialized wanted to have the fastest Climber und fastest Racebike back. it’s clear to me. give a little Extramoney to Zwift to put “a Bug” (😉) in the game to slow Competitors. Why am I sure, that it is no Bug? a Bug would be fixed sooner or later. but Zwift has no Intentions to fix that. So why they put the Dura Ace 9200 to the Tarmac SL7 you ask? I think because the Bike is cheap (<700k Drops) and early available… Read more »
i’m sorry, I forgot 1 Word:
the Venge S-Works has ‘NO’ new Groupset.