Zwift drafting is broken

Right. As I said, I think that the ‘delay’ in the draft effect is part of that problem. I see it on descents and on flats. If you were affected by the draft the moment you were in the draft zone, I don’t think this would be happening to the same degree.

Thanks Tom I feel I understand the point of your posts now. I suppose I replied saying ‘you need to match their speed not their power’ when I might have been more accurate saying ‘it requires a combination of speed and power’.

You mention using your steering to place yourself in the bunch. I don’t know how Zwift works but I’m sure as soon as you are in the bunch you expect that draft. Zwift is not IRL but of course IRL the bunch normally don’t let you in so you remain generally on the outside of the bunch, not fully in the draft hit the back of the bunch and find you need to put out more than expected to stay attached.

I don’t know if there are differences with draft when moving forwards and when moving backwards?

Yeah, the ability to magically slide into the middle of the bunch is a zwift-only thing :smiley: But given that my avatar can be placed there, I would expect the draft that should be available in that spot to be available.

IRL doesn’t let the bunch chasing me take hairpins nearly as fast as they do in Zwift, so there are definitely differences. Solo rider should have the advantage on descents through corners. But things like that are fine–they’d have to completely overhaul how the bunch works to make things like that happen, including not allowing avatars to phase through each other.

To be clear, it’s not so bad that I don’t keep using Zwift, and I’ve know that it happens pretty much from the start, so I just adapt. But I do see it as something that could be fixed with drafting all the same.

I’ve seen a similar sort of delay from within the bunch too. When I’m just happily sitting in a bunch, I find that the best way to go isn’t to try to maintain a set position. Rather I drift forwards and backwards. And the way to do it sounds counterintuitive, but it’s a similar–or maybe the same–delay.

For example, just picking a point in the iteration as a starting point: I’m at the back of the bunch of lets say 30 riders, getting (at this point) the full draft effect, and I start sliding forwards through the bunch. I will shift down to an easier gear, my power will correspondingly drop, but I will continue to slide forward through the bunch. I’ll reach the front, or near it (not always right on the front, in fact I try to manage it so I don’t hit the very front, and thus am still in a draft), this way, and at that point, having geared down, I’ll stop moving forward. A couple of seconds later, I’ll start sliding backwards. At that point I’ll shift up to a harder gear and put in slightly more power. But I’ll continue to slide backwards for a good while…eventually slowing my backwards drop, and then starting to slide forwards again. So, downshift, less power, keep sliding forwards–when I start to slip backwards, upshift, more power, keep sliding backwards, and repeat.

It’s the continuing to slide backwards while I increase power, and continuing to slide forwards while I decrease power–that’s the unintuitive part. But as with trying to catch on to the bunch, there’s a delay, and then the change in power starts to have an effect.

Maybe there’s an in-game explanation that shows IRL physics is actually being followed. But it doesn’t feel that way. Not a huge deal, I manage it, it’s fine. But it’s odd.

I also see what @Tom_J explains on descents.
Recently I rode Island Hopper with pace partner a few times over several days.
After the Sisha Sprint, there is a descent. Several times, at the same corner, the group was behind me, and blew past me, and I had to really up the power to keep from losing the group.
The same thing happens every time on the descent coming off the south side of Tempus Fugit, past the Titans Grove turnoff heading towards Ocean Boulevard. Any small group of riders will fly past me, even when I’m putting out over 200 watts.

Yeah this is “normal”. You have to anticipate the arrival of the group and really crank up the power when you’re like 20-25m ahead (depending on how fast they are going). Any delay will result in disaster.

it is. the inertia you pick up from the draft, you have to wait a little longer for. it really isn’t much different in real life, at equivalent speed

with experience and timing you can pretty easily account for this by bringing your speed closer to the catching bunch before they make the catch… it sounds obvious, but it takes time to accelerate. maybe a little more in zwift than in real life, but it’s the same principle

the opposite also holds true. if you’re making the catch, you can more or less sit up at 15m and coast on. you already have the inertia, and now you’re “gaining” the draft too

however, to keep it on topic: its almost impossible to do this at 90kph on zwift on a -10% gradient. probably in real life too, but in real life you can make up the difference in the coming corners

I’m not sure how you’re using inertia here. An object’s inertia is determined by (directly proportional to) the mass of the object. Wind resistance combats the speed of course. But the mass of the object isn’t changing, neither are any other frictional forces (tires on the road) or gravity, as the wind resistance changes. There’s no change in inertia, only a removal of a force combating the forward motion of the object. Once that force is removed, the change in the forward motion of the object should begin immediately.

Imagine we setup a treadmill running at whatever speed, and put a strong fan blowing from in front of it. We put a remote controlled car on the treadmill, and we set the power of the car so that it was maintaining equilibrium–driving ‘forward’ on the treadmill, into the wind from the fan, and not moving forward or backwards. And then we shut off the fan. The speed of the car would be immediately effected–it would move forward on the belt, right away. It’s true that the acceleration we’d see would happen over time–no acceleration is instantaneous. But we wouldn’t see a delay of 4, 5, 6 seconds while the car slowly adjusted to the lack of the wind resistance. It wouldn’t just stay in the same place on the treadmill with no effect for that long, and only then start speeding up.

What I’m saying is that there is no noticeable effect of the draft in those initial seconds. It would be understandable if for example the group overtakes you at 10kph faster relative speed. The wind resistance would go away, and your acceleration from e.g. 30kph to 40kph would have to happen over a duration. But it wouldn’t be the case that there would be no acceleration for 5 seconds, and only then it would start. But that’s what I see happening. And I see this happen at very similar speeds–1kph, 2kph, etc. A change of 1kph doesn’t take that long at reasonably power either.

What I’m seeing does not seem to be this: Enter Draft Zone → begin to accelerate → eventually reach new speed.

What I’m seeing seems to be this: Enter Draft Zone → No change in speed or acceleration → No change → No change → Begin acceleration → eventually reach new speed.

It should be the first way :slight_smile:

Autobraking seems to have been kicked up a notch or two I’ve noticed in the past month.

I’ve been getting kicked out of packs on flats and descents whenever heavier higher wattage riders kick it into gear.

I don’t know what’s changed, but I’ve seen more flashing red watts on my screen in the past 2-3 weeks, than I had since August during the 4.1 release.

No idea if anyone else has noticed it recently; but I certainly have. Something has changed with PD / autobrake this year, not sure when it started; but much like the “steering is now fixed, steering riders draft non-steering riders”… it’s definitely not fixed.
And I wouldn’t be totally surprised if this is when this increased braking situation was knocked on as well.

Your recording though is absolutely weird; that looks more like the game’s physics code breaking because it doesn’t know what to do with two people in supertuck when one ‘technically’ shouldn’t be in supertuck.
Almost as if your game thought you weren’t in supertuck when you visually were; but the physics side of the game decided you weren’t.

In your case however, I 100% blame steering devices being enabled, seriously.

As for OP, welcome to being a lightweight in a world not meant for “us.”
Learn to push over hill crests and keep momentum up in Zwift. It does matter.

As I showed in my data analysis above, the OP was coasting on the flat part of the hill (even in the 1% bump) that is why he could not keep up.

It is very hard to gain speed on a downhill since you have very little resistance.

If ever there was a BS statement…try climbing at 90-100kg and see if you feel the same way.


before i read all of that, which i will, i basically just pick words at random and assume people know what i actually meant. it usually works, so forgive me

it is true that zwift is slower than real life in this respect. at least, i share the same feeling. the principle though is the same. as to why it’s slower, i can guess. 1: it’s an online mmo, using a global tick system… i’m not a dev, so i won’t butcher an explanation of what that is and why they’re used, but i think the broad strokes are “money” and “it just be like that”. and 2: the overall speeds on zwift are just really quite high compared to irl, which does magnify the effect. there’s probably a 3 and a 4 etc but for me, it’s consistent and learnable, so i just take it for what it is. things only start to bother me on a personal level when they’re inconsistent

it’s probably worth pointing out also that you can’t physically “feel” draft on zwift, there’s no feedback like you would get from say, the backdraft of a truck close passing you at 70mph

Doesn’t mean there couldn’t be more implemented to give a better understanding of what’s going on.

Things like increasing wind noise when not in the draft / decreasing when in draft, along with visual cues other than the “close the gap!”, could and absolutely would go an extremely long way.

As silly as this statement might sound for many, Grand Theft Auto V has a surprisingly good arcade racing mode built into it, and it has basically what I’d refer to as wingtip vortices that appear behind the cars in front of someone drafting, which is a highly effective piece of feedback to know if you’re doing the right thing or not; along with simple things as I mentioned… less wind noise (very common simple effect in racing games; and accurate to real life too for the record).

So it’s not to say Zwift can’t have things to help a situation like this… it just doesn’t [for now] (if ever).
One can wish though.

It’s not about whether speeds are realistic, it’s about whether things largely behave as they should. To put it another way, it’s about precision as opposed to accuracy.

If Zwift made us go more slowly on descents than on ascents, or if getting in the draft made us slower instead of faster, we wouldn’t shrug and say “well the speeds aren’t IRL accurate anyway”. We’d say “things don’t work like that”.

In any video game, there’s a certain level of realism that has to be achieved in the physics, even if the characters are jumping 50 meters in the air and shooting lasers. In Zwift, there’s a certain level of realism that should be (and that they try to make sure is) maintained, even if our speeds are a few kph off IRL. To me, this ‘draft delay’ is a borderline case. It shouldn’t work like that, but it’s not a big enough break from reality to ‘break the game’. But it does effect my riding and racing, so it does bother me.

Compare that to those times when my wheels seem to be halfway embedded in the tarmac at intersections. Things also ‘don’t work that way’, but it doesn’t affect the mechanics, so it’s not an issue.

the two things are actually directly related. granted, i deal in bro science, not actual science, but there’s a lot to be said for broscience, which i will define here as “more or less sort of knowing how stuff works after doing it a million times and skim reading a few papers, ignoring the ‘MRa’ part of IMRaD”

nothing in zwift, how the physics work, or what happens in zwift in any given situation on any given day, in any given event, catches me by surprise. to the immense credit of zwift, there are very few errors, glitches, or bugs in that sense (i won’t say none). so realistic or not, i don’t mind stating confidently that the model they use is both precise and accurate, at least in the sense that it’s true to itself. we agree that it’s not the same as outdoor cycling, well, i think basically everyone with an opinion on zwift at all does

cda, aerodynamics etc, are something i have very little actual understanding of, but i’ve spoken to people who do, and have been told that i am correct on this: when the speed is abnormally high, which the speed often is in zwift (fun is fast etc), the amount of attention, power, and precision required by the rider to adjust for anything that happens is amplified exponentially. if a group passes you at 25kph, you can ride onto it after the fact with just a little effort, it’s no big deal. if they pass you at 55kph, you have to be ready for it, early. it works the same way in real life. but in zwift, if they pass you at 100kph, well, i hope you were doing at least 98kph yourself when they passed you. The broad assessment by people i’ve spoken to who know cda, aerodynamics etc, is that zwift are extremely generous in the cda values they assign to all riders at any given moment, leading to fast speeds, leading to an exaggerated requirement to be able to antipicate things ahead of time. apologies if i sound like i’m trying to explain 2+2 to a scientist here, but this is literally how i see the world myself and so far so good, if i’m being honest.

but these exaggerated speeds cause problems in certain scenarios on zwift, like specific sections of specific descents, because the speeds are so high that the user input required to adjust and anticipate is beyond what can be realistically compensated for. the descent of the radio tower, that one section on the the innsbruck kom through the village that everyone has definitely been dropped on at least once, bologna etc. they could also just be scuffed too, there’s always that.

I get what you’re saying, but I don’t think it’s the same thing. That speeds are exaggerated doesn’t mean that air resistance shouldn’t work like air resistance works. Again, it’s that ‘threshold level of reality’ that I was talking about. If my zwift bike wheels turned into pineapples at random intervals–not wheels that looked like pineapples, but pineapples with pineapple physics–I wouldn’t use the app. If being lighter made you descend faster, I wouldn’t use the app. If lighter bikes made you climb slower, I wouldn’t use the app. Those aren’t just a matter of ‘degrees of accuracy’, they are just ‘wrong’.

But the speed being ‘slightly too high’ is not the same thing. That’s just a matter of small degrees.

And to me, the draft effect being delayed by a bunch of seconds is borderline. I don’t have much more to say about it, so I’ll just let it drop, as I’m just saying the same thing over again here. :slight_smile:

i think it does. as a bro scientist, i cannot for the life of me remember the title of the paper, but i can remember zhq staff, davidp i think in one of the revised pack dynamic beta threads, referencing a specific paper detailing the aerodynamic model that the zwift physics engine is based on in a forum post, and more importantly the distinct lack of posts in reponse to that post telling him “that model sucks and it’s wrong”, the kind of stuff that normally follows anything a zhq staff member posts on these cursed forums. I think there are some quirks in practise that make it feel unusual or maybe even unintuitive in a virtual cycling setting, which i have some personal opinions about, but i also agree with you that i think we are both at the end of our limits of our vocabulary here trying to find the right words to explain what on earth they are

1 Like

I guess I’ll add this–I don’t doubt that the intention is a good model. I just think that it’s not working right in this one particular way. There are so many moving parts to the code running Zwift, it is never surprising to me when they discover interactions that make things not work as intended.

If it was me left to program this, you wouldn’t even get Excitebike levels of software as a result :laughing:

Zwift’s drafting has never been anything close to IRL. Somehow Zwift figures that racing is improved by altered draft physics.

Originally we had “half draft”. Then we had “double draft” - which was in fact “full draft” but only for the 1st rider following…? I have no idea where we are now. (Latest Zwift Insider testing indicates that the maximum draft benefit is probably about 36%.)

I’d love to see a realistic draft mode. Make it optional for race/event organizers and see what happens.