Team TT: strange differences in experience/avatar position

I ride with three team mates in the TTT on Thursdays. We are slowly getting better/more organised. Last week, for most of the race I was hanging on to the back of my team for dear life and having to put down a lot of power (for me) just to stay there, even though we are all roughly similar w/kg. I reckon I only did about 20% of the race on the front. Chatting with my team mates after the race, they said I was “on fire” and had been scrambling to hold my wheel and that I had been on the front for over half the race. Their experience and my experience of the same race was different, even though we stuck together for the whole course. Any explanations/suggestions most welcome…

2 Likes

Yeah we had exactly the same experience and have been having so for a few weeks now .

TTT is exposing a real issue in a way that makes the event extremely challenging if Zwift dont acknowledge and fix ( We think this has started m or got significantly worse on the 21st Jan release .

That’s because Zwift doesn’t show the same position of riders to every user.

Are you stating this as an expected ( and accepted ) behavoiur ?

If so the game is not sustainable , and think about it , no other game would be either . Think Call of Duty when you tried firing at someone and actually they were not there at all … No one would bother playing.

Thankfully for online interaction in all those games accurate plotting of avatars on the screen is a given and minimum standard which is delivered .

So what gives @ZwiftHQ

For sure i would not expect it to be 100% but it has to be reasonably accurate or else group racing and especially team time trialing becomes a nonsense .

The experience anyway has substantially degraded over the last few weeks.

Yes it’s expected behaviour. Check out this finish from ZRL04 the other week. Ovett crosses the line first by half a bike length but is 2nd in the official results - Zwift Racing League Season 2 // Race 4 - 2019 UCI Worlds Harrogate - Points Race - YouTube

And here’s a game of spot the difference for you, this is two PC’s on two different internet connections following the same rider at the same time. Look at how much variety in the rider positions there is

4 Likes

Pretty damning evidence on the game mechanics . Spotting difference , Its not only relative position ( e.g that could be put down to network lag) there are riders on completely difference tracks and formations relative to each other .

The the ultimate discrediting evidence zwifts own broadcast demonstrating the bug/flaw for all to see !

What I originally thought was a bug looks more and more like an out and out flaw now :frowning:

2 Likes

Well, its reassuring to know that others have seen this and we are not going mad, but I agree, it really does detract from the race experience. I hope they are planning to fix it, but it doesn’t feel like an easy/simple fix…

Agree with Mike - totally expected. Been there since my day 1 and I’ve even lost podiums (and probably won them too) because of it. And yes, if you watch streamed races where you can see multiple views, its pretty obvious, as is Fan viewing a race and switching around riders. Someone from Zwift mentioned a few years back that it was network latency, and given the variances that Zwift tolerates, completely makes sense with the current engine. Call it a bug or design flaw, cant see Zwift addressing this anytime soon.

It certainly seems embarrassing for Zwift given that are trying to raise the profile of e-racing crappy broadcast quality pictures and commentators calling out one result while the screen shows another. When I get wound up about it I just remind myself its just a game.

1 Like

So Dean, its not related to the recent upgrade? Sounds like you have been seeing this for some time? At first, I thought this was down to different network latency between team members but the effect seems too marked/consistent for that to be the case. Would be good to get a definitive position from Zwift “this is intended behaviour, we do not plan to fix it” vs “this is a known issue and is scheduled for a fix in Q3” or something. Meantime, I guess its the same for everyone and per Graham’s post, there are team behaviours we can adopt that might reduce the impact…not ideal though.

Could well be getting worse (after all, this is Zwift!) but to be fair, I’ve not been racing for past few months (nor watching many streamed events) due to summer where I am. Given a few of you are again mentioning this, it could well be getting worse, or, it could be a new bug exaggerating the old.

That will never happen. The only bugs they tend to admit to are the ones that impact significant portions of the paying community (and then its only when the masses have got their pitchforks out).

Zwift has a bunch of issues that impact us all. Some of us notice, others dont. I take solace in the fact that they generally impact us all so at the end of the day, its mostly a level playing field.

What you guys could do is take your recent evidence (logs, vids etc) and log them with Support and see what happens. Its worth a shot given this hasnt popped up for a while. In your case Chris, get your colleagues to provide you with their log files for the event that you referenced above, as well as your own and perhaps get each of them to write a quick paragraph on what they witnessed.

Agree re level playing field, and I am guessing that the number of peeps doing TTT compared to the entire use base is relatively small (although there were over 5000 participants in last Thursday’s TTT across all timezones, presumably most if not all paying subscribers?). I thought about assembling the evidence from all four of us and submitting a support ticket, but not sure we want to go to that effort if there is no prospect of a fix.

What we could do is capture video from different team members next week, at least that will prove it.

1 Like

I do think it has got worse in the last few months. Doing a group ride the leader will call out their position as 140th and I’m in 120th place 10 metres behind the yellow beacon. You could be riding near the front and one moment your position is 5th then the next it’s 20th and you haven’t changed position.

Also, the nearby rider list can often be out of order or displaying the wrong data. A rider 7 seconds back will be ahead of one listed as 4 seconds behind for example. Which is correct? The time difference or the position? Or is it all just a rough estimate anyway?

3 Likes

OK lets predict then its a flaw . I think Zwift need to step in here and give us there perspective . If Zwift dont want to fix the fundamentals that they have to be listening to the fact this issue has got worse in the last month , seriously exposing this problem .

And lets be clear whatever they want to pin it on . The simple fact is that there are other hugely massive MMO RPG games that can deal with infinitely larger user bases without this . And a rival product appears to be able to do so also .

The main problem there is that the absolute position, time gaps and the sort order are updated independently of each other rather than all at the same time. So in isolation each couldl be correct but seen together they can get weird.

This really depends on how the data is mastered and delegated and if it has good reliability patterns built into it . We really are now into speculation and guesswork .

Its really up to Zwift IMO to explain what is going on , why it is happening , and how they intend to fix or at least mitigate it to bring this into acceptable levels . If it cant be totally fixed at least give users some indication of how much latency in there connection.

However look at the specific picture , there seems to show this is NOT a network latency issue ( alone ) . Different clients are not just seeing time delayed rendering and positioning they appear to be calculating the postions locally and that can only end up with absolute chaos if it is actually what is happening .

agree, doesn’t look like a latency issue. I can forgive Zwift getting a bit random as it tries to figure out the real time placing of a 20 person blob in a 2000+ person group ride, very dynamic/complex situation. But 4 riders, essentially on their own? If TTTs are going to be worth the considerable effort, that needs to be reliable. If each client really is computing its own (different answer), suggests that not enough data is being distributed from the center (which of course would make latency a bigger problem) and the client has too much automomy?

We’ll capture some video and raise a support ticket and see what the response is.

1 Like

There’s certainly some interpolation going on client side. Have you ever seen an avatar do a full 360 while riding with a group? I assume that is the Zwift client continuing to move an avatar without recent data. When the client gets the data from the server the rider might have slowed and is now a bit further back than the client had assumed so the avatar does a 360 to circle back to the new updated position. Without actually looking at network traffic I doubt Zwift is updating each rider’s position from the servers to the clients more than once a second. It’s probably more like once every 2 or 3 seconds.

I also suspect that changes Zwift made to handle the large volume of riders during Tour de Zwift may have extended the update rate as well. Ideally there would be some sort of dynamic system to adjust the update rate, 5,000 people in a ride would need a longer update time to avoid overwhelming clients, 4 people in a TTT should be updated more frequently. Does Zwift dynamically adjust update rates? Your guess is as good as mine but considering the general quality of Zwift coding my guess would be no.

2 Likes

I think you are hitting on a very plausible theory in your final paragraph that indeed they have attempted to solve some scaling issues and done it in far to simplistic and badly targeted approach . So simply upping the base sync rate across the board and not smartly (or with an algorythm that needs serious refining) , we would surely see the results . Interestingly our teams with 3 riders in the TTT were far more inconvenienced than the 8 riders teams adding weight to this .

Check out this footage from a broadcast Zwift race. Look how much the riders rank changes in comparison to the position of the avatar in the pack - ZCL ZRL 230221 - is it 1st, 21st, 10th, 13th, 18th | ZCL ZRL 230221 - is it 1st, 21st, 10th, 13th, 18th | By Zwifting Bloopers | Facebook

1 Like