himespau 

Twice in the last 2 weeks, Zwift/Zwift Power has screwed up my results in races (that I have evidence for). What about my 200+ other races where I haven’t looked as closely/don’t have recordings and those of 10000s of other racers? Should we trust results? Are they just random nonsense?

2 weeks ago in the Herd Summer Racing League (results based on FTS points on each lap for primes and finishing points) on a 2 lap race of Casse Pattes, I entered the Marina Sprint at the end of the first lap at 37:27 and finished the sprint at 37:41 as shown in the following link (set to start at the entry into the sprint): . After I finished, the results list on the left hand side of the screen said I’d finished the segment in 13.50 seconds, which is in line with the start and finish times if you take into account fractions of a second. Not spectacular, but fast enough that it should have been 3rd in FTS for that segment on that lap.

So you can imagine that I was surprised after the race was over to check the Zwift Power results https://zwiftpower.com/events.php?zid=2980572 and see that I was listed in 6th for that segment with a time of 17.040 and a couple of the people ahead of me were people that I started the sprint segment after and finished before (i.e. I passed them during the sprint).

I contacted the race organizer who said that it was extremely unlikely that I’d have exactly the same time for the 1st and 2nd time through the Marina Sprint (1st time through I also finished in 13.50 according to the onscreen display and Zwift Power) and the difference must be due to the difference between what my local computer was showing and what the server had. I thought it was equally unlikely that lag would cost me 3.5 seconds in a ~15 second sprint (20-25% of the segment) and, if it did, how the heck are we supposed to trust when we should be sprinting in any segment?

I didn’t push it with him because those few prime points that I missed out on wouldn’t have pushed me from 2nd to 1st in the final standings.

Then there was last night’s fiasco. I race in Americas East, so I wasn’t part of the Oceana race where people got sent off in 3 different directions. Instead, I had just your average crappy race (got dropped from the front pack up the 1st KoM, caught back on, got dropped up the 2nd KoM, and came within a couple seconds of catching on right before the final sprint but didn’t quite do it) until the finish when everything turned to a flaming dumpster fire. I crossed the finish line in 15th at 38:47 (or so I thought) with a couple FTS points for the sprint (link to just before the finish – please pardon the anemic “sprint” to the line: ), but, when the results popped up onto the screen, my name was nowhere to be found.

Yes, I did go all the way through the blue virtual banner because why else would the results pop up on screen? Confused, I checked the Zwift Power results, and I wasn’t there (except for 8th in the sprint FTS). Looking at the live results, it said I was still out on the course.

I figured that somehow it didn’t register my finish and I wanted to earn at least some points for my team (3 from FTS and 1 for participation is less than the 23 I should have gotten, but better than none), so I rode the course for ~1.5 more laps before quitting so that it would show me crossing the finish a 2nd time (and doing the full distance a 2nd time) in case that was what was needed to show up. I finished and saved the ride after 63 km. In the ride summary (where it shows your PRs if you got them and all the ride ons you got) it shows “Completed Event” but I’m still not listed in the results.

Even this morning, when I go to Zwift Power and look at the live results for the event (https://zwiftpower.com/events.php?zid=2994808 ), it shows me as abandoning the race at 26.1 km (race length was 26.145) 1 min 30s after I crossed the finish line, when my ride actually continued for 37 more km and over an hour more time as is shown in the recording and my Strava results.

So that’s twice in 2 weeks that I haven’t gotten full (or any in one case) credit for my efforts. If I were to go back through all my other races, would I find more of the same? What about if everyone else did? Would there be lots of this? Would it render the results of all races meaningless? How am I supposed to trust Zwift and Zwift Power for results when they can’t get this sort of thing right?

Viich

Paul - part of the confusion is the side leaderboard seems to switch randomly from showing your last time through the segment and your best in the past hour....

I agree that the sprint timing system is a little wonky. I think it's likely that the displayed banners and actual timing points don't actually link up.

himespau 

Yeah the sprint timing is definitely weird. The leaderboard usually does that where it alternates between the best time in the last hour of everyone in the race (or on the course if you're not in a race) and your best times over the last 30 days. It was weird that my leaderboard was showing 13.5 seconds for both of my times through in that race but Zwiftpower had a time that was 3 seconds slower for the second time through. Oddly, when I rode my bike through that course once as warmup on Sunday, the leaderboard showed me as having the times that Zwiftpower listed for that day (so 13.5 and 17 rather than 13.5 and 13.5). That 17 doesn't line up at all with what's on my recording, but that's clearly what Zwift has on its servers for that time. I wonder if I somehow had a very brief internet dropout that coincided with me finishing that sprint and so my computer recorded when I actually crossed the line and the Zwift servers recorded the next time they saw me after I crossed the line when my internet signal went through again and I was 3 seconds further down the road.

After the whole my finish not showing up in the finishing results happened a 2nd race in a row, I talked to someone from Zwift who looked at my file and for the 2 races last week that it happened, there appeared to be a UDP dropout a few seconds long that happened at the time I crossed the finish line. Not sure what the odds are that I'd have the exact same dropout twice only at the finish line and only for a few seconds, but I've done 2 races since then when it didn't happen and the sprint timings are believable, so I don't know.

It's strange that if my timing for the sprint segment was just increased by 3 seconds due to a dropout (just my hypothesis), that Zwift wouldn't also just make the timing for my races 2-3 seconds longer and use when my connection was restored as my finishing time since the whole dropout was only 2-3 seconds both times. I wonder if it's something about the event was still going on for the primes but not for the finish time. Or maybe just weird **** happens sometimes and it just happened for me twice in a row after never happening before in the 200+ other races I've done.

Viich

Yea, the temporary dropout disconnection really sucks - you think you've finished, but it drops you out. You may well be right about that affecting your sprint times as well - if one of those short disconnections that you can recover from happens as you hit the finish line, it should delay your finish of the segment to that point.

I've done a lot to clean up my in-house RF environment to help my wifi, ANT & Bluetooth reliability. You'd be surprised what in your house is super noisy, and lining devices up in a straight line to the router can have some odd effects as well. There are some good things out there about 2.4GHz wifi channels, etc. - I'd say first thing is to force devices to 5GHz wifi if they're nearby to your bluetooth & ant+ stuff.

Do you have any new hardware? Fan, microwave, air conditioning on for the first time this year, new phone charger plugged in to same plug....... like I said, almost anything CAN be a source of RF interference. Might be worth a router reboot if you haven't done that lately as well.

Ed Wiser

Everyone is riding the race on their own computer and not on a sever. Zwift needs to make races to be on the same server if they want to get the results correct. Other cycling platforms already do this.

billridesbikes

Have you pulled the Log file from the session and interrogated it to see if you have ANT+ /BLE disconnects or excessive network traffic or Wi-Fi dropouts. I think the Wi-Fi can disconnect and reconnect in a few seconds and not have all the riders disappear. I agree with one of the other posters that I this could be a network or connection problem.

I think Zwift insider has a log file analyzer tool if your uncomfortable navigating the log file.

spelger

that's what i thought. i think that was one of RGTs claim, that the physics happens on their server.

himespau 

I used the Zwiftalyzer tool and it said that there weren't any bluetooth or ANT+ dropouts, just the UDP packet drop about the time of the finishing for a couple seconds and a very short duration increase in latency at the same time. I don't think I've added any new devices in the last month or so that would add additional RF traffic anywhere in my house. Neighbors (we live pretty close), I don't know. I had swapped from ANT+ FEC to Bluetooth for controlling the trainer a month ago because I found it easier do spindown calibrations in Zwift than in the Elite app, so I switched back so that there would be one less set of transmissions, but I doubt that's what's causing the issues as the timing doesn't line up that closely.

I have rebooted my modem (with my main route and my mesh routers - one of which my Zwift computer is connected to via ethernet cable - powered down) and then rebooted my wifi main router then mesh routers. I read somewhere that UDP dropouts can be caused by poor damaged connectors (and I do have cats that like to crawl anywhere they want, so I can't rule out them brushing on cables and pulling on the connectors), so I replaced all of my ethernet cables (modem to router, wifi mesh router to Zwift computer) with brand new cat 8 cables with gold-plated connectors.

Oddly, about the time that these dropouts happened, I noticed that it's been taking me longer to ping the AWS servers than it used to. Not sure if that's related or not.

billridesbikes

UDP is designed to be ‘unreliable’ in the sense that dropouts happen occasionally and apps and servers are expected to deal with it.

If you have a 5G connection and phone you can at least rule out the home network by running the next race on a 5G hotspot from your phone. Since Zwift mostly runs locally the amount of data isn’t much, 30-35Mb/hour something like that.

Ed Wiser

so does mywhoosh and VirtuPro.

spelger

TCP is designed to be reliable, that does not mean UDP was designed with the opposite in mind. but you are right, applications are to accept that fact and deal with missing frames. and the problem with this type of missing frame is that it might have gotten lost way before it even got to your router/computer. heck, the source of it may not even have sent it.

billridesbikes

I just knew that that word ‘unreliable’ was going to get me in trouble!

himespau 

That's kind of my problem in trying to track down where it is occurring to prevent it from happening again. In the last month or so, I've also had issues with websites occasionally needing to be reloaded (sometimes several times) before they will actually load on both my zwift computer (connected to a mesh router) and the computer that is connected via ethernet cable to my router, which is connected via ethernet cable to my modem, and this never used to happen (oddly, it doesn't seem to happen on tablets connected via wifi or streaming sticks or other smart devices).

That makes me think the problem (if it's related to the UDP dropout that appears to have led to 2 race finishes not being registered) is either at the level of my router, my modem, or upstream of my house (once I replaced my ethernet cables and ruled them out as the problem).

What I need to to next diagnostically (assuming the page reload requirement issue is related to the UDP dropouts), is bypass the router and directly connect my computer to the modem for the one computer that is within reach of ethernet cables and determine if that causes the problem to go away. My router (Netgear Nighthawk X6 R8000) is only just over a year old (so just out of warranty), so I hope that's not the problem. If I can show it's upstream of the router, I can complain to my cable company and see if it's the modem (also a bit over a year old).

spelger

finding the source of dropped frames is not so easy and expensive if you don't have the equipment. here at work i just get another switch with port mirroring functionality from IT and capture traffic then compare. it can be very tedious though.

the web uses HTTP and that is typically over TCP, not UDP. dropping a single TCP frame is not a big deal and TCP will re-transmit if needed.

if the phones and tablets work fine over wireless then that is a clue, it tells me that the modem is very likely OK as well as *some* part of your router. try to get your zwift machine to use the same wireless network. some wireless routers are multi band so make sure you are on the same band. it is also possible that the dropouts are caused by your wireless adapter in your zwift machine. you can try a real Ethernet cable if practical.

atwl77

Netgear, huh.

Might just be a coincidence but in a simracing game that I play, iRacing, it is well-known among the community that Netgear routers don't play well with it, the developers tried to work with Netgear to resolve it but that didn't pan out, so anytime someone on the forum asks about network issues one of the first questions will always be, "are you on a Netgear router?"

himespau 

Huh, interesting. I found out the hard way after building my "gaming" pc to use in Zwift that Zwift doesn't work well with AMD processors or GPUs. Not working well with my network would be just the whipped cream on top of it all. It actually hasn't had a problem until a month or so ago, and then my desktop that I don't zwift on and my zwifting computer have been having issues, so I'd guess it's either my modem or router that is failing (or something upstream of my house in the signal passage). Just looking right now, the light for internet coming in on my router is a very constant flicker (almost solidly on, but not quite) and the ethernet out is the same (while all the wifi lines out are solid white so I can see the difference), which might imply to me that it's a ethernet signal coming into the router issue.

unterhausen

I think it's a fair way to say udp isn't guaranteed to be reliable. The funny thing is how many people have reinvented tcp in the effort to make udp more reliable. I have used both, there isn't much difference if it's on a local network. Zwift would no doubt crash if they used tcp though.

himespau 

Same stupid thing happened to me again. Third time in 2 weeks. And this is a race that I would have won (by 26 seconds - an iTT that I finished in 18:09 and the winner of my cat finished in 18:35) had my result counted. Super frustrating. Was well rested and feeling in good form (probably could have pushed a bit harder, but from 10-18 minutes my power only dropped by 3-4 W. Had I held it for 2 more minutes, it would have been a new all time best 20 minute power, so it's a shame it didn't count.

Rxffy

Howdy, you might-could use a program like Wireshark or Iperf to determine if your packets are being dropped/rejected/not sent, and the relative latency involved.

It's not extremely simple to use, but it isn't too complicated either

himespau 

Thanks for the idea. I finally got around to bypassing the router and running the ethernet cable straight from the modem to a desktop that is usually attached to the router. Suddenly, any page loading issues that I'd been having went away. Put the router in the loop, and they came back. Swapped order of the cables and the same thing happened. So that tells me that, at the very least, my router is acting as a problem. As these things usually go, the manufacturer's warranty on the router expired 2 months ago. Fortunately, I went out of my usual habit and bought the extended warranty when I bought the router. So now I just have to figure out how to deal with Best Buy's extended warranty program.

spelger

see post #14. don't assume the wireless is ok in your PC or laptop (whatever you are using).

himespau 

My zwift machine (home-built Windows 10 PC with 16 GB RAM, AMD Ryzen 5 processor, and AMD Radeon RX560 GPU - built before I knew Zwift doesn't play nicely with AMD) has a wifi adaptor built into the motherboard, but I'm using that machine with an ethernet cable connected to the local node of my mesh network instead because running that way cut the lag from what I was doing on my pedals to what I would see my avatar doing on the screen from 2-3 seconds (when the machine was exclusively running on wifi) down to 1-2 seconds (same everything but machine was connecting to that node of the mesh via ethernet instead of connecting to it via wifi). Since my tablets seem to lose the internet (often while saying they're still connected to the network) from time to time, I haven't tried racing exclusively on wifi after this problem started because it didn't seem like that would improve signal reliability.

It could just be that one node of the mesh network has gone bad, but that wouldn't explain that the machine directly hooked to the main router has problems loading pages (some pages more than others - oddly it has a lot of trouble with Zwift Power and other Zwift-associated pages) when connected via ethernet cable to the router but not when connected via ethernet cable directly to the modem and the fact that tablets sometimes randomly transiently lose connection to the internet (without usually losing connection to the wifi network) no matter where they are in the house (much closer to the main router or the other node doesn't prevent the drops).

Because, when the tablets lost connection to the internet they often said they were still connected to the network, I had been assuming that the problem was with my modem (or even further upstream with the cable company) rather than my router, but, because the page-loading problem (I understand, that's TCP vs UDP and might not be related) went away when I connected that desktop directly to the modem, I'm now expecting that at least some of the problem is to be found in my router. Unfortunately, the Zwift machine is too far from the modem to physically connect via ethernet cable. I had considered trying to use a powerline connector, but I want to know where the source of the problem is because installing it downstream of the problem wouldn't fix anything and just be a waste of money (though, should I be able to fix the problem, I've been told that running a powerline adapter directly from my router would decrease the lag even more).

It has seemed weird to me that I'd be having these problems with my computers and tablets, but it hasn't seemed to affected my Alexa-devices (echo and echo dots), my streaming sticks, my smart TV, or my internet-connected remotes. I don't know whether those use a different type of signal, smaller packets of data that can slip through even with dropouts (or just be re-sent automatically until they get through), have more buffering and are less sensitive, or what.

Until I started not getting results recorded for races due to apparently getting dropouts at the end of races (did I mention how pissed I was that I had the fastest time of anyone in my cat in any of the 4 instances of the HErd HEries individual TT on Tuesday only to have my result not be recorded?), I didn't really spend too much time trying to hunt it down because, while annoying, the problems weren't really overwhelming - web page issue is usually solved by just hitting refresh a bunch of times, tablet wifi issue usually solves itself in a minute or less or when I switch to a different network frequency (router puts out 2 x 5 Ghz channels and one 2.6 GHz channel and I have an extender covering part of the house that's not on the mesh network - a relic of when my kids and a couple others in our pod did schooling at our house during the shutdown).

And yes, I've checked to see if Netgear has any firmware updates for my Nighthawk X6 R8000 router.

spelger

i'm a little confused and i don't really want to read this whole thread again, i thought earlier you said that your wireless tablets/phones were working ok, but now they are getting drop outs? and so too is your machine used for zwift? what is common here...the wireless router. borrow another wireless router and try that.

you also indicate that going ethernet directly zwift works fine. also points to the wireless router since the tablets have this problem as well.

had to look up the power adapter thing, i would be surprised if that helped at all, more than likely i owuld expect it to help cause more problems than fix any. house hold wiring was not designed with high speed communications in mind. you have no idea what those wires are next to and none are twisted, they might owrk to a certain degree but i would not put much faith in them.

all of your other devices appear to work more than likely because they use TCP and not UDP. and having larger packets is more efficient than smaller ones.

himespau 

Sorry, had a detailed response and lost it.

Long story short, the tablets working fine that I posted much earlier was because I only read long form journalism on the tablets while half-watching a show I don't care about, so I don't notice if there are connection issues, but I got that there is a problem from talking more with my wife who uses them around the house. It seems that there are times when it still says it's connected to the network but there isn't any internet. Then, it either fixes itself before there's time to do anything or you just have to switch to one of the other networks (each frequency is listed as its own network) and it connects just fine.

The Zwift machine I can't hook up to the modem as it's on the opposite end of the house on a different floor with too many finished walls/ceilings between. The machine next to the modem in my office loads pages just fine when I bypass the router, but frequently throws up "This site can't be reached. DNS_PROBE_FINISHED_NXDOMAIN" errors when the router is in between the two (reload the page a couple times and it goes through).

The problem with testing for Zwift working is that there is no visible sign that there's a problem (in the first 2 races I could see racers all around me as I sprinted through the pack to the finish line - the third was a TT and I was alone) until I cross the finish line and my name isn't to be found among the list of those who finished that pops up (the list pops up, so it knows I finished locally, just Zwift can't tell that I did or when on its servers so it pretends I didn't). In the video below, you can see one person 4 seconds ahead of me and then you can see other people (from different races who were held invisible when I was in my race) appear once I cross the finish line, but my name isn't one of those on the leaderboard that pops up.

But yeah, I think everything is pointing to it being a router issue. I just hope that I've gathered enough convincing evidence that the Best Buy extended warranty folks will replace it without too much trouble. IF they do that, I just have to set the network back up (takes me a while), and then, hopefully, things will work again.