Login required to started new threads

Login required to post replies

Does Zwift actually read the feedback it solicits for Pace Partner rides?
Quote | Reply
At the end of every ride on which one rides with a Pace Partner, Zwift solicits input. One question for one's experience, and the other for whether one would ride with a Pace Partner again. Furthermore, Zwift provides a prompt where one can provide input. I'm wondering if Zwift actually processes any of this, or possibly decides to do anything about the input.

From what I have noticed, other than superficial changes such as labeling the ride by w/kg of the respective pace partners, as opposed to broadly indicating the category ranges (e.g. 4.2 w/kg, as opposed to 4.0 - 5.0 for the A-level pace partner), there does not seem to be any meaningful change made to the pace partners and the group riding dynamics.
More importantly, I don't know if anyone else has done this, but the rides are a lot more difficult at the busier times than they are at less busy times (say ~18:00 CET / 12:00 Noon EST for the former, vs 1:00 CET / 7:00 PM EST for the latter). In the B-level ride, for the former, the group with the pace partner is easily 30 strong, and average speed through the flats is about 45 kph. In contrast, for the latter, the group with the pace partner may have ~10 riders, and average speed through the flats is about 42.5 kph.

It's just ridiculously unrealistic to have a group chugging at 45 kph, despite the fact that no one at the front is doing more than 4.0 w/kg. Those who have done Zwift TTT knows just how ridiculous that is, given that the lead puller has to do at least 5 w/kg (if not more) to go 45 kph through the flats. The worst part is that given the pack more or less moves along at ~45 kph, whenever it approaches other riders, the front of the pack actually experiences a lower CdA that allows it to speed up even further, whereas the rear of the group sees no effective decrease in CdA, thereby producing a very stretched group should the rear not pick up power drastically (as in doing at least 4.0 w/kg, if not more). Then when a gap is opened, which is almost a certainty at the busier times given the stretching, one has to chase onto a pack moving at 45 kph, as opposed to one moving at ~42.5 kph. In contrast, there are a lot fewer gaps that get opened at the less busy times

The same type of effect is observed in the C-level ride, where the group chugs along at ~41 kph on the flat, despite almost no one doing over 3.0 w/kg. It would seem that the unrealistic speeds and the pack getting stretched is almost invariably due to the large number of riders than anything else. Worst is when the pace partner sticks actually helps to split the group by sticking to the front (still doing 3.2 w/kg) while those chasing at the rear barely makes any gains while going at 4.0+ w/kg.

I'm sure Zwift knows of this (because Zwift solicits opinions at the end of every ride on which one rides with a pace partner), does it plan to do anything about it?

Also, while we are at it, I could really do without the inanities of the pre-generated quotes from the pace partners, some of which are just downright awful.
Last edited by: echappist: Mar 6, 21 19:03
Quote Reply
Re: Does Zwift actually read the feedback it solicits for Pace Partner rides? [echappist] [ In reply to ]
Quote | Reply
The larger the pack, the lower any given rider in that packs CDA is. It's not a great method, but it's the Zwift method, so it is what it is.

In response to the question in your title, I'm not convinced Zwift reads any of the feedback it gets, about anything really.

If you are interested in a more realistic draft, I'd suggest giving RGT a try. It has it's own set of worts, but it's a much better cycling simulator (as opposed to cycling game).
Quote Reply
Re: Does Zwift actually read the feedback it solicits for Pace Partner rides? [echappist] [ In reply to ]
Quote | Reply
The dynamics of the B and C pace groups are very different:

-In the C group, Coco is almost never on the front and always has a draft, so the 2.5W/kg is what someone in the middle of the pack needs to do.

-The B pace rider is sometimes drafting, and sometimes pulling (riders have figured out that you can slow the pace/rest a bit by forcing the pacer to the front). Also, since the pacer always does 3.2W/kg, the climbs you hit on the Titan's Grove section tend to be easier and the descents harder. Most people's natural output fluctuates on the rollers, the pacer's doesn't.

The A pacer is the complete opposite of C. It's almost never drafting, just dragging a few tortured souls around Watopia.


Personally, I'd love to have the pacer be unable to draft (similar to wo mode or a tt bike), but riders can draft the pacer. That would go a long way to stabilizing the group paces.

ECMGN Therapy Silicon Valley:
Depression, Neurocognitive problems, Dementias (Testing and Evaluation), Trauma and PTSD, Traumatic Brain Injury (TBI)
Quote Reply
Re: Does Zwift actually read the feedback it solicits for Pace Partner rides? [echappist] [ In reply to ]
Quote | Reply
echappist wrote:
Other than superficial changes such as labeling the ride by w/kg of the respective pace partners, as opposed to broadly indicating the category ranges (e.g. 4.2 w/kg, as opposed to 4.0 - 5.0 for the A-level pace partner), there does not seem to be any meaningful change made to the pace partners and the group riding dynamics.

More importantly, I don't know if anyone else has done this, but the rides are a lot more difficult at the busier times than they are at less busy times (say ~18:00 CET / 12:00 Noon EST for the former, vs 1:00 CET / 7:00 PM EST for the latter). In the B-level ride, for the former, the group with the pace partner is easily 30 strong, and average speed through the flats is about 45 kph. In contrast, for the latter, the group with the pace partner may have ~10 riders, and average speed through the flats is about 42.5 kph.

It's just ridiculously unrealistic to have a group chugging at 45 kph, despite the fact that no one at the front is doing more than 4.0 w/kg. Those who have done Zwift TTT knows just how ridiculous that is, given that the lead puller has to do at least 5 w/kg (if not more) to go 45 kph through the flats. The worst part is that given the pack more or less moves along at ~45 kph, whenever it approaches other riders, the front of the pack actually experiences a lower CdA that allows it to speed up even further, whereas the rear of the group sees no effective decrease in CdA, thereby producing a very stretched group should the rear not pick up power drastically (as in doing at least 4.0 w/kg, if not more). Then when a gap is opened, which is almost a certainty at the busier times given the stretching, one has to chase onto a pack moving at 45 kph, as opposed to one moving at ~42.5 kph. In contrast, there are a lot fewer gaps that get opened at the less busy times

The same type of effect is observed in the C-level ride, where the group chugs along at ~41 kph on the flat, despite almost no one doing over 3.0 w/kg. It would seem that the unrealistic speeds and the pack getting stretched is almost invariably due to the large number of riders than anything else. Worst is when the pace partner sticks actually helps to split the group by sticking to the front (still doing 3.2 w/kg) while those chasing at the rear barely makes any gains while going at 4.0+ w/kg.

I'm sure Zwift knows of this, does it plan to do anything about it?

Also, while we are at it, I could really do without the inanities of the pre-generated quotes from the pace partners, some of which are just downright awful.

It boosts egos, even if unintentionally. That is good for membership. Therefore I doubt it will be fixed.
Quote Reply
Re: Does Zwift actually read the feedback it solicits for Pace Partner rides? [MRid] [ In reply to ]
Quote | Reply
MRid wrote:
The larger the pack, the lower any given rider in that packs CDA is. It's not a great method, but it's the Zwift method, so it is what it is.

In response to the question in your title, I'm not convinced Zwift reads any of the feedback it gets, about anything really.

If you are interested in a more realistic draft, I'd suggest giving RGT a try. It has it's own set of worts, but it's a much better cycling simulator (as opposed to cycling game).

While I do rant about Zwift a lot, at this point, I'd settle for the bad simulation if the Pace Partner rides occur with everyone else invisible. As is, these rides are highly influenced by others on the road (which is what leads to the stretching groups).
Titanflexr wrote:
The dynamics of the B and C pace groups are very different:

-In the C group, Coco is almost never on the front and always has a draft, so the 2.5W/kg is what someone in the middle of the pack needs to do.

-The B pace rider is sometimes drafting, and sometimes pulling (riders have figured out that you can slow the pace/rest a bit by forcing the pacer to the front). Also, since the pacer always does 3.2W/kg, the climbs you hit on the Titan's Grove section tend to be easier and the descents harder. Most people's natural output fluctuates on the rollers, the pacer's doesn't.

The A pacer is the complete opposite of C. It's almost never drafting, just dragging a few tortured souls around Watopia.


Personally, I'd love to have the pacer be unable to draft (similar to wo mode or a tt bike), but riders can draft the pacer. That would go a long way to stabilizing the group paces.

That's quite interesting. Thanks for sharing. I didn't realize the different behaviors of Cadence vs Brevet. I just thought it was absurd how in the C-group, while no one is riding above 3.0 w/kg, the group is still flying.

Also your suggestion to make the Pace Partner unable to draft might be a good idea. Or have the speed of the group be much more influenced by the Pace Partner.
burnthesheep wrote:
It boosts egos, even if unintentionally. That is good for membership. Therefore I doubt it will be fixed.

I'm not sure how ego boosting there is when one gets dropped by the pace partner driving the split. It actually gets a bit irritating when one wants to do a steady ride and instead has to put out power spikes just to stay in contact.
Quote Reply
Re: Does Zwift actually read the feedback it solicits for Pace Partner rides? [echappist] [ In reply to ]
Quote | Reply
Using the C group as a reference, I typically average 190W for a ride. During the ride, there can be times when it is as low as 160 and as high as 220; so a pretty significant variation (approx. +/-15%). It really depends on if Coco has a good/bad sticky draft, and if I'm in a good or bad one.
Zwift's drafting dynamics aren't bad for 2-3 riders, but big pack dynamics are very different than IRL.

ECMGN Therapy Silicon Valley:
Depression, Neurocognitive problems, Dementias (Testing and Evaluation), Trauma and PTSD, Traumatic Brain Injury (TBI)
Quote Reply
Re: Does Zwift actually read the feedback it solicits for Pace Partner rides? [echappist] [ In reply to ]
Quote | Reply
Wouldn't just having more pace partners solve this dilemma? Instead of 4, you could have 10 with a much smaller difference between levels, and offer some variety in terms of route etc.

Obviously the ABCD lettering might need to change as they're obviously keyed to event difficulty.
Quote Reply
Re: Does Zwift actually read the feedback it solicits for Pace Partner rides? [echappist] [ In reply to ]
Quote | Reply
i think your observation has more to do with zwift's drafting algorithm rather than pace partner per say, but in regards to whether they actually listen to feedback i can say yes for sure on pace partners. it used to be that regardless if you were in front of or behind the partner it would show your bike trying to "catch up" to pace partner (even if you were way ahead.) i know a lot of people commented on this, and now it accurately reflects your position. if you are ahead of PP, it says, " let coco catch up" and shows your bike ahead of PP. opposite for if you are off the back.
Quote Reply
Re: Does Zwift actually read the feedback it solicits for Pace Partner rides? [Titanflexr] [ In reply to ]
Quote | Reply
Titanflexr wrote:
Using the C group as a reference, I typically average 190W for a ride. During the ride, there can be times when it is as low as 160 and as high as 220; so a pretty significant variation (approx. +/-15%). It really depends on if Coco has a good/bad sticky draft, and if I'm in a good or bad one.
Zwift's drafting dynamics aren't bad for 2-3 riders, but big pack dynamics are very different than IRL.

I’m glad to read this, I was thinking I was going crazy. Every time I try to ride with a pacer, my power is all over the place. Having to coast to let the pacer catch up and then a few km later having to hammer to stay in the pack.

I’ve just decided to use the group until I warm up and then do my own thing.
Quote Reply
Re: Does Zwift actually read the feedback it solicits for Pace Partner rides? [timbasile] [ In reply to ]
Quote | Reply
timbasile wrote:
Wouldn't just having more pace partners solve this dilemma? Instead of 4, you could have 10 with a much smaller difference between levels, and offer some variety in terms of route etc.

Obviously the ABCD lettering might need to change as they're obviously keyed to event difficulty.


That would be a good start. I think even better would be to implement one or more of the following:

-Have only riders in the group be visible (which, iirc, means only riders in the group affect speed). Right now, anyone can come and go as s/he wishes. I get the point of that, but the downside is that everyone else on the road affects the drafting dynamics, which is what leads to the stretching phenomenon mentioned above. This is also one of the reasons why most group rides these days show only riders in the group.
-Cap Pace Partner speed. When there is a gap, it takes a lot bigger effort to close down a gap to a group moving at 45 kph than it is to a group moving at 42 kph.
-Remove ability of Pace Partner to draft others, as suggested above.
-Make Pace Partner sit toward the back half of the group, like the way Coco Cadence behaves.


Also, in terms of user interface, allow people to join a ride without having to do a restart. I often find myself wanting to do 30 minutes steady after doing intervals, and I either have to quit and join (takes at least two minutes) or go in direction opposite of the pace partner and hit an U-turn to join.

---------------------------------------------------

jflan wrote:
i think your observation has more to do with zwift's drafting algorithm rather than pace partner per say, but in regards to whether they actually listen to feedback i can say yes for sure on pace partners. it used to be that regardless if you were in front of or behind the partner it would show your bike trying to "catch up" to pace partner (even if you were way ahead.) i know a lot of people commented on this, and now it accurately reflects your position. if you are ahead of PP, it says, " let coco catch up" and shows your bike ahead of PP. opposite for if you are off the back.

I forgot about that part. In which case, it seems to do the superficial fixes without contemplating the more substantive issues.
Last edited by: echappist: Mar 7, 21 9:28
Quote Reply
Re: Does Zwift actually read the feedback it solicits for Pace Partner rides? [echappist] [ In reply to ]
Quote | Reply
yeah that was low hanging fruit, and made obvious sense to change. as far as drafting dynamics, that is a big ball of string to untangle, and could fundamentally change everything if they mess with it and don't get it right. they are testing new drafting AI in races, read about it here: https://zwiftinsider.com/new-pack-dynamics/
Quote Reply
Re: Does Zwift actually read the feedback it solicits for Pace Partner rides? [timbasile] [ In reply to ]
Quote | Reply
timbasile wrote:
Wouldn't just having more pace partners solve this dilemma? Instead of 4, you could have 10 with a much smaller difference between levels, and offer some variety in terms of route etc.

Obviously the ABCD lettering might need to change as they're obviously keyed to event difficulty.

One hack is to change bikes. Keeping up with the C group on an MTB (or a no draft TT bike) is harder than riding B on a tron bike. Going from an aero road bike to a gravel bike to an MTB provides a lot of intermediate levels of difficulty between groups.

IMO, having big riding groups in the pace packs is a benefit. It rains ride-ons, and it's good practice for races/group rides for getting used to Zwift's quirky pack dynamics.

ECMGN Therapy Silicon Valley:
Depression, Neurocognitive problems, Dementias (Testing and Evaluation), Trauma and PTSD, Traumatic Brain Injury (TBI)
Quote Reply
Re: Does Zwift actually read the feedback it solicits for Pace Partner rides? [Titanflexr] [ In reply to ]
Quote | Reply
Titanflexr wrote:
timbasile wrote:
Wouldn't just having more pace partners solve this dilemma? Instead of 4, you could have 10 with a much smaller difference between levels, and offer some variety in terms of route etc.

Obviously the ABCD lettering might need to change as they're obviously keyed to event difficulty.


One hack is to change bikes. Keeping up with the C group on an MTB (or a no draft TT bike) is harder than riding B on a tron bike. Going from an aero road bike to a gravel bike to an MTB provides a lot of intermediate levels of difficulty between groups.

IMO, having big riding groups in the pace packs is a benefit. It rains ride-ons, and it's good practice for races/group rides for getting used to Zwift's quirky pack dynamics.

That’s a good point. I’ll swap out my tron bike for a MTB or steel bike and ride with Coco instead. I could probably average right around 3 wkg

Speaking of ride-ons, personally don’t care too much for them, but it would be nice to auto-reciprocate or automatically give them out to riders who have ridden together with me for, say half an hour
Quote Reply
Re: Does Zwift actually read the feedback it solicits for Pace Partner rides? [echappist] [ In reply to ]
Quote | Reply
echappist, you are a hard yakka aren't you? :)

Agree on the bot commentary!

You've made valid points, but overall, the pace partners have been a huge improvement.

Sometimes I use the Coco group for an easy ride out to LAX (I have no idea what it's called in Zwift!) and organize a little breakaway to do an interval at 3.5w/kg all the way back to the banner. Then recover, let Coco catch up..and repeat. It's a solid workout for a long ride and passes the time better than doing LONG intervals on my own. And unlike group rides led by humans, no one's ego gets bent out of shape by fliers "ruining" their ride. Great fun and found like minded/similar ability riders on zwift doing this.

As for Zwift listening to feedback on the pace partners...I'm pretty sure they listen to exactly ZERO user feedback unless it's about a world championship race they want to organize.
Quote Reply
Re: Does Zwift actually read the feedback it solicits for Pace Partner rides? [echappist] [ In reply to ]
Quote | Reply
Agreed with many/all of your remarks.
For me the pacer-rides also seem to be unrealistic (esp. an hour with B even feels harder than an hour with A for some reason (guess the NP with B is relatively high due to group dynamics as well as the titans grove roller coaster??)
anyway my requests would be
1) make the bot not draft the rest of the group (indeed putting her/him on a TT bike would simply fix this)
2) allow for a longer warm-up before joining the bot, let's say upto 10 minutes with a possibility to "join now", like with an event or race
3) lower power-output of Bot on downhills (maybe make it proportional to grade*10? so -3% grade = -30%power)
4) don't make A-bot go through the jungle
Quote Reply
Re: Does Zwift actually read the feedback it solicits for Pace Partner rides? [Kempenaer] [ In reply to ]
Quote | Reply
Darren325 wrote:
echappist, you are a hard yakka aren't you? :)

Agree on the bot commentary!

You've made valid points, but overall, the pace partners have been a huge improvement.

Sometimes I use the Coco group for an easy ride out to LAX (I have no idea what it's called in Zwift!) and organize a little breakaway to do an interval at 3.5w/kg all the way back to the banner. Then recover, let Coco catch up..and repeat. It's a solid workout for a long ride and passes the time better than doing LONG intervals on my own. And unlike group rides led by humans, no one's ego gets bent out of shape by fliers "ruining" their ride. Great fun and found like minded/similar ability riders on zwift doing this.

As for Zwift listening to feedback on the pace partners...I'm pretty sure they listen to exactly ZERO user feedback unless it's about a world championship race they want to organize.

That's a good suggestion, I'll give that a try.


And also, I agree with your overall point, which is that it's still better to have the pace partners around. Back when I used to do an hour or so at upper endurance, I either need to find an actual Zwift group ride (which doesn't always happen when I'm available) or watch something to distract me. The Pace Partners are almost good enough at replicating the former.

Kempenaer wrote:
Agreed with many/all of your remarks.
For me the pacer-rides also seem to be unrealistic (esp. an hour with B even feels harder than an hour with A for some reason (guess the NP with B is relatively high due to group dynamics as well as the titans grove roller coaster??)
anyway my requests would be
1) make the bot not draft the rest of the group (indeed putting her/him on a TT bike would simply fix this)
2) allow for a longer warm-up before joining the bot, let's say upto 10 minutes with a possibility to "join now", like with an event or race
3) lower power-output of Bot on downhills (maybe make it proportional to grade*10? so -3% grade = -30%power)
4) don't make A-bot go through the jungle

Out of curiosity, have you typed in those suggestions after your rides? They all seem very sensible. E.g., it's particularly irritating getting dumped out the back on a descent, while still maintaining 3.2 w/kg or more.

Also agreed that the A-bot in the jungle is just pure meanness. I could stay with her just fine, if I were on my road bike, but I do that pace partner ride when I'm on my TT bike (where I might eke out 4.1 w/kg for 20 min), and it's a tough slog. @Titanflexer's description of the A-bot dragging poor souls around Watopia certainly applies to me; the ride feels like a task from purgatory.
Quote Reply
Re: Does Zwift actually read the feedback it solicits for Pace Partner rides? [Titanflexr] [ In reply to ]
Quote | Reply
Titanflexr wrote:
One hack is to change bikes. Keeping up with the C group on an MTB (or a no draft TT bike) is harder than riding B on a tron bike. Going from an aero road bike to a gravel bike to an MTB provides a lot of intermediate levels of difficulty between groups.

IMO, having big riding groups in the pace packs is a benefit. It rains ride-ons, and it's good practice for races/group rides for getting used to Zwift's quirky pack dynamics.

I got around to giving this a try today. First was with a steel bike and normal wheels. Still had no issue keeping up; even managed to get ahead of the group a bit (by riding at ~3.1 W/kg at 220 W). I did the old hit "A" button trick to send me to the back of the pack for a few times.

Next up was on the MTB, and it certainly was difficult. Probably did ~3.3 W/kg, and I still got dropped. I should keep this in mind for future reference, when I need to do a hard-ish ride. With the TT bike, I'm probably doing ~39 kph at 3.2 W/kg, which is a bit too slow for Coco, when there are numerous other riders around.

Didn't get around to trying the gravel bike yet, but I'm hoping that it would allow me to get a decent workout at ~3.1-3.2 W/kg when riding with Coco.
Quote Reply
Re: Does Zwift actually read the feedback it solicits for Pace Partner rides? [echappist] [ In reply to ]
Quote | Reply
IIRC it took me 2.9-3.0 W/kg to stay in the Coco group on a gravel bike. The differences between bikes in a class (ex. all TT bikes) is small, but the difference between bike classes (e.g. road vs TT vs. gravel vs. MTB) is huge.

ECMGN Therapy Silicon Valley:
Depression, Neurocognitive problems, Dementias (Testing and Evaluation), Trauma and PTSD, Traumatic Brain Injury (TBI)
Quote Reply
Re: Does Zwift actually read the feedback it solicits for Pace Partner rides? [Kempenaer] [ In reply to ]
Quote | Reply
Kempenaer wrote:
Agreed with many/all of your remarks.
For me the pacer-rides also seem to be unrealistic (esp. an hour with B even feels harder than an hour with A for some reason (guess the NP with B is relatively high due to group dynamics as well as the titans grove roller coaster??)
anyway my requests would be
1) make the bot not draft the rest of the group (indeed putting her/him on a TT bike would simply fix this)
2) allow for a longer warm-up before joining the bot, let's say upto 10 minutes with a possibility to "join now", like with an event or race
3) lower power-output of Bot on downhills (maybe make it proportional to grade*10? so -3% grade = -30%power)
4) don't make A-bot go through the jungle

Longer warm up, yes absolutely. I've also suggested they create an E group for the "old and in the way" pack, for those days when someone wants a pacer at a more leisurely pace. Also I didn't realize that the TT bike would not get a draft while riding with a pacer, so I'll try a different frame next time.

Live long and surf!
Quote Reply
Re: Does Zwift actually read the feedback it solicits for Pace Partner rides? [Giant Steps] [ In reply to ]
Quote | Reply
Giant Steps wrote:
Longer warm up, yes absolutely. I've also suggested they create an E group for the "old and in the way" pack, for those days when someone wants a pacer at a more leisurely pace. Also I didn't realize that the TT bike would not get a draft while riding with a pacer, so I'll try a different frame next time.

A loosely related problem I've noticed is really big rides, particularly C/D level, e.g. rides with several hundred. With that many people there's a large distance to the start line for the people who join the pen late. So the rider leader starts at the prescribed pace right at the front. But people way behind see the beacon up the road, and jump hard to go catch it. This creates gaps, then everyone starts complaining that it's a "hammer fest" and badmouthing the ride leader about why they're going 4 W/kg on a 2 W/kg ride, but are still dropped. I always feel bad for the rider leader in these situations.
Quote Reply
Re: Does Zwift actually read the feedback it solicits for Pace Partner rides? [trail] [ In reply to ]
Quote | Reply
trail wrote:
A loosely related problem I've noticed is really big rides, particularly C/D level, e.g. rides with several hundred. With that many people there's a large distance to the start line for the people who join the pen late. So the rider leader starts at the prescribed pace right at the front. But people way behind see the beacon up the road, and jump hard to go catch it. This creates gaps, then everyone starts complaining that it's a "hammer fest" and badmouthing the ride leader about why they're going 4 W/kg on a 2 W/kg ride, but are still dropped. I always feel bad for the rider leader in these situations.

Yeah, that's quite the no-win situation. 150 riders is about the upper limit for what is manageable for a ride leader (and even then, it has to be on conducive terrain). Anything larger will most likely produce splits.
Quote Reply