UPDATE (April 5, 2023): After lots of tweaking and testing, Zwift has taken Pack Dynamics 4 live game-wide. Read more about it here >
On Monday, October 24, Zwift held its first-ever public ride using their experimental “Pack Dynamics v4”. The ride was James Bailey’s weekly “Ride with James” and its description explained the overall goals of PD4:
The Goal:
- This new version of PD has the goal of creating a more realistic racing experience (compared to IRL racing). The main changes introduced focus on reducing the forward and backwards movement of cyclists in the pack while also reducing the overall speed of larger groups.
- We encourage racers to try a more “aggressive” approach and try breakaway opportunities to create more dynamic racing scenarios.
What you can expect:
- Less speed in groups of cyclists where there isn’t a concerted effort to “pull” the group by taking pulls at the front.
- Double draft is activated by default therefore riding in a group should feel a bit easier than with the standard draft model.
It’s safe to say that the vast majority of Zwift racers, and probably the vast majority of Zwifters, are supportive of Zwift investing more time into improved pack dynamics. While the current pack dynamics (v3, rolled out in October 2021) are a great improvement over the previous version, with riders moving more like a “school of fishes” as Wes Salmon used to say, they aren’t perfect. Overall group speeds are still too high thanks to the endless slingshot-n-churn on the front, making breakaways nearly impossible. And sitting in the group draft is harder than you would find IRL.
I participated in that first ride on Monday, and recorded it:
But that first ride wasn’t the best test of PD4. It had been misconfigured for the wrong route, which meant much of the ride was spent climbing or descending a dirt KOM. And the overall ride was only 16km long, which simply didn’t leave much time to get the “feel” of PD4.
So I volunteered to use PD4 in the Zwift Insider Tiny Races happening the following Saturday. Testing PD4 at “race pace” with a sizeable pack of riders would give us a much better picture of how it behaved in comparison to PD3.
I recorded the full set of four Tiny Races, including me on the suffercam:
Experimental Feature
I want to be sure to note here at PD4 is very much an experimental feature at this point. It’s not ready for prime time, and Zwift knows this. That’s why it’s only being used on a very limited set of events, and participants are being asked to share their impressions in Zwift’s forum once they’ve completed their ride.
See the Pack Dynamics v4 thread >
This is exactly how it should work, of course. Zwift can develop new Pack Dynamics and test them in-house to some degree, but the best feedback and testing will come from a much larger body of experienced riders found in the Zwift community.
My Notes on Pack Dynamics v4
Objectively testing something like Zwift Pack Dynamics can be challenging, because you’re really comparing what you’re experiencing to what you remember from past experiences. Does PD4 feel like PD3? How does it compare to outdoor riding?
Everyone’s impressions will vary slightly. But there are also some definite themes I’m seeing in feedback from other riders which matches the impressions I had. Here are my notes on my PD4 experiences thus far, written as objectively as possible.
Draft Feel
Simply put, if you’re not near the front of the front group, the draft with PD4 feels unrealistically weak. While Zwift says double draft is built into PD4 and “riding in a group should feel a bit easier than with the standard draft model”, that wasn’t my experience.
While Zwift Racers know they can tailgun at the back and still enjoy a strong draft with PD3, PD4 gives you no such luxury. In the Tiny Race 1 I stayed in the front pack, and to be honest things didn’t look or feel much different from PD3 up there.
But in races 2, 3, and 4 I found myself toward the back of the front pack. And that’s where things got ugly. I was dropped in the Jungle in race 2 (more on that below), dropped in Central Park in race 3, and dropped along the Thames in race 4. It was painful!
It could be argued that the front pack just contained very strong riders this week, and I didn’t have what it took to hang in there. But after racing Tiny Races almost weekly for 14 weeks, and looking at pack speeds and power numbers, I’d say this wasn’t a case of strong riders stretching out the group. The pack speeds are still high, and the draft doesn’t feel strong enough, so you end up (as a B racer) feeling like you’re riding a TT if you’re trying to get back into the middle of the large front group.
Reviewing my race video, it’s possible that this isn’t a draft problem as much as a speed problem with PD4. There are times in my race where I’m just off the back of the front group, pushing high watts, but my speed is lower than it would typically be on Zwift with PD3. So perhaps PD4 is slowing me, as part of its attempt to keep pack speeds slower, and that is what causes me to get dropped from the front group.
Speed Variation
While PD4’s draft feel doesn’t feel right to me, the biggest problem I’ve seen with PD4 is something completely different. I’ll call it “Speed Variation” for lack of a better term, but basically what I’m seeing is rider speeds becoming unrealistically fast. It seems to happen on descents for me, and the best example I can show is my descent through the Jungle in Tiny Race #2. Here’s the video cued up to a good spot:
You’ll notice the video starts with me in 105th place. I’ve been dropped from the front, which is really a snaky line of riders. As the descent gets a bit steeper (going from -1-2% to -2-3%) my speed inexplicably picks up from ~39kph to ~53kph. I’m not pushing hard, either! It felt like I had a steamroller powerup, or perhaps an anvil. Without even pushing I magically moved from 105th to 45th by the time the Jungle dirt turned to Alpe pavement.
105th to 60th is a huge move, too, considering the pack was very strung out and I was catching riders one at a time. Also worth mentioning: I was on the Tron bike, while many riders were on faster gravel bikes!
I saw it with other riders as well. Here’s me getting dropped in New York. Watch the rider in the Jensie kit (“T Ward”) pass me, then watch the “A Vella” rider in the default grey kit pass as well. Both aren’t doing any more power than I am, but they fly past (and A Vuella just keeps going!)
I noticed this on Monday’s initial test ride, too. On my descent from the Temple KOM, I picked up momentum and began flying past other riders, despite putting out pretty low power in comparison to those around me:
Forward/Backward Movement
On a positive note, comparing videos of PD4 Tiny Races with PD3, I would say there is less forward/backward movement with PD4. That’s a good thing, since Zwift’s notes for PD4 say, “The main changes introduced focus on reducing the forward and backwards movement of cyclists in the pack…”
Here are two example videos, cued up to an ideal comparison spot:
Pack Speeds
Zwift’s notes for PD4 also include “reducing the overall speed of larger groups.”
It’s hard to compare group speeds with much precision outside of a controlled test, but just looking at Strava times, Tiny Races in weeks past, I would say the speed of the front pack isn’t appreciably slower. I compared the first 2 km of a few different events held on similar routes, and their times were nearly identical, with PD4 races actually slightly faster in some cases.
High pack speeds are further evidenced by the lack of any breakaways in our Tiny Races, although admittedly these aren’t the best environments for testing “breakawayability” since the pace will be quite high due to short race durations. I’d like to see PD4 used in an hourlong flat/rolling race, to see if breakaways are possible.
If Zwift is looking to reduce the speed of large groups (and I agree they should be looking at reduce pack speeds by 5-10% if they’re going for IRL realism), I don’t think PD4 is there yet.
Pack Shape
One final point (and I’ll admit this feels a bit nit-picky since PD4 isn’t trying to address this issue): I would love to see more realistic pack shapes at race speeds.
Then the pack is flying along (like you’d see in a 5-minute race), the front of the pack should be very narrow – just one or two riders – and it should widen as it goes back. A super-wide pack with, say, 15 riders all in the wind is what you see during neutral starts, or when everyone is taking it easy in a race.
I have a theory which I would love to test: that if riders could feel the draft in Zwift, it would encourage better pack positioning (more drafting, less in the wind). Right now if you’re in the wind, your trainer resistance is the same as when you’re in the draft. But if you felt yourself hitting the wind by getting more resistance, that might help you subconsciously ease up and get back in the draft.
More on this topic: “Opinion: Let Me Feel the Draft On Zwift” >
Conclusions
Zwift’s stated goals for PD4 are laudable. They show that Zwift knows the shortcomings of their current pack dynamics, and they want to make them more realistic.
That said, PD4 isn’t fully baked quite yet. Hopefully the feedback above, and the feedback Zwift is receiving from hundreds of Zwifters on their forum thread, will help Zwift’s developers dial PD4 in and thus improve the pack riding experience for everyone.
(Oh, and for you Tiny Racers: due to buggy speed inconsistencies mentioned above, we won’t be using PD4 in its current state in future Tiny Races. But once Zwift updates PD4, it’ll be time for another Tiny Race experiment!)
Your Thoughts
Share below!
Q- when you’re say 2nd rider in a pack what’s the draft? Are you still is DD mode or is it like irl and you’re doing more work than someone sitting 8-9th? I’m curious for both PD3 and PD4?
With PD3, 2nd rider does more work than, say, 4th rider. See https://zwiftinsider.com/speed-tests-ttt-pd3/
But it’s not as strong of a draft as you’d get IRL, where in a larger pack you’re saving well over 30%.
Posted my PD4 thoughts/observations from the Tiny Race Series (V2 Cat C) in the PD4 thread. I shared my notes with a couple of my Teammates that raced earlier in V1 and they both agreed.
Just this little about the speed; No wonder why T Watt flew by you, he did 4.4wkg for a while and then 7.3wkg for a short while before cruising by you at around 3-3.5wkg. True zwiftracer style
What were the rider weight differences on the descents? Could that be part of the speed differentials? 65kg rider v 75kg?
This does affect speed, yes. But not as dramatically as what I saw in the Jungle and Temple KOM descents!
Be curious to watch this with Sauce for Zwift compared with a PD3 race to see the numbers
I actually ran Sauce for some of my Tiny Races with PD4, to look at the draft number. It didn’t seem much different from PD3, but then again, Zwift says that’s not actually a draft number… so who knows what we’re looking at! 🙂
Sauce will still be using the draft numbers from PD3 I think, so it will just tell you what draft you would have got in that position if PD3 was in operation.
I suspect the draft number is some calculation around what percentage of your speed is coming from your watts vs some other force – draft or downhill.
Thanks for your thoughts, Eric. They match up with my experiences and that of some team mates and rivals that raced Tiny Race this weekend. Your note that draft reduced further back in the pack definitely lined up with my experience. The parts of the race where I sat up front felt normal and those where I was at the back felt like there was almost no draft at all. I didn’t make the connection at the time, but since you put it that way, I realize that’s exactly what it was. I had team mates get dropped doing more… Read more »
While I havent’ raced it yet, a number of my teammates have and shared their experiences in depth on WhatsApp. Based on what I’ve read and seen through the videos, PD4 could be very real-life like when it comes to differences between Pro/Cat1 crits and say Masters Cat 1-2-3-4 races. I can race a Masters race IRL and hang on the back or even sit up in the top 10 and finish a 45-min race with an average pace of 27-28mph. It’s hard, but not unbelievably hard. If I want a rest, I go to the back of the pack… Read more »
Hey Eric, My experience was similar to yours, but without the gaining 60 places. lol Life was great at the front of the pack, but horrid when I lost focus and dropped back a bit. Having said that, I love the fact that you took a chance to test PD4 out on such a popular race series. Despite the race experience being less fun, the trade-off of getting a lot of data on PD4 quickly was worth it for me. If you’re comfortable taking the slings and arrows for the occasional sub-optimal experience. I’m willing to suffer whenever the next… Read more »
James from Zwift has mentioned auto-braking is being used in PD4 and I suspect that is the problem with the speeds. As Eric sped up down the jungle it’s probably because he was no longer being auto-braked while the others around him were. Personally, I think if Zwift has to artificially brake riders to fix a problem with pack draft then they’re using the wrong tool for job. ie the pack draft dynamics has some fundamental problems (hint: you shouldn’t be able to sprint your way through the middle of a pack enjoying a free draft, you should be forced… Read more »
I picked a gravel bike (Grail + Enve) for race 2 (zone 2 cat C). I weigh 67kg and had a horrible time during the first 2km downhill. Kept slipping backward and seemingly nothing I could do about it. I was about to quit, but decided to give it one more go. Suddenly it flattened out and all changed. Between 2km and 3km I went from 98th right to the front. Then got in a breakaway, which had been my objective all along and managed to get 7th. At least 2 others on the PD4 thread at Zwift forums had… Read more »
My experience with the PD v4 in the Tiny Races was a bad one and I agree with most of Eric’s and other riders’ thoughts. In the middle or at the back of the pack it felt like having no draft on a TT-Bike … but it also often felt like having a “negative draft effect”, so an active disadvantage like Star-Zwifter Joakim Lisson already mentioned in the forum. I got dropped several times. Then I had to accelerate hard to rejoin the pack with the hope of some recovery-seconds with lower Watts. But reaching the end of the group… Read more »
I raced the first three races in zone 2 with you Eric. My experience was also very similar to yours. On the forum thread I commented that it seemed as though individual riders were stuck in arbitrary bubbles of either super draft or super anvil. But arbitrary is not fun or predictable. On your video of race two at the .5-.6 km mark you will see me get dragged from the front to the back in seconds, despite going downhill at 3.9 w/km (TFC jersey). Almost like getting caught in a riptide. Race three was better until about the halfway… Read more »
awesome that zwift is looking for community feed back here
from watching your vods (yea i watched both in full), i can def see there is work to be done. i just hope they work with the community moving forward!
awesome coverage eric
Apart from the hard work trying to stay on the back of a group, I experienced 1 part of random speed boost, and thankfully caught on Eric’s recording. @Marco mins, I’m just behind Eric, and both me and him put in a bit more power to get on the group in front. I ease off as I pass him, and I cruise through them all. I was doing around 1w/kg less than the group when I hit the back, and went through them all, and off the front, and it was a long group. While Eric was still pushing much… Read more »
That’s meant to say @Marco mins, don’t know where Marco came from
Just incase something weird is happening, it’s fifty five
“I have a theory which I would love to test: that if riders could feel the draft in Zwift, it would encourage better pack positioning (more drafting, less in the wind).” I feel like this is the main issue. If you feel the differents in draft, the riders will automaticly put in different watts they are doing now. Then you can position yourself way better in a group. It would be harder to ride on a front of a group and people will try to decrease in power to not be on front row in pack. It would directly mean the pack… Read more »
Decending in Zwift sucks.
Heavier rider careen downhill..zooming around corners at speeds they would never atain irl…because they would crash….. no brakes in zwift eh… nothing to see here.
Meanwhile lighter riders pushing big watts are going backwards fast….compared to the pack.
This needs attention…it absolutely the least realistic part of zwift…
How to solve…maybe a Max speed on corners?
Good to see Zwift looking at improving the draft/pack dynamics further, one thing I would suggest if you speak to them about it @Eric Schlange would be for them to turn on the visible “Draft Heat Map” (which they sometimes display during ZRL etc.) for their DP4 Test Events to help riders visually see the draft in certain positions, then potentially do back-to-back test events on the same course with one using DP3 and one using DP4 both with the draft heat map visible, and preferably most/all of the same riders. That might give a better visual understanding of how… Read more »
Representing being “in the wind” by increasing trainer resistance in order to induce a DECREASE in effort to move back into the draft could be confused with “going slightly uphill,” a condition that should cause riders to INCREASE effort to keep pace on the climb.
It definitely feels very beta/early days for this. I found myself “crabbing” sideways in the pack too – only a visual artifact but pretty distracting. It would be interesting if they actually released the physics/model to allow people to review and feedback. Rather than gradually refining and iterating a physics model of drafting/pack dynamics each time, it certainly feels from the outside as if Zwift they’re starting with a fresh sheet of paper each time they re-work pack dynamics, rather than attempting to model real-world physics with increasing accuracy and ironing out errors. I can’t help but feel that they’re… Read more »
Hi @Eric Schlange … would you have info on PD4 changes; change log perhaps? It appears it is not at v4.4 News on final release date? thanks.
Road a ride with PD4 this morning. It basically feels like If you aren’t in the very front of the pack the draft really doesn’t help at all. Riding mid pack if a small gap opens up in front of you two things happen 1) you lose the draft and 2) you take the full brunt of the wind. Also, they have turned on auto braking (watts turn red). Brakes can be activated at any time….not just on downhill curves. This includes being in the draft on a straight section of road. So as of today it doesn’t work well… Read more »
I am a little late to the game, but I sort of like 4.4. I will go as far as to say that it has been my favorite draft algorithm yet. For context, my race was crit city, B cat, ~30 riders. For the first time ever in my experience in crit city (B Cat), there were points in the race (other than the start of course) in which there were “splits” in the group as opposed to just “attrition” losses. It might have been me, but it felt like a fast group went faster and a slow group went… Read more »