Zwift took their latest pack physics version (Pack Dynamics v4 or PD4) live on April 4. PD4 was developed to meet the following goals:
- Reduce pack speed somewhat for large packs
- Make it less punishing if you’re dropped/easier to bridge back up
- Make breakaways have a slightly better chance of sticking if the attackers work well together
- Reduce the number of bunch sprint finishes in races
- Reduce churn and rider movement in large packs
Now that it’s been live for 2 weeks, I thought it would be helpful to dig in and see if each of those goals was attained.
But first, a disclaimer: figuring out precisely how PD4 has actually affected Zwift pack dynamics is no easy task! Yes, precise tests can be done to show certain effects, but they don’t show the whole picture. There are many, many variables at play. So what you see below is a mix of conclusions from Zwift Insider tests, my own experience riding on Zwift, and feedback I’ve seen from many other Zwifters. I believe my conclusions are mostly right, but given the lack of solid analytics data, I won’t claim they’re perfect. And I would love to hear if you disagree with any of what I wrote below!
#1: Reduce Pack Speed

This is sort of the linchpin goal in the list, isn’t it? If Zwift can reduce the speed of large packs it would make it easier to bridge back if you’re dropped, it would give breakaways a chance, and it would reduce the number of bunch finishes.
There’s just one problem: it doesn’t seem to have happened.
It’s hard to find empirical evidence showing that packs are now faster with PD4 than they were with PD3. There are many variables in play, and I don’t have access to the full universe of Zwift data. But if you look up category-separated races that consistently run on the same route, you notice a trend.
Example: Team DRAFT Races
Look at the times for B and C categories (the largest categories, so most comparable across events) of Team DRAFT’s Thursday Races just prior to PD4’s release (Mar 23 and Mar 30) compared to the race using PD4 (Apr 6). PD4 times are faster for both categories, despite fields being larger in most of the pre-PD4 events.
Similar results can be seen for Team DRAFT’s Monday Races. Mar 27 and Apr 3 show significantly slower times than Apr 10 and Apr 17.
Why are pack speeds higher with PD4? It’s not because Zwift’s fundamental individual rider physics have changed – we still have the same wind resistance, Crr, CDA, etc. Tests with a solo rider at constant power produced the same results before and after PD4’s release. A single rider sitting on the front of a pack still turns in the same times as well.
The big change is the stronger draft. Double-draft is now used game-wide, so riders sitting in can now hold lower power yet maintain the same speed as before. (See our recent TTT speed tests for details.)
The stronger draft does two things that affect overall pack speeds:
- Riders get more rest in the draft. Putting it another way: the combined power of the group is now lower for the same speed. This means a race group holding the same power as it did pre-PD4 will now move faster than it did pre-PD4.
- The slingshot is stronger. When you slingshot around the front rider and into the wind, double-draft shoots you out into the wind stronger than the old draft did. This increases your speed just a bit more, which lets the pack move a bit faster. Multiply that over hundreds of “slingshot events” in a race and you’ll have a slightly faster pack.
The RoboPacer Question
Many Zwifters are saying PD4 has slightly changed their RoboPacer experience, making it harder to hang with the pack and increasing the speed of the group. This makes sense if you consider that many Zwifters have learned to try to stay ahead of the RoboPacer in the pack. This means the RoboPacer isn’t on the front setting the pace – rather, the bot is in the draft.
Increase that draft while keeping the bot’s power the same, and now the bot is moving faster than before. This means riders will need to put out 5-10% more power than before to stay on the front of the RoboPacer group, while sitting in the pack draft shouldn’t feel much different even though you’re going a bit faster.
#2: Easier to Bridge Up

Zwift wanted to make it a bit easier to bridge back to a group if you’re dropped. Has this happened with PD4? I would say yes… and no!
Yes: if you drop just a bit off the back of the group, the stronger draft gives you a bit more “cover” than before, so you can chase back on.
But if you drop too far back, the increased pack speed (see above) means you’ll be chasing an even faster pack, without any help from the draft. This means bridging from one group to the next (say, a group 20 seconds up the road) is now harder than before.
#3: Give Breakaways a Chance

Before PD4, seeing a small breakaway win in a flattish Zwift race against a pack of 30+ was a rare thing indeed. This is because the pack would churn on the front, driving its speed up, while the small breakaway wasn’t churning nearly as much.
PD4 doesn’t help solo breakaways – in fact, it hurts their chances, since pack speeds are higher, but solo speeds are not.
But PD4 does give small bunch breakaways a better chance, since double-draft lets a well-organized breakaway work together TTT-style to efficiently keep speeds high. If a small bunch can rotate in TTT formation, using the double-draft slingshot to boost speeds each time a new rider goes to the front, they stand a better chance now than they did pre-PD4.
#4: Reduce the Number of Bunch Finishes

Are we seeing fewer bunch finishes with PD4? I’m not sure. It’s another hard thing to test given the data and analytics tools I have on-hand.
But my gut (and recent Tiny Race stories) say we’re seeing a slight increase in bunch finishes, and we’ll see more as breakaways learn to work together with Zwift’s updated physics.
#5: Reduce Churn

The goal of reducing forward/backward movement of riders in the pack is, I think, the one PD4 goal in Zwift’s list which has most clearly been met. You only have to hop into a RoboPacer group to see there’s less of that churn, and a more realistic feel of needing to modulate your power to move forward/backward in the group.
But there’s a problem: we’ve still got too much churn, and it’s driving up pack speeds (see above).
The good news is, Zwift knows this, and they’re working on it…
Pack Dynamics 4.1
There’s a massive forum thread on PD4, and David P. from Zwift, who headed up the development of PD4, has been chiming in. Here are some excerpts from what he said in a post just yesterday:
I guess it was already said and explained, and I can confirm, the so called “churn” is responsible for the higher speeds of the packs, and with PD4, like the ZI tests demonstrated because of increased draft if riders keep a constant pace the speed will increase.
I’m currently exploring what Dejan previously suggested, of implementing a more aggressive type of churn prevention, by only allowing to overtake to the front (someone not drafting), if you overcome the power to maintain the same speed if not drafting. So… if you want to go to the front of the group you need to put out more power (net power, not absolute w or w/kg) than the person in front.
So I can confirm I’m working on a possible set of PD 4.1 improvements, only testable on racing events, at least initially. ETA unknown.
Summing It Up
So what does it all mean?
I’d say Pack Dynamics v4 is an improvement to Zwift, but there’s more that could be done. The work David P. is doing seems to be exactly what’s needed in order to tighten pack physics even more, so Zwift’s stated PD4 goals can be emphatically met.
Your Thoughts
Share below!
I agree with allot of this but here are a few things I have noticed. IF the pack is not trying the breakaway (solo or small) has a much greater chance because of lower churn. What I have experienced is if the pack is 30+ it isn’t much different in breakaway potential but if the pack is mid-sized (20ish) you can get a break going. This is just because the less people in the pack the less likely you will get someone willing to work. If the entire pack “sits in” the pace is lower now where previously a pack… Read more »
I will add – the Chasing series races from before and after PD4 are a good example to look at. These are mixed cat starts and prior to PD4 we almost always had B’s making it with the lead(large pack) to the finish.
Now we are almost never holding onto the front A’s if there are any hard/lumpy spots and the packs at the finish are mostly 10 or less rider clumps.
These races feel and long significantly different now with PD4 and we are often start 100+ riders in them.
It is really hard to keep the category together now. As I see it, these races are category against category, and dropping many riders on the first bump, because the group is stretched out, is a big problem. If it is only the strongest riders, riding in a small group from early on, I think it makes it less fun to attend these races for all but the strongest.
I agree 100%. I consider myself a strong B rider, but since the release of P4, I find myself struggling just to hang on. As I noted above, I think this has a lot to do with Zwift now relying on their zFTP for category enforcement. I know a lot of really strong A riders that have for some unknown reason been bumped down to cat B. As a result, the B races are absolutely brutal to say the least. Once again, Zwift has gone and screwed the pooch.
okay that explains my downgrade. bottom of the barrel B to top of the line C. sorry C’s thems the rules
if you cant hold onto a higher cat, you are not meant to be there. too many people think they can jump at the start and sit on a high cat for a free ride away from the rest of their own cat.
The point: „implementing a more aggressive type of churn prevention, by only allowing to overtake to the front (someone not drafting), if you overcome the power to maintain the same speed if not drafting.“ will hopefully help and come soon.
Another simple solution would be to have some kind of „chaos factor“ in the speed calculation whereas that factor grows linearly or exponentially with pack size and acts as some kind of speed resistance.
I have an idea: Right now if someone moves up through the pack they go through the center of the road, pushing slower riders outside. What if, like in the real world, if you want to move up you will have to go on the outside as there are already riders blocking the center (of course they could still drop by reducing power, giving space in the middle). So now the riders aren’t slingshot right through the draft, instead they are catching wind on the outside. That would turn down the churn I think. And it improves realism a bit.… Read more »
That sounds like a great idea. As someone just trying to hold a position, I often find myself pushed to the outside and then back as I taste the wind. Then I have to push back up and hope to go through the middle. That seems to be the exact type of churn this model was supposed to prevent. Making the people move up through the outside would prevent that churn as those of us happy where we are would maintain our spot in the draft. It would also reduce the need for the autobraking as the wind would do… Read more »
Already proposed – zwift can’t do it without completely re-writing all of the mechanics.
Yeah I thought so, it’s too obvious to not been suggested before so there had to be obstacles.
The effect of forcing advancement on the outside can be done without actually forcing someone to the outside. It’s what DavidP is considering now.
You’re a many multi million dollar company. You bring in how many millions a month in subs? Hire some damn coders and write it properly. Or , someone else will.
PD4 doesn’t seem to be that improved for racing from my limited experience. It’s change a little bit how sprinting is, can’t get to the front as early as before.
What I really dislike is that it’s harder to zone out and ride with groups at the same time. Long group rides and multi-tasking (watching a show, texting, etc) is much more difficult. Kind of an oversight as I think there are a lot of junk mileage kings that aren’t enjoying it as much.
I definitely noticed that getting dropped from a group seems more sudden and violent than it did before. If you fall off the back, the change from double draft to zero is quite abrupt and will create a huge gap very quickly. Combine that with the pack constantly accelerating away and catching up is all but impossible. Nothing more deflating than seeing a gap develop, upping your power to 1 wkg more than the people you were just rolling side by side with, and still watching them leave you in the dust. Double draft makes getting dropped less likely on… Read more »
I agree with you Eric, While trying to bridge a 6 second gap it just kept growing 1 vs 8. I was pushing 4.5 vs 2.6-2.8. I also found when sweeping rides that it seems the sweeper and those being pulled have to put in more effort. It also takes more than a solo sweep now to pull those who fall out of the draft and are more than 6 or 7 seconds behind. The left right washing machine where you can go from 10th to 15 in perfect position. Then you notice that you are moved far left or… Read more »
They really need to get rid of the right left motion business and say, if you’re moving up, you move to the outside, if you are staying constant, you move toward the middle, and, if you’re moving backward you are somewhere between the 2.
So if “double draft” is likely the culprit causing these higher pack speeds at this point with PD4, would the desired effect be achieved by keeping the current autobraking criteria and reducing drafting effect by a certain amount equal to or slightly greater that was in effect in PD3? Not sure of the answer, but just spitballing here…
This would be great if they made the categories smaller or more competitive. The problem is that if you’re at the bottom of a category this actually made things more likely to fall apart earlier due to things like breakaways or hills, which means less fun for that rider. I wish they’d figure out a proper matching system before disadvantaging bottom category riders.
I don’t think 1 and 2 are correct.
There is visibility, empirically, less churn.
Drafting does not make a pack go faster, the speed of a pack is determined exclusively and logically by the speed of the riders in the front.
With those two things being true, it is fair to assume that if a pack is travelling faster with pd4, it is because riders are working harder when they are on the front. PD4 encourages riders to work harder, but there is no longer free speed.
You would think so. BUT… see the results of our “churn” test 1 here https://zwiftinsider.com/road-bike-drafting-pd4/ and here https://zwiftinsider.com/tt-drafting-pd4/ If speed is only set by the rider on the front, why is the speed in our churn test so much higher with PD4 than with PD3? The individual rider isn’t faster (a solo 300W rider hasn’t changed speed, neither have single-file formations where the front rider is at a constant wattage). The only logical explanation I cane come up with is that speeds are being bumped up as riders slingshot to the front of the group… and that slingshot is stronger… Read more »
Hey @Eric Schlange ! It’s interesting to note that some are complaining that when the pack is all doing the same wattage, the PD4 packs are sitting six wide across the road and when a few strong riders are there, the packs are getting strung out. That’s exactly what should be happening, rather than the non-deformable blobs of the past. Additionally, it’s just as hard to catch back on, but easier to tail gun. Um, that’s also like real road racing. I have a lot of fast Zwift group ride experience and real road racing experience. PD4 is far from… Read more »
no idea why updates are getting pushed live across the entire game which do the exact opposite of the stated goals. clearly needed longer in the oven before being forced upon everyone. double-draft probably the culprit.. wonder how the update would have performed without it. solo breaks no chance now (as before) lul – but maybe that is for the best from a “spectator sport” point of view..? sprint finish tends to be more exciting to watch.
You realize it was tested for months with opt-in events and then in Makuri ? These things aren’t as clear-cut as you think, and the increased pack speed item was not noticed. I was doing some of the test races and following the discussions and it was simply not noticed. Sometimes everyone tries hard, and it doesn’t work out. No need to cast aspersions on the effort though – effort was expended, testing was done.
did they have double draft enabled during the tests?
Agree with pace partner point (i hate to use robopacer) it seems a lot more difficult to stay together. Either i got dropped very quickly and need to punch power to stay in a group or i just overshoot the pacer, even i am on a 0-10% trainer difficulty on recovery day
Pace partners with decent size groups are now much harder to keep pace with. So much as a sneeze and you’ll be slingshot from front right to the back and if your attention drifts even for a split second either you go deep into the red zone to never recover on the rest of the ride or just simply get spat out and done for the ride. This is supposed to be realistic?
Hi Eric
I cannot get over 1.5wkw pedal hard will not go faster no resistance until stop peddling I have new Zwift trainer
can you help.
Make sure you have the latest firmware (connect it to the Companion app on your phone). Pedal for several minutes then stop for 30 seconds to make sure it auto-calibrates.
If that doesn’t do it, make sure it’s paired as a controllable trainer in the pairing screen.
Lastly – reach out to Zwift support!
Don’t forget when looking at the speed now compared to before in for ex B is that now zwift and ZP is using Category Enforcement witch has resulted in the many riders that previously was A cat now can race in B. So more strong riders in thet cat that helps getteing the speed higher.
Agree that it seems like a lot of people who were low in their category before are at the top of category below now, and that’s making those races harder at the front.
One Question I have, that I don’t see an answer to, is how far the draft goes back in a pack. any ideas Eric? What I mean is – in a situation with just 2 riders, the rear rider gets draft if they are within 5m. Or thinking of it the other way round, as it is probably calculated for you – you get draft from a rider 5m or less ahead of you (the closer the stronger) But is your draft always calculated just by looking 5m ahead? Say you are in a group that is 20m long, and… Read more »
Nope, the draft goes much further back than 5 metres at speed, a good rule of thumb is any wheels touching the helmet of your riders is drafting distance. People seem to spend far too much time in the front of the pack with pd4, all spread out across the road pushing wind 6 riders wide.
It is super important for both a fun experience and realism that they reduce overall pack speeds and make it easier to breakaway or bridge back up! It looks like PD4 has failed this so far.
I realized that something changed since two weeks ago. I have to go at about 4 w/kg to stay with robopacer Jacques, it never happened before, it was so hard form me to ride an hour with Jacques as ever in my zwift experience. Hence changes introduced pack riding do not work well, if not at all.
I have liked the new pd4 due to the draft. Every race or tow ride there are zpower or someone doing 450w with 130bpm zooms off the front. I can now draft these and stay away from the group of “racers” who think we’re all cheating and sneak out a good result. On a side note what is this new method of cheating I see all the time where people are doing crazy watts for low bpm? Are they intercepting the signal and changing it? Their hr goes up and down with effort but it’s just 100-200w higher than realistic.
It’s not new, likely poorly calibrated or inaccurate trainers/power meters
Pack speed is most definitely NOT reduced. In fact, it is faster and much more difficult to stay with the group, as strong riders on the front stretch out the blob into a line resembling the debris trialing off a burning meteor. Personally, I think this has a lot to do with Zwift’s new category enforcement which has, for some reason, downgraded a lot of strong A riders to cat B, and these new sandbaggers are taking full advantage of it, ripping apart races. I much prefer zwiftpower’s legacy category set up.
I’m right in the middle of Class C and to me the biggest difference with this change is that before I would usually be in the front pack until I got tired when I’d drop to the second pack, and now there usually is no second pack, just a long string of mostly isolated riders behind the main pack. I think the double draft effect must have contributed to this all or nothing type of grouping.
Interesting that A’s under CE are now B’s. I’ve been racing the ZRL ladies at C and am seeing the opposite at this end of the power spectrum. Long standing D’s upgraded to C and C’s upgraded to B. This difference probably is due to the extrapolation of the new algorithm which gives you a lower zftp if you have higher 1-5 min power. I would speculate that the change to CE will make the proportion of women winning mixed races smaller.
It is really hard to have an apples to apples comparison between PD3 and PD4 pack speeds and I would take any of them with a grain of salt. First—there’s a huge placebo effect. We were told that the racing would be more dynamic, and in the few PD4 races I’ve been in so far, it has been attacking-all-the-time. Is this the new norm? Or is this just people testing out whether you really can establish a break better? If people decide you can’t establish a break better, will the blob speeds drop as a result? But more importantly: Zwift… Read more »
Also, for example, the Team DRAFT Thursday B category races you link to on 23 March and 6 April have the following interesting characteristics. It looks like in the PD3 race (23 March), only 9 riders finished with the front group (I’m cutting off the gap from between 10s and 30s, since 10s can be opened in the sprint, but probably not 30s). That front group averaged only 219 watts. In the second race, if we cut off at the first significant gap between 14s and 31s, that gives us an estimated 14 riders in the front group who averaged… Read more »
Finally, Zwift has stated on the PD4 thread on the forum that it is not double draft, it’s 1.5X draft world-wide. So stronger than it was before, but less strong than the old double draft events were. This is still, by the way, below IRL draft, since Zwift was always halving the measured draft effect of real life drafting and “double draft” was just the actual correct draft simulation. When I see complaints about PD4 I get the impression too many Zwifters don’t ride outdoors enough and have unrealistic expectations for how easy it should be to establish a break… Read more »
Double draft has always been 1.5x, actually. They just called it double draft because it sounded fancy, I think.
Some really good points here, John. The move from ZwiftPower categories to CE is something I hadn’t taken into account – it certainly reshuffled the deck on some level!
That said, I think it’s pretty clear that packs are moving faster. Maybe not a LOT faster… but even just RoboPacer groups seem to be scooting along faster than ever, and this with a lot less riders on the platform than we had 4 months ago.
I think the new pack dynamics are great. Much more natural and less chaotic in robo pacer rides. I didn’t even pay much attention to the pack speeds. All the other things seemed right to me. Also, I’m not a racer, so the racing aspects of the pack dynamics are largely lost on me. Having said that, it seems intuitively obvious to this old rocket scientist that by increasing the draft effect (double draft, I believe you said), thereby reducing energy dissipation to the air by the vast majority of riders in a large pack, it would make that pack… Read more »
Any idea if Apple TV is going to get that cool Video Capture update?
I’ve recently raced two KISS B grade races and managed to get into breaks during both events placing p1 and p2. With PD3 this would have been virtually impossible so whatever zwift has done it is working to reduce bunch sprints on flat / rolling routes. Riders need to improve racecraft accordingly and not assume a 2/3 person break will fail. Work to bridge the gap and you never know!
Thanks Eric, you always keep us updated through your hardwork. I noticed changes to the Robopacers in Makuriland over a month ago…riding with Genie became much tougher, he keeps accelerating for no apparent reason, even when he was already at the front. So to stay with him I have to keep accelerating also. It makes the training much harder and hopefully more effective, but I have to ride with Jacques much more often now! There used to usually be around 10 or so riders in the Genie group but now it’s quite often just me. Your update has made me… Read more »
It needs to be realistic, and Zwift is moving further and further away from that. With every update, it gets more unrealistic. In Cycling, as in other endurance sports, it comes down to your fitness, but on Zwift, you get disadvantaged. It starts with the unrealistic PD and the positioning in the group. In my case, my avatar always places himself on the left or right side of the pack, hardly ever in the pack itself. I get pushed aside all the time, and I lose positions like that, making me work way harder than others…the list goes on and… Read more »
I’m 1,66 meter and my weight is 66 kg (on Zwift, irl it’s 64-65). Maria was 144 watt for me, now it’s aprox 150 watt. Coco went from 168 watt at 180 watt aprox.
At least in the A’s, the higher pack speeds are definitely not from people sitting in and being fresher, it’s because the new pack dynamic randomly spits a few people off the front and then they start gaining ground despite putting out the same numbers as the front row. This causes everyone to have to surge to cover this artificial gap. Rinse and repeat for the whole race. Instead of looking at pack speeds across races, you should look at average and normalized power. What I’m seeing is a significant increase in average power in races and a reduction in… Read more »
Hi again, I’ve just been with Genie in Makuriworld for the first time since the last update and he’s now behaving himself…no jumping around or accelerating for no reason like he’s being doing for a couple of months..excellent.
“Increase that draft while keeping the bot’s power the same, and now the bot is moving faster than before.”
I may be wrong but I thought the RoboPacer was meant to be unaffected by draft? Its speed should be controlled by power and not aerodynamics? at least that was my understanding when it first came out.
False. The bots have always been able to draft, except for a brief experiment that lasted less than a month.