Someone had commented in this morning’s 0600 EST race that once the pack started to fall apart, it seemed that the groups forming were all producing similar W/kg. This natural selection may show that we could use a few more categories… at least 2 more.
Thanks for clarification. For sure a just rolling along pack ought to be easy as heck just sitting on, even at really high speed, if Zwift’s idea is to get closer to real life.
1 Like
Felt like draft was back to front, sitting more than a 1/3 back from front meant having to work harder than being on the front, and then getting braked while trying to hold wheels.
1 Like
Glad to hear your perspective Jamie being a lighter C-cat rider (I have a vested interest).
I didn’t race yesterday but I was “watching” your race on the “live” feature of ZP. Looking at the field, which contained quite a few well-ranked riders and (as usual for C) quite a lot of heavy/powerful riders, I was wondering if a lighter rider might struggle. Obviously not, well done.
Even as things currently stand with PD3, I’m having to ride at FTP, or above, just to hang with the front group of a C-race for any length of time, and that’s with efficient drafting etc. I’ll reserve judgement until I’ve actually tried PD4, but I fear for my future as a mediocre C-cat 
In zwift in general or PD4? I find drafting very easy on Zwift. Not saying it’s the same as real life, or course it’s not, holding the draft in both are two compeltrly different skills, but eith doubled draft on, zwift is very very easy.
Zwift has no idea whether you are at the front, middle or back of a pack. The autobraking is only applied when all the criteria mentioned before is met, so if you are not being autobraked, the only thing making it harder at the back is simply the fact that everyone is riding harder than normal. All the other variables (draft for example) are the same. Overall speed of the pack itself is slower.
[quote=“Ned Bowen [DRAFT] 
, post:155, topic:596602, username:Edward_Bowen”]
Previously, if you were in the middle of the string behind the bunch (the string being all the riders not tucked in tight near the front), you could put in a reasonably hard effort (similar to an effort that would stretch the bunch at the front), then pedal easy to make the extra speed created carry you forward to the back of the tight bunch. With PD v4.4, if you try that method, as soon as you back off pedaling the extra speed you created will be removed by autobraking the first rider you try to pass.
[/quote]
That I think is the reason why it was so hard to just hang onto the pack and probably is a major point to work on. Approaching a rider while there is enough space to move around him you should not get auto braked unless you go to <50 W or so. I think it can be tweaked with the existing parameters, but maybe needs to introduce another one.
Condition 3 (power drops below last 10 sec average): The threshold of just one (or just some) percent decreased power really low. Especially if I ride over FTP, power fluctuations of 10% happen quickly even when I think I am still pressing on and trying to move up.
Condition 1 (draft exceeds a threshold): This might be the second parameter to look at. The threshold is low enough that it can be met at the back behind a longer string of single or double file riders; you don’t have to sit right in a huge bunch apparently.
Generally I think the algorithms needs to take another variable into account: how many riders are within a bike length in front you, i.e. is there room to pass. That probably is not an existing value already and might require some more coding, but the information is there.
@DavidP I am sure you have smart guyys working on this problem, but I cannot help but propose some thoughts. These may be conditions for auto braking to discuss and try:
-
You do not want to pass:
1a. You are approaching: You are faster than the rider in front of you (or the fastest rider within one or two bike length in front you.)
AND
1b. You are backing off: Your power is below x W, e.g. < 50 W, i.e. you don’t want to pass. -
You cannot pass easily
2a. You are approaching: You are faster than the rider in front of you (or the fastest rider within one or two bike length in front you.)
AND
2b. The road is crowded: There are more than x riders within one or two bike lengths in front of you.
AND
2c. You are not pushing hard enough: You power output is below a threshold depending on your speed. It could the power you would need to maintain your speed or accelerate without draft.
I think 1b would work better than a small percentage below last 10 sec average because it owuld give riders more control. If I don’t want to pass I soft pedal below 100 W or frewheel. If I still do above threshold power, but just lower than the last 10 seconds does not mean I am relaxing.
2c is tricky, but it may work better than the +10% compared to last 10 secs. The idea here is that in order to move up and get to the front, you would have to put in enough power to overcome the wind resistance without draft. If you dont push hard enough to being able to ride to the front in the wind, you will not be able to overtake.
I did exclude the current condition of the draft exceeding a threshold as I think it interfers to much with being in a tail behind a large pack. A pack or 2 single file riders in front still apparently give very high draft and in that situation there is no reason to get autobraked. Instead I added condition 2b.
3 Likes
About the missing ZwiftPower results, that should be sorted now.
There is an old rule in ZP that if the event contains Test in the name the results are all N/A… 
2 Likes
- if you are holding steady power you should NEVER be auto-braked as you run up on slower rider(s)?
Why not? that’s exactly what happens in the current PD3. You hold a steady power, you get the draft from riders in front and overtake them like if they aren’t there.
- is auto-braking triggered differently if you have a clear path left/right of the rider immediately in front? In other words, when you are on the outer edge of the group or the group is strung out single or double-file maybe you should never be auto-braked (because you have a path to pass)
It depends. Depends on the total amount of drafting you get from the riders in front and around you (diagonally). In theory if you have a clear path you shouldn’t get enough draft. For now we are trying not to enter the realm of collision detection and keep things more simple.
- is the draft threshold a proxy for how many riders you have in front of you or does the client know this specifically?
Not sure I understand, but I think the answer is the client knows this specifically.
- how does the system determine which rider to auto-brake to? Could you be auto-braked to a rider drifting back through the group instead of those holding the group speed? or a rider being overtaken by several riders including yourself? (subject to the 3kph limitation)
It’s the rider giving you the most amount of draft, so either the one directly in front or even if a bit pushed to the side diagonally.
- is it “instant power” drops below 10 sec average power? Might it make sense to apply a trigger threshold (10%) on both sides?
Yes and yes
#1 I’ll have to think about it…
#2 I’ll have to think about it…
1 Like
At the moment if you are the ‘progressing’ rider, you go up the middle (or go straight, and as soon as there is a gap, you adopt the central line). Others riders are pushed outwards. Almost like a convection current.
@DavidP I wonder if flipping this logic would help? So the rider progressing always moves to the outside, and the slower riders to the middle? It means to progress you mostly have to lose the draft.
7 Likes
I will have to discuss that with PDv3 expert.
7 Likes
I’ll analyse some of those ideas.
The main problem with using absolute watts in rules is to account for everybody from the lighter / weaker riders to the heavier / stronger
Sure. Instead of a fixed number of watts, a percentage of FTP could be used.
I appreaciate very much the transparency of your test and your intake of comments from the users!
I would prefer zFTP since FTP can be adjusted.
1 Like
Hi David, how happy are you with the way things are progressing at the moment and do you think we’re nearly there with PDv4?
I was just wondering what the timeline is for full implementation (or is this a “how long’s a piece of string” type question?).
If you really want feedback from a much wider demographic, maybe you could introduce it exclusively for a series of races within “your” control such as the ZRacing Monthly series (lots of different riders of all abilities riding the same course). That way you could make minor tweaks based upon feedback & speed-up up the iterative testing process?
If the intent is to have PDv4.0 as the new default across all activities on Zwift, then it would surely be beneficial to test as many scenarios as possible, as soon as possible?
The “close the gap” logic is also broken in that it only pops up when the road is completely clear in front of you. If there is someone in front of you but all the way to the other side so you get zero draft benefit from them, the only feedback you get is that you have to push harder.
The lane choice logic is not working properly either, on a recent non-PD4 ride in some situations my avatar simply would not move right behind the rider I was following but rather would stay far away to the side no matter what I tried in terms of modulating my power. Plus there is also the problem of getting pushed from the inside to the outside of a blob and into the wind for reasons unknown.
In a complex system like this, changes to one part of the system can quite easily have unexpected consequences in a different one, so I wouldn’t dismiss comments about drafting becoming more difficult with a simple “drafting wasn’t changed”. (As far as I’m concerned, drafting is broken in PD3 as well…)
4 Likes
how happy are you with the way things are progressing at the moment and do you think we’re nearly there with PDv4?
I’m not sure. There are polarizing opinions about it. This is not like the previous version that was essentially a visual change. With PDv4 it will change the way races are played.
There is no fixed timeline at the moment.
If the intent is to have PDv4.0 as the new default across all activities on Zwift, then it would surely be beneficial to test as many scenarios as possible, as soon as possible?
That’s the end goal, and we have things ready to test on all scenarios if we want to. In fact I could just enable it right now globally with a push of a button 
I hope to setup for next week another round of tests events, with changes in parameters (mostly reduced threshold when backing off power) and also without double draft (maybe regular draft or a custom draft value).
2 Likes
Do it next Tuesday please https://emoji.discourse-cdn.com/twitter/rofl.png?v=12
@DavidP I think double draft is fine, and not dictating the pack speed because overall the test events from yesterday showed average power was much higher across the board for the whole pack. Then the riders in front were pushing much harder as well. It might be worth a check that the max draft conditions are set for double draft as well, and not for standard draft.
Also, I had an epiphany moment this morning. In PDv4.4, if all the conditions are met the rider is brake checked and forced behind a wheel, scrubbing all momentum. What if, instead of scrubbing that momentum, there was a speed limit set on passing the slower rider. So, not changing anything else, but the brake check adjusts speed differential to 1KMH, or even 0.5KMH. If all conditions aren’t met, then you don’t get brake checked, but if all conditions are, you don’t lose all momentum. It’s almost the same thing, but visually it might cause less uncertainty with racers.
2 Likes
Maybe there’d be some value in turning PD4 on for for a few hours in all of Makuri? Currently it’s not a guest world (so limits exposure), but you could garner feedback from the Robopartner riders who all ride there (though somehow you’d have to spread the word you’re doing this). Unlike a race, maybe some key learnings watching what happens with a group following the given PP pacing.
3 Likes