I suppose the truth is that I am struggling towards the end of “very hard” VO2 workouts. Cadence is dropping and power is less than target. Without realizing it, I am taking advantage of delayed trainer response to complete the interval before spiral of death would kick in. Trainer smoothing hides all of that and gives a false impression of perfect compliance.
"The biggest lies are the ones we tell ourselves …
… unless you include smart trainers with power smoothing enabled. Those are the real whoppers."
My typical go to for overunders at least, is usually my warmup recovery or recovery interval is at >105% of target at least… so as the countdown to start , i ramp the % closer to 105-103% for the start of the under interval (good little bit of over to start) , but then as the interval progresses I ramp it down and it definitely doesn’t have a reduced start quite as much. I also shift one gear up, usually results in a surge not a slow rise. Plus power through with high cadence for a few seconds after interval. Seems to help account for the lag if it happens.
Broadly speaking, I always find it odd that ERG Smoothing is telling lies while many riders focus on single digit levels of accuracy for power & other metrics. For anyone that desires that precision, flipping off smoothing seems essential.
And besides that, I can’t count the number of times I’ve seen and replied to users who upgrade from a Wahoo trainer to get a real power meter and instantly assume the power meter is busted because they’ve used ERG mode exclusively with the Wahoo default of smoothing on. They’ve never seen real power to understand that jagged fluctuation is the norm, not the exception 
Those are reasons I despise the setting overall, and particularly that it’s the default for Wahoo trainers / smart bikes.
I doubt that the trainer smoothing is actually a smoothing (time averaging) algorithm. It seems to simply lock into the target Watts unless the actual Watts is outside some significantly larger window. It makes a mockery of the claimed 1-2% trainer power accuracy.
Totally. When I vary cadence in ERG mode, the trainer has to compensate by varying resistance to maintain the set power, and that will of necessity have some lag to it. I also use cadence when doing workouts in resistance mode or outdoors. Rather than chase the power I’m supposed to hit, I look at where I am and if I’m supposed to be at higher power, I increase and hold my cadence a little higher, and vice versa. It’s much easier to try to hold a constant cadence than a constant power.
So this is actually the opposite of what I want to do. I want to be intentional about the cadence I’m holding. Like on over unders I like doing the unders at about 90~95rpm and the overs at 95~120rpm (depending on the target power and duration). I know for me personally doing this allows me to feel like I’m “resting” on the subsequent unders (or at least fool myself into thinking I’m resting 
). So ideally I’d spin at the RPM I want and erg mode pulls me to the power I should be at.
It feels more like when I’m actually riding or racing outside. When I’m riding at a high power before a surge or sprint I don’t want to be spinning at my sprinting cadence before the actual jump. Or at least that’s how I think about it.
Actually I think we agree, but I was making a simplified point about just maintaining a given power. But yes, there are times I want to also change cadence, and over/unders can be one of those. That’ll be in ERG mode, or when riding in resistance mode or outdoors. Outdoors, of course, that’s what I use gears and shifting for. I’ll sometimes also use shifting in ERG mode to make big change cadence. That helps minimize the change in flywheel speed that the trainer has to deal with. And, using a medium low gear (like 34 ring and 17 cog) reduces flywheel speed so changes in flywheel speed can happen quicker.
Yes and we’re all glad you’ve lifted the veil, but don’t think this means you’ve been doing a bad job to date. Your muscles don’t know if you had a jagged line or a flatter one, and there is no such thing as perfect.
I make an effort to avoid big peaks and troughs, and don’t like it when the average is a few watts under target. Beyond that is the minutiae imho.
Agree with the sentiment … but seeing the real power data shows that my average power for each VO2 interval is about 7% below target. This is coming from averaging the peaks, troughs and a general struggle towards the end of each 50 second effort. A lesser contribution is from the ramp up/ramp down. This is more than a few Watts. TR was right to conclude that I struggled.
My previous mindset was that if I complete a workout in erg mode without breaks, back-pedalling or dropping the percentage then I’ve 100% hit the workout objectives (because erg held me to them). Now that I’ve actually turned off the trainer smoothing, it’s clear how much I’ve been kidding myself (for >= VO2 workouts anyway). 
I’ve definitely struggled and failed toward the end of very hard intervals, but it hasn’t been by the spiral of death in a very long time. What I do when I’m struggling is hold and slightly increase cadence so that the resistance decreases rather than reduce cadence and have resistance increase. When I do fail, it’s due to running out of steam and not to running out of torque. This is how I approach ramp tests too. Each time the power steps up, I hold or slightly increase cadence. When I do blow up, it’s at ~95-100 rpm.
That’s exactly right and I aspire to that. But when I’m deep into 50 second intervals at 125% of FTP and starting to struggle, it’s hard to execute. You have to push even harder meaning more torque and more power to shift the cadence up, then hold it until the trainer backs off the resistance (which takes so long relative to a 50 second interval…) to give you the relief of faster cadence and lower torque. More likely I’ll end up in trainer oscillation flailing around the target and then crack.
Or I can turn on trainer smoothing and go back to blissful ignorance, probably like 90%+ of the user base.
It does make me wonder how often this fundamental source of inaccuracy from indoor trainers ends up misleading the AI algorithms.
Or you could turn erg mode off 
I am confused by your comments in regard to Wahoo’s Power Smoothing. My understanding is that whether Power Smoothing is off or on, erg mode works exactly the same. In other words, there is always real fluctuation in erg mode. Power smooth just changes the display to a 3 second average rather than instantaneous reporting, making the display look smoother. But there is otherwise no difference. Your comments make me believe that you are sensing that your erg mode is actually functioning differently when Power Smoothing is turned off. https://support.wahoofitness.com/hc/en-us/articles/4402734188050-What-does-the-Power-Smoothing-function-do
I agree with this. Erg mode works the same. It uses a feedback algorithm to modulate training resistance to try to hold you to target power. The feedback loop is delayed due to delays in measuring cadence and torque (hence power), and time to change the resistance.
I don’t believe that Wahoo trainer smoothing is a 3 second average - it does not behave like that in my judgment. It’s a much more aggressive heuristic that is locking the reported power to the target power as long as you are within some percentage of the target power. It is fine under steady state conditions, but hides the peaks and troughs, and can mask workout compliance issues.
I don’t mind so much that the displayed power is inaccurate, but I really object to the reported power saved to the FIT file being inaccurate. This is supposed to be the accurate record of the workout to 1-2%. When head units apply power smoothing, they apply it to the value on the screen, they don’t change the power reported into the FIT file (or equivalent). The problem they have in a TrainerRoad or Zwift context is that there’s only one power number in the protocol.
I know why they do it - to reduce support calls because without this crutch, many folks would think their trainer was broken.
Your link is to power smoothing in the Wahoo Systm app, equivalent to a head unit. It is not the trainer power smoothing that we are talking about.
Hope this makes sense, thanks.
Here’s a better link, though I think they are intentionally downplaying the situation - https://support.wahoofitness.com/hc/en-us/articles/204281764-Trainer-control-modes-for-KICKR-CORE-SNAP-or-BIKE-in-the-Wahoo-app
ERG Mode Power Smoothing: When the trainer is controlled in an ERG mode (Target Power or similar), this setting smooths reported power values to eliminate small inconsistencies in pedal strokes, making power graphs appear less erratic.
Enabling is a personal preference. It will not impact overall trainer functionality, and not required.
Compare these graphs for similar VO2 max intervals:
Wahoo trainer smoothing enabled:
Wahoo trainer smoothing disabled:
The first graph is a work of fiction. It exactly matches the TrainerRoad workout. Erg mode behaves the same in both cases. My “performance” is similar in both cases. I am struggling in both cases. The giveaway is the cadence (white line) starting high and degrading in every single interval because I am struggling to hit the target.
Beautifully put 
Well you learn something new every day. I didn’t realize it actually falsified the power file. That is insidious. Glad I turned this off on day 1.
Here is an article that is very explicit on how it works. https://zwiftinsider.com/wahoo-erg-mode-power-smoothing/
“Wahoo’s Erg Mode Power Smoothing feature is toggled on and off on the trainer itself, so it actually changes the power numbers sent to Zwift (or any other device). That’s right: it’s not just changing the way your numbers appear. It’s actually changing the way your trainer reports power.”