I’ve often wondered about this on the ramp test.
Would the ramp test spit out different numbers if it actually ramped, instead of having steps?
Not really related to the OP. 
Sorry for the misquote. It kinda sums things up though.
I suppose you could also add that an actual FTP result isn’t necessarily accurate either.
I suppose because as we know FTP is not a number but a range 
I’m stuffed if I know the correct definition of FTP. Everyone seems to have their own interpretation on the original term.
No matter which way you try to determine the number it will be dynamic.
I really hope TR come up with a way to blend the ramp with something better than 1m max power.
Garmin, Xert (on the Garmin) and intervals icu seemed to have no problem providing a decent estimate from my recent ramp tests.
I had to find my last ramp test and apparently intervals.icu estimated it from last 6:30 min. It was done the day after KM test to test the difference between tests. TR last min FTP was 10w lower than my real FTP and intervals.icu estimation was 1w higher so spot on. So at this particular moment the multiplier in TR should be 0.77 for me. But the ramp test is basically random numbers generator in my eyes
A lot of people are saying that Xert, icu etc will underestimate etc because you’re following plans which don’t push you to that level.
Surely you are all training for some kind of event, or do bunch rides which put you on the limit to give you some data ?
Of course power based estimates need good data to give a good estimate.
For anyone doing structured training + a decent amount of racing, xert intervals.icu should be pretty close. Mine is normally not far off a test.
Racing is the key part in this equation. So you are doing some efforts close to maximal. It is not important where they come from. The only point of training is to get better. To check if you become better - the maximal efforts are needed and then there is a bunch of software that estimates your FTP pretty close to actual value.
I think that’s the problem for a lot of people. They aren’t racing, or they don’t ride hard outside very often.
Maybe I’m misremembering but I feel like Nate mentioned wanting to do away with ramp tests in the future? I would think TR has enough data that they could pretty accurately predict what the next ramp test based on current performance.
A number dont race…there was a thread on this some time ago.
Depending on how you ride and what efforts you are doing you may not actually push your limits or your power to trigger even a bump in FTP based on the modeling done by say Xert or Intervals.icu.
My group rides are not hard rides. So I can do a couple of SSB rides and a group ride in a week and rarely go above my FTP. Rides may be long in duration but not result in a sustainable effort above FTP that would necessarily model a bump in FTP.
My own riding this year had such a period. Intervals.icu had me declining in FTP. Lots of long rides. Good effort but steady and not over FTP. You could track the changes in my heart rate and power and see the improvement. It wasnt until I did an FTP test that intervals.icu showed the increase which was above previous FTP. The models are reactive to how hard we ride. They do not project what our FTP will be based on are riding.
I’m on the same page there. It’s not really a problem as such. Unless you’re expecting an accurate number from one of the many estimates available.
I can imagine the peeps doing traditional style HV base plans fall into this category. Even SSB1 struggles to get good results with Xert and the like.
You can change your “decay rate” in xert to “No Decay” if you are planning on a long period such as early season base with no max efforts. This will still provide you with increases in TP and LTP as your training load progresses but with prevent the big decreases that you might see with no max efforts.
From xert’s manual:
When you change the Signature Decay Method to No Decay – Training Load Matched , the system does a number of calculations to help you better manage your signature during times where you haven’t done or don’t plan to do any breakthrough efforts. (In control systems theory, we would say that the system is going through a period of dead reckoning ).
Under the covers, Xert establishes the relationship between your training loads and your fitness signature parameters. When you have robust historical data, without large gaps without power data and without power meter errors / changes, Xert matches your historical breakthroughs with individual training loads.
And in my experience it really works.
Nate has also mentioned and hinted at adaptive plans and big data / machine learning / AI.
One way to close the loop on the result of a plan’s effectiveness is to use the ramp test’s estimate of 5-min vo2max / aerobic capacity (its the last 1-min power from the ramp test).
The current plans don’t include max efforts, except for the ramp test, so if you are only doing inside workouts then “current performance” can only be estimated by ramp test.
I would guess that FTP tests will soon be things of the past. This problem seems very amenable to machine learning/statistical modeling, and all of the major players have enough data to train models. Intervals.icu is pretty close already. I’d wager that before long you won’t need an all-out effort to get an estimate that’s within 5% of true FTP.
If any of the big players were to open source anonymized data, there’d likely be a model available within a few weeks. Meantime, we have to wait for TR to hire an ML engineer 
But without non-workout rides, Xert will constantly decrease your FTP. That is, if you studiously followed a TR sweet spot plan, then Xert would say your FTP is going down.
There is no way to estimate FTP without efforts that essentially “express” your new fitness. Which is essentially what the ramp test is
I had a very different experience with Xert: without constant break-through / really hard rides, if I just trained, Xert would constantly decrease my FTP. This is why I dropped Xert (for the 2nd time): for me, I couldn’t use it to “eliminate” FTP tests / use its “machine learning” to inform my training
I think any model is going to need efforts to calculate off. Even if TR got rid of the ramp test, the system would still need to prompt the rider for a 1, 3, 5, 10, 20 minute or whatever max efforts every now and then so it could calculate FTP.
I have considered looking at HR along with power to try figure out what the max version of a given sub-maximal effort might look like. That could work but HR varies a lot so would give quite a wide range of results. So I decided just to stick with max efforts for now.
I can get a pretty good idea myself of where my FTP is just by watching how my HR rises on a 14 min climb near my house, without having to go full gas.
Maybe, but I’m guessing that having features like “HR as a percentage of max”, RPE, HR minus resting HR (perhaps divided by wattage), diff in HR divided by diff in watts, etc. would get us most of the way there. Even max efforts aren’t guaranteed to be max efforts (people have bad days, some people are able to push themselves closer to their limit, training fatigue plays a role, etc.), so it’s possible that a model might even be better than an FTP test for some riders on some days.
You obviously didn’t select “no decay” for your fitness signature