the first 30 seconds of an interval is not accounted by XP/NP. XP & NP needs the preceding 30 seconds in order to do the (exponential/power) averaging
np is supposed to represent like training equivalent to better capture higher-power bursts and the non-linear stress they cause relative to steady riding.
so... I guess I'm wrong. No idea how you would ride to get 380AP and 300NP..
oh, there you go @echappist got it.
anyway, from what I've read NP is really only valuable for time periods longer than 10 or 12 minutes.
twitter.com/ygduf
strava.com/athletes/ygduf
Normalized power uses 30 second averages and takes them to the 4th power to figure out how hard a ride is to average out the easy parts, once you average based on the 4th power it brings it back and gives you a new number. Basically it will inflate your power numbers artificially to estimate the difficulty of a variable ride. For anything less than 10-20 minutes NP just tells you how well paced your effort and is not as good of a metric to score your intervals. Once you hit 20 minutes it tells a pretty good story about how much power you could put out in TT mode over the same distance vs just riding hard and easy and mixing it in.
If you have a lot of variability in NP and average power during a race it just means it was a hard race and the NP is a good number to use along with the difference to show how hard it was. If it is during an interval, especially one that is on the short side, it means that your effort was not paced well and you should try and even things out a bit more. I use the NP number that I get back from an interval and shoot to hit that for my next average number and use it for pacing. Under ideal situations you should be able to hold your NP number for the duration, once you get NP = average you are pretty much as efficient as you are going to get for that effort.
I have found that by targeting my NP for the next interval I get a much better workout in and it has made my intervals much more efficient. So use np, but I score my short intervals based on average power and my longer efforts (the whole ride) and races based on NP and the difference between NP and average.
oh ... and if x power was NP in your chart then I will assume that your software does not handle the shorter intervals very well and the missing 30 seconds really ruins the numbers and makes them near useless. My software will use some type of estimation to still create a valid NP number for shorter intervals hence allowing me to use it for pacing those as well. However the additional estimation of the first 30 seconds really makes it a less accurate number with the shorter distances, but still gives me a good pacing indicator. You don't seem to be getting that from your software.
There's a lot of variability there. I can do 6x5x1s on this course (hilly and always windy) at exactly the same average power for each work interval.
kindablue: you should consider recovering at 70% FTP instead of JRA, at least for the first two minutes of each recovery interval.
I think its actually just from the fact that the powertap stops my garmin later than the quarq. As soon as you come to a stop on a Quarq its auto pause. With the powertap, it takes a few more seconds as its still reading some speed. It's hard to explain but thats why the powertap has an extra 40 seconds of moving time than the Quarq.
-Cat-3-o-meter: TBD :/
makes sense
"Insanity: doing the same thing over and over again and expecting different results." Einstein
haha nah I just wanted to see how different they were because the powertap is going to be my main road powermeter. I'm moving the quarq over to the track bike. I hope to have the track bike in my hands Friday or the weekend. The wheels are being built at the shop but everything else is already together(bars, stem, etc etc).
In my living room I've got the Bergstrom Adapter plate sitting patiently, waiting to be attached to my Quarq
-Cat-3-o-meter: TBD :/
Hey I got a question. I'm trying to use a spreadsheet to calculate ATL, CTL & TSB.
When I look online, I find the same general formula for ATL & CTL, but I can't find a consistent time constant and I would rather not take the time going through derivations. I find these three constants popping up:
(1) ATL constant = 1/7, CTL constant = 1/42
(2) ATL constant = 2/(7+1), CTL constant = 2/(42+1)
(3) ATL constant = exp(-1/7), CTL constant = exp(-1/42)
I don't know how to compute exponentially weighted averages. Number 3 seems way off - seems too close to one.
Number one seems best. Starting out with extreme negative, balancing out to +/- 10 most of the time, and being negative most of march with a low of -32 last Saturday.
Edit:
Also wondering what's a good range of TSB to hit during recovery.
Using (1), last three blocks recoveries I hit barely positive (then hovered around 0 for the whole block), +25 recovery (decreased steadily to ~0 through block), and +10 recovery (most of last block was negative).
Last edited by aaronmcd; 04-03-14 at 07:34 PM.
CTL is just a 42 weighted average of TSS and ATL is a 7 day weighted average of TSS. You must add them up one line at a time to get the weighted average.
Chronic Training Load = [Todays TSS * (1-e^(-1/42)] + {Yesterdays CTL * (e^(-1/42)]
ATL is the same but with 7 (well 5 to 10 days depending)
You basically need a spreadsheet to do this without some complex database calculations. Using Excel you just need to calculate your score for your CTL using your previous CTL and current TSS scores and add in todays score.
I assume you are using a spreadsheet to do this, without it would be very hard.
Note that exp(-1/42) is pretty close to 41/42, and exp(-1/7) is pretty close to 6/7. You can reasonably just calculate today's ATL as 6/7*yesterday's ATL + 1/7*today's TSS, and CTL as 41/42*yesterday's CTL +1/42*today's TSS.
ATL/CTL, and TSS itself, are obviously just estimates. Getting the decay constant "right" to some number of significant digits isn't as important as being consistent with the inputs, and most of all actually using the information in some way to guide your training.
My TSB usually gets up around +10 or +15 by the end of a recovery week. It's a balance between losing enough fatigue that you can maximize your hard work during the next block of work, and still maximizing your overall fitness.
Last edited by globecanvas; 04-03-14 at 08:50 PM.
I would agree 100% and think you did a better job of explaining it than I did. And I obviously missed the first sentence where he said he was using a spreadsheet. I somehow was thinking he was not.
For CTL you are not going to see much drift at all using the non exp formula, for ATL you will seem some drift as that number is much further off (7% give or take). As you are using a spreadsheet you might as well use the exp(-1/42) formula, but as they say 5 to 10 days for ATL, it is really a swag and once again the simple equation might be just as good as long as you are consistent. I think it is important to note your RPE and basic feelings on days that feel odd so that you can see how well the form/fatigue numbers work for you.
I have found that the form numbers don't really work that well for me and I am still trying to tweak my spreadsheet to better predict how RPE and results equate to the TSS and other numbers. TSS and CTL vs ATL have been very useful, just not in the simple way of CTL-ATL = form.
Jmikami thanks for the insightful response!
Agreed, a lot more variability than I would have liked, intervals 4 and 5 were much too low. I was trying out a new spot for my 3' intervals. Its on a 5% hill, and agreed WR, I need to ride 2' more uphill following. Unfortunately I have to ride down hill as well to get back to the start, which makes it a bit cold at times.
Nothing should come between you and your chamois -- lawkd
Also found this to be true.
FWIW, I did the Allen and Coggan 20' test protocol about 2 months ago on a trainer and averaged 223W on the 20' interval, for a calculated FTP of 212, and at 75kg, not so good. Fast forward to last weekend, I went on one of my first group rides of the season with some fast locals. One is a pro. 45 mile ride, hilly (3k elevation). After going pretty hard for the first 2/3s of the ride, I got dropped on the long run in home that was mostly uphill, nothing over 3%, and headwind, smooth pavement, no stop signs. Not only did I have the mild uphills and head wind, but I had another dropped rider ahead of me to chase. I decide to do a hard time trial and averaged 283 watts for 30 min back into town. I'm highly skeptical my threshold would truly climb that much after 2 months of base training minus 1.5 weeks with stomach flu, so now I'm leery of doing any testing on a trainer.
My initial trainer solo test was similar to yours, and we're roughly the same size. However, Looking at my results from a 50 minute crit and a 94 minute road race confirmed that I can crank out somewhere between 240 and 250 wats for an hour at a time under race conditions (likely a little more if I wasn't trying to pace myself and save a little for hills and sprints). So I'm using 240 as my baseline for training and pacing. That puts me at 3.2 w/kg, which is in line with a 50 yr old cat 5 newbie. Given that they hand out trophies based on when you cross the finish line and not what your power meter says, I'll live with that number until I see consistent results at least 10% higher.
BB
www.beancotton.com
Formerly Fastest of the Slow Riders, Currently Slowest of the Fast Riders
http://veloviewer.com/athlete/2615827/
(3) is correct. the time constant is 42 and 7, not exp(-1/42). If you think about it, exp(-1/42) ought to be correct as CTL is primarily stuff from the previous 42 days, with stuff from the last day matter little. Therefore, the contribution from the previous day should account for most of the new number. exp(-1/7) should be less as the more recent events are weighted more.
THe actual expression are as follows
ATL (today) = ATL(yesterday) * (exp (-1/7)) + TSS (today) * (1-exp(-1/7))
CTL (today) = CTL(yesterday) * (exp (-1/42)) + TSS (today) * (1-exp(-1/42))
also, one last note, for globecanvas's point, an exponential may be written as an infinite series, names exp(x) = 1 + x + x^2/2! + x^3/3! +...; you cut it off after two terms, with x = -1/7, you get exp(-1/7) ~ 1 + (-1/7)