ATL and CTL Constants Adjustment
- Thread starter kaparzo
- Start date
kaparzo said:
It sounds as if you are constantly overriding the calculated CTL and ATL with new "seed" values. (I don't think you are referring to the time constants, because those are constrained to much lower ranges than most anyone's average TSS over 7-42 d.) In essence, that means that you're using the Performance Manager chart just to track your shorter-term rolling averages, rather than having it reflect the longer-term exponentially-weighted moving averages as intended.
To answer your question more directly: unless you're new to using a powermeter and/or are missing large amounts of data for some reason, you don't need to adjust the seed values at all...just leave them at 0.
Concerning the time-constants, has anyone experimented with them?
How were the 7 and 42 day constants established?
If they were the result of analysing some statistics from multiple riders, what was the spread in values for different riders etc?
In other words, how likely is it that different riders should have to adjust these time-constants? and if so, what could be a protocol for doing so?
If I understand correctly, ATL-time is the timescale on which the FATIGUE from a training session fades, and CTL-time is the timescale on which the BENEFIT from a training session fades?
(Obvously if you want to adjust a 42 day time-constant you need to spend multiple 42 day periods to have enough data.)
(I'm not serious enough about training to need this, but just curious at heart)
How were the 7 and 42 day constants established?
If they were the result of analysing some statistics from multiple riders, what was the spread in values for different riders etc?
In other words, how likely is it that different riders should have to adjust these time-constants? and if so, what could be a protocol for doing so?
If I understand correctly, ATL-time is the timescale on which the FATIGUE from a training session fades, and CTL-time is the timescale on which the BENEFIT from a training session fades?
(Obvously if you want to adjust a 42 day time-constant you need to spend multiple 42 day periods to have enough data.)
(I'm not serious enough about training to need this, but just curious at heart)
For the curious at heart: http://home.trainingpeaks.com/power411.aspxhmronnow said:
See items #12 & #13 in particular for the questions you asked, but the rest is all interesting as well.
There were quite a few discussions on time constants on the Wattage Group forum (requires membership in the group).hmronnow said:
As indicated in an earlier post by frenchyge, http://home.trainingpeaks.com/articles/cycling/the-science-of-the-performance-manager.aspx discusses the rationale behind the default TCs.
There appears to be less inter-athlete and intra-athlete variability for the CTL TC. Consequent to this, I have not played with the CTL TC in my PMC. It is set at 42 days.
Conversely, there is evidence to suggest there is greater relative variability in the ATL TC - both inter-athlete and intra-athlete. Factors that may impact an athlete's ATL TC include
Consequent to this, I played with my ATL TC after my 2007 season and found that a 10 day TC allowed my PMC to reflect my actual season peaks and valleys better than the 7 day TC. I played with a higher TC primarily due to my age - 50+. My ATL TC has remained set at 10 days since the Fall of 2007.
I have not played with an intra-seasonal variable ATL TC in PMC yet. While I am toying with the idea of running two PMCs this season for this purpose, there are a number of issues:
While I'm rambling on, I'll share another thought on the interplay between the CTL TC and the ATL TC: because the of the longer decay rate with CTL than ATL, when one finds that their FTP has bumped up and makes the change in WKO+, there will be a resultant period of relatively erroneous TSB (positively skewed) while the newer "price" of TSS will impact ATL relatively greater than CTL (lower relative ATL than CTL) until perhaps one ATL TC duration has passed, before the TSB is accurate again. This might prompt one to shorten the ATL TC after an FTP bump, and then gradually increase the TC back to the pre-FTP bump setting.
All relatively little things in the grand scheme of things, but perhaps leading to some fine tuning of an already pretty damn good tool.
There appears to be less inter-athlete and intra-athlete variability for the CTL TC. Consequent to this, I have not played with the CTL TC in my PMC. It is set at 42 days.
Conversely, there is evidence to suggest there is greater relative variability in the ATL TC - both inter-athlete and intra-athlete. Factors that may impact an athlete's ATL TC include
- relative age
- musculoskeletal stress (tissue strain, not muscular metabolic stress)
- perhaps off-the bike fatigue (lack of sleep, weight work)
- perhaps higher intensity work (L6)
- perhaps greater time working at the upper end of one's force/pedal velocity curve
Consequent to this, I played with my ATL TC after my 2007 season and found that a 10 day TC allowed my PMC to reflect my actual season peaks and valleys better than the 7 day TC. I played with a higher TC primarily due to my age - 50+. My ATL TC has remained set at 10 days since the Fall of 2007.
I have not played with an intra-seasonal variable ATL TC in PMC yet. While I am toying with the idea of running two PMCs this season for this purpose, there are a number of issues:
- what would be the determinant for changing the ATL TC (eg, perception of general body and general muscular fatigue?)
- how many days up or down should a TC variation be?
- how frequent would one change the ATL TC?
- will the PMC even be sensitive to these changes?
While I'm rambling on, I'll share another thought on the interplay between the CTL TC and the ATL TC: because the of the longer decay rate with CTL than ATL, when one finds that their FTP has bumped up and makes the change in WKO+, there will be a resultant period of relatively erroneous TSB (positively skewed) while the newer "price" of TSS will impact ATL relatively greater than CTL (lower relative ATL than CTL) until perhaps one ATL TC duration has passed, before the TSB is accurate again. This might prompt one to shorten the ATL TC after an FTP bump, and then gradually increase the TC back to the pre-FTP bump setting.
All relatively little things in the grand scheme of things, but perhaps leading to some fine tuning of an already pretty damn good tool.
right, my question was posed because I gradually move away from using PMC as the season wears on (i'm an mtb racer). thus, i have a lot of missing data and was wondering if adjusting the values to respectively reflect both ATL and CTL would give more accurate results, given the lack of data. i start the season at scratch data wise for the pmc.
The seeds are just a way to estimate your starting load for exactly the situation you have coming off a mountain bike with no PM and moving to off season road training with power data.kaparzo said:
All you're trying to do with the seeds is to estimate your starting CTL and to a lesser extent your ATL so that the curves don't start from zero. If your estimate is reasonably accurate you shouldn't need to adjust them over time. The only real reason to reset the starting seeds after training for a while is if you're convinced your estimates weren't very good.
I generally tell folks to start with matching CTL and ATL seeds. This assumes you started from a neutral TSB or not particularly tapered nor deeply fatigued. If you start your PMC immediately after your MTB season wraps up and you're knackered then consider seeding ATL 10 to 20 points higher than CTL. Or if you start your PM based training after a couple of weeks of rest consider seeding ATL 10 to 20 points lower than CTL. Either way it doesn't matter much as it only takes a few weeks for ATL to settle even with a poorly choosen seed value. Just don't put too much faith in ATL or TSB during the early weeks of your PMC.
CTL is more important to seed well if you've been training but without PM data. It takes three to four months for CTL to settle if you've missed on guessing your starting seed. If you've seeded CTL too low relative to the workload you can easily sustain day in and day out then your CTL ramp will appear really steep (greater than 8 to 10 TSS/day/week) but you'll feel fine and not not overly stressed. In that case you might go back and bump the CTL seed upwards to reflect your normal day to day training load.
Similarly if you seed CTL too high and resume typical training loads your CTL curve will start flat or drop for a while before trending upwards with increasing load. IOW, your CTL curve will show a taper even when you're working normally or even increasing load. If that's the case then it makes sense to go back and lower the seed value.
But basically all you're doing with the seeds is estimating you typical day to day training load, not what you might be capable of if you kill yourself nor an easy rest period load but what you're currently sustaining when you start collecting PM data. I don't know what class or the level and length of your MTB races or your training patterns, but I'd guess that if you didn't rest for too long after your MTB season you'd want to start your CTL seed in the 70s or 80s or even higher if you train a lot and have done so for many months. If you've taken a big break after your MTB season than a CTL seed value in the 50-70 range might be more appropriate.
Good luck,
-Dave
also, the seed values only change the *starting point* for the chart in question. So, changing them on a weekly basis would basically be retro-changing all the previous data as one progressed through the season.
As more "good data" finds its way into the chart, the seed values becomes less important. As Dave describes above, take a stab at the seed values, then leave them alone as good files are generated.
As more "good data" finds its way into the chart, the seed values becomes less important. As Dave describes above, take a stab at the seed values, then leave them alone as good files are generated.
As guide, when power meter data is not available and you want to pick some seed values, I go with:
CTL = 8 x average weekly hours for past 2-3 months
ATL = 8 x average weekly hours for past 2-3 weeks
It'll get you in the ball park.
CTL = 8 x average weekly hours for past 2-3 months
ATL = 8 x average weekly hours for past 2-3 weeks
It'll get you in the ball park.
