TSS w/ HR training?

ccrnnr9 said:

After reading an article on TrainingPeaks by Coggan, he says that his TSS idea is based off of the concept developed by Bannister called TRIMPS (training impulses?). I have done some searching and am trying to find exactly how one could calculate TRIMPS for a certain workout based on HR (as I do not own a power meter). Here is the only thing I can find:

"TRIMPS = exercise duration x average HR x an intensity-dependent
weighting factor

---

That being said, the first two factors in the equation are straight forward, but what does he mean by "an intensity-dependent weighting factor"? Any help would be appreciated.
~Nick

Wether you are familiar with the TSS calculation algorithm, the 30 s rolling average powers are ^4, then all samples are averaged, and the average is ^(1/4) to get the normalised power. I'd assume that similar biquadratic weighting factor could be generated by some other method as well.

Edit: Assuming power & HR be:
Watts HR
200 135
240 150
280 165

Then riding five minutes on each power, and each HR. I'd get NP to be 248 Watts, and the "normalized HR" to be 152, which would quite well correspond the 248 Watts. However, there are not included the lags in the power-HR responses, any powers exceeding the max HR power, cardiac drift, etc. But yes, some HR based algorithm can be developed.

This will almost seem blasphemous to many of the power training folks around here, but I too have looked for a way to gain TSS data from my workout information pre-powertap. I just spent a year cycling my butt off and want to be able to make comparisons to my fitness/taper performance over the last year's worth of racing.

The reality is, the PMC, TSS, ATL,CTL are all formulas held under lock and key - for obvious reasons. The PMC is a money maker. But the concepts are out in the open and your ride information is your own.

But really, if you're only comparing you to... well.. yourself... the actual numeric value isn't important except for the points where you cross a "Zero" line and the overall trends. All this in mind, I took some data that I have from my powerfiles and made some averages.

First, we know there appears to be an "average" relationship to your calories/kilojoules burned to your TSS. Over the course of time there will be a rough average of the number of calories you burn per hour. This assumes you vary your workouts on a seemingly routine basis. My taper weeks end up being a slightly lower calories per hour, while more intense weeks obviously the other direction. (Note, the polar HRM calories per hour was WAY higher than what my powertap reports, so I threw those numbers away in exchange for using my average calories/hour)

Based on this average calories per hour, I took the hours on the bike over the last year ride-by-ride. Likewise, assuming OVER TIME, an average comparison of TSS to calories burned (again - higher for more intense, lower for less intense, but over all looking for TRENDS, not exaction). Taking this data and plugging it in the worksheet you can find through a thorough search of the forums, I was able to plot "general" TSS values across the 2006 season and look at the trends established. I can say that there are definitely similarities in the trends, and most importantly - crossing the "Zero line" of TSB seems to correspond with my tapering for events throughout the year.

I can't stress enough that this is my tool for examining the trends of last year's training and not the exact stress any particular ride caused. But I'm excited to see the closeness between the trends. Here's an image of a comparison between the real values of the original TSTWKT sheet based on TSS and the same information using estimated TSS values - based on calories burned over duration.

[img]cycleiwakuni.comtom thumb kJvstss.jpgOpen ↗[/img]

Likewise, this is what I would end up with for a 2006 TSTWKT using a TSS value estimated by calories/hour. The tapers match up and in general the results match with the values on the chart - even to include the points at which TSB crosses the zero line during tapers. Uncanny even.

[img]cycleiwakuni.comtom thumb 2006tstwktkjbased.jpgOpen ↗[/img]

I haven't given real thoughts to TRIMP yet, but in the next few months I will :

- Write a paper on it in collaboration with Dr.Jim Carter (Trimp expert)
- Implement TRIMP calculations in my little power file analysis tool

Until then, attached you'll find a little introductory text on TRIMP.

NomadVW said:

The reality is, the PMC, TSS, ATL,CTL are all formulas held under lock and key

Uh, no: all of the formulae are in the public domain. The terms and acronyms have been trademarked by TrainingPeaks LLC, but even that doesn't prevent someone else from using them in a commercial context - they just have to acknowledge the owner of the trademark should they do so.

acoggan said:

Uh, no: all of the formulae are in the public domain.

Ah.. huh.. Sorry! Learn something new every day 🙂

SolarEnergy said:

Until then, attached you'll find a little introductory text on TRIMP.

A little near-instantaneous peer review:

1) For completeness, you might want to give the formulae for y (which as you know, is different for men and women).

2) It's never been 100% clear to me from the studies I've read, but I believe that many (if not most) of those using TRIMP break down each workout into segments, calculate TRIMP for each separately, then sum them up to arrive at the final value.

3) Are you sure that your formulae for calculating fitness, fatigue, and performance are correct? There seem to be some summation signs missing...

ccrnnr9 said:

After reading an article on TrainingPeaks by Coggan, he says that his TSS idea is based off of the concept developed by Bannister called TRIMPS (training impulses?). I have done some searching and am trying to find exactly how one could calculate TRIMPS for a certain workout based on HR (as I do not own a power meter). Here is the only thing I can find:

"TRIMPS = exercise duration x average HR x an intensity-dependent
weighting factor

Since HR is essentially linearly related to oxygen uptake (metabolic
rate), the product of the first two factors in the above equation is
proportional to the amount of energy expended, or (since efficiency is
relatively constant), work performed. The third term then takes into
account the intensity of the exercise, since many physiological
responses (e.g., glycogen utilization, lactate accumulation) increase
non-linearly with increasing intensity."

That being said, the first two factors in the equation are straight forward, but what does he mean by "an intensity-dependent weighting factor"? Any help would be appreciated.
~Nick

Nick,

I don't have a PM (yet) and I estimate all my TSS. I've been down the same road as you, exactly actually. Here is a search for TRIMPS:
http://www.cyclingforums.com/search.php?searchid=553657

You'll see 2 threads started by me. First is the same exact question you ask:
http://www.cyclingforums.com/t368193.html

Later I bagged the TRIMPS idea and went with estimating TSS:
http://www.cyclingforums.com/t370109.html

So far this works for me. I use common values for TSS/hour, as such:

L2: 64
L3: 81
L4: 90
L5/L6: 100
I also made up one for my mountain biking: 72

I do my L4+ workouts on a trainer, so it's not terrible inaccurate (IMO). For L3 I use the trainer or the road. I occasionally do L4 on the road as well but only as longer hill repeats. MTB is all outside, obviously.

I find RPE better than HR for measurement. I have an HRM but gave up on it when I did the exact same L4 workout on back to back days and saw large differences.

Is this perfect? Maybe, but highly doubtful. I'm learning alot about myself, for sure. And I think the key is that each week my riding is basically the same composition. I ride 1 off-road ride, 1 road ride, and 3 trainer rides. So this metric is being compared on a weekly basis to a metric which has been generated in past weeks using the same (potentially shoddy) methods. Point being, if my 72 TSS/hour estimate on my off-road rides is off by 25%, no big deal since it's being measured against a 25% inaccuracy from the previous week, and so on.

Hope this helps.

acoggan said:

A little near-instantaneous peer review:

1) For completeness, you might want to give the formulae for y (which as you know, is different for men and women).

this post here http://www.cyclingforums.com/showpost.php?p=3034698&postcount=4
may lead to believe that I know (or knew) while in fact all I know is how to copy/paste formulas. 😄

But yeah, I have noticed that there are different values for males and females.

acoggan said:

2) It's never been 100% clear to me from the studies I've read, but I believe that many (if not most) of those using TRIMP break down each workout into segments, calculate TRIMP for each separately, then sum them up to arrive at the final value.

hmm... I'm going to have a party implementing this in a program. Thanks for getting me dreaming 😉

acoggan said:

3) Are you sure that your formulae for calculating fitness, fatigue, and performance are correct? There seem to be some summation signs missing...

You know what? I really haven't gave this document a real thought. Sorry. It's been sent to me by Jim Carter. Our interview is in progress now, but it is still in its early stage. In depth review of the formulas is yet to come.

But here (http://www.triforum.ca/articles/general/TahaandThomas.pdf). Here's a (private) link to an other technical document on Trimp. This one has been sent to me by someone in which I have total trust 😉

I've bookmarked this thread and intend in comming back to post any interesting material whenever it becomes available.

Thanks a lot Andy for your feedback.

normZurawski said:

... I use common values for TSS/hour, as such:
L2: 64
L3: 81
L4: 90
L5/L6: 100
I also made up one for my mountain biking: 72...

I've been using a similar method for estimating TSS with crosstraining, mostly XC classic and ski skating. Your per hour estimates seem a bit high.

Remember TSS = IF^2 * 100/hour.

By your estimates an L2 ride would have to average to .8 IF across each hour and your L3 rides would average .90. By definition a ride with a .8 IF is higher than L2 and .90 puts you in L4. Yeah, IF and TSS is based on NP, not AP but still those estimates seem high to me.

I also base my estimates on RPE since I've got enough HR vs. Pwr data now to show me how unreliable HR is for estimates like this. When I'm out skiing or running or mountain biking (no PM on the mountain or cross bikes) I just ask myself how my pace relates to my one hour pace for that activity. I also try to take into account any real hard sections like steep climbs that would drive my NP up even if the AP isn't all that high. So basically I estimate IF and plug it straight into the TSS formula above.

NomadVW said:

First, we know there appears to be an "average" relationship to your calories/kilojoules burned to your TSS. Over the course of time there will be a rough average of the number of calories you burn per hour. This assumes you vary your workouts on a seemingly routine basis. My taper weeks end up being a slightly lower calories per hour, while more intense weeks obviously the other direction.

There's no way I could make this work for me. See the spread in my TSS v kJ relationship.

There was a discussion on using HR on wattage:
http://groups-beta.google.com/group/wattage/browse_frm/thread/e1b39a07586a9ad8/0ca7eca8172dbeaa?lnk=gst&q=estimating+tss+with+heart+rate&rnum=4&hl=en#0ca7eca8172dbeaa
but just do a search on "estimating TSS with heart rate"

SolarEnergy said:

here (http://www.triforum.ca/articles/general/TahaandThomas.pdf). Here's a (private) link to an other technical document on Trimp. This one has been sent to me by someone in which I have total trust 😉

Funny - I was going to mention the Taha and Thomas review as another source of information...

Anyway, see Eq. 8 of that paper, and note how it differs from the continuous transfer functions that led to its development...

daveryanwyoming said:

By your estimates an L2 ride would have to average to .8 IF across each hour and your L3 rides would average .90. By definition a ride with a .8 IF is higher than L2 and .90 puts you in L4. Yeah, IF and TSS is based on NP, not AP but still those estimates seem high to me.

Those IF values are straight from the source, based on typical outdoor rides at the specified levels. A dead-center L3 ride with an AP of .83FT and a VI of 1.08 has an IF of .9, which seems pretty reasonable to me.

cyclingpeakssoftware.comdefined.aspOpen ↗
[b][QUOTE] [/b]

Typical IF values for various training sessions or races are as follows:

• Less than 0.75 recovery rides
  
• 0.75-0.85 endurance-paced training rides
  
• 0.85-0.95 tempo rides, aerobic and anaerobic interval workouts (work and rest periods combined), longer (>2.5 h) road races
  
• 0.95-1.05 lactate threshold intervals (work period only), shorter (<2.5 h) road races, criteriums, circuit races, longer (e.g., 40 km) TTs
  
• 1.05-1.15 shorter (e.g., 15 km) TTs, track points race
  
• Greater than 1.15 prologue TT, track pursuit, track miss-and-out
  

frenchyge said:

Those IF values are straight from the source, based on typical outdoor rides at the specified levels. A dead-center L3 ride with an AP of .83FT and a VI of 1.08 has an IF of .9, which seems pretty reasonable to me...

Looks like I'm the one who's underestimated IF for various workouts. :o Thanks for the reference....

acoggan said:

Funny - I was going to mention the Taha and Thomas review as another source of information...

Anyway, see Eq. 8 of that paper, and note how it differs from the continuous transfer functions that led to its development...

They look different but if "t" is discrete it's easy to show that Eq. 8 is equivalent to Eqs. 5,6,7 together (ej. by math induction).

In fact if you start from the discrete version of the original differential Eq.2 using Z-Transform it solves to the recursive expression.[font=Times-Roman]
[/font]

Alex Simmons said:

There's no way I could make this work for me. See the spread in my TSS v kJ relationship.

I get a similar relationship if I use simply TSS to kJ as you depicted on your chart. But when I move over to kJ per hour to TSS, I end up with a much more linear relationship. For practical purposes and training from here out, I'll obviously continue to use the PMC and power based TSS. Nor would I consider comparing absolute TSS value of a kJ/hour estimate to a TSS from Pnorm.

But for the purpose of displaying trends, my relationship of roughly 771 kJ per hour has demonstrated on paper a reasonably close comparison to the trend lines of the PMC. Some days this "estimated" TSS is Ultimately, the use of the individual is to gain a consistent metric that he/she can use for their own training. In the case of power based training, the metric has been defined.

Fortunately, I only have one year of "back tracking" to look at, and now I don't need to worry about it. It's trivial for me at this point.

NomadVW said:

I get a similar relationship if I use simply TSS to kJ as you depicted on your chart. But when I move over to kJ per hour to TSS, I end up with a much more linear relationship.

I must be thick - I don't get the difference.😕

You made one comment earlier about more intense weeks having higher kJ. I dunno about that for me. If I'm doing higher intensity track work (L5,6,7), I would probably burn far fewer kJ (even on an kJ/hour basis given volume of L1 recovery between efforts, esp L6/7) than when in an endurance phase but TSS would probably even out.

Like you say, it's no big deal. You have the power!

MMDV (my mileage does vary 😉)

Alex Simmons said:

I must be thick - I don't get the difference.😕

Honestly, I think I musta drawn a bum chart or something. I re-drew a chart tonight and it certainly shows a rather liner TSS to kJ relationship. Take a peek.

I think at this point I'm just having fun with Excel more than anything else. Something to do to keep me awake for absolutely no reason.

[img]cycleiwakuni.comtom thumb kJvstss2.jpgOpen ↗[/img]

NomadVW said:

Honestly, I think I musta drawn a bum chart or something. I re-drew a chart tonight and it certainly shows a rather liner TSS to kJ relationship. Take a peek.

This is the result you tend to obtain when all of your workouts have fairly similar VI values. Throw in some highly variable training (or racing), however, and the difference between TSS and total work becomes more evident. Indeed, even your chart shows that the relationship isn't strictly linear, but is in fact curvilinear (due to the fact that TSS is proportional to IF^2, and longer rides tend to be at a lower overall intensity).