Agreed, there are definitely times (events) where b2b2b hard days can be really beneficial.
Personally I would find that physically and mentally tough week after week.
I guess the POL counter point is: I’ve done plenty of 4hr rides so 5hrs isn’t that much longer. This climb looks like a VO2 max effort - yep done plenty of that too. Not a problem - lets keep going 
IMHO there is no right / wrong - just what works for the individual.
I’m having trouble understanding your graphs as when I try and guess the values you’ve used for each zone I get different PI values to what you’ve stated. Are you able to share the raw data used? Thanks.
Yet another possible metric but read part 5 and look at the limitations and you’ll see yet again there are a multitude of factors that could affect outcomes, and thats before you even look at the fact that currently only a few HR straps produce data that is even close to being sufficiently artifact free to be used for any quality results.
I’m sure the research and tech will get there at some point but we are still a long way from that right now. But once again I honestly think its missing the point to some degree - if you train to some intensity level that ultimately impacts your ability to perform your high intensity sessions, then you’ve got it wrong. It doesn’t take long to know how hard to go on easy days to allow yourself to hit the z3 days fresh enough to nail them. It’s not like trying to bash out another SST session when you feel a little tired. Trying to nail 4x8m @105% will quickly tell you if you went too hard in the previous days…
Well, here’s hoping I got this right…
For each of the Z1 percentages (from 100% down to 0%) I back calculated the ratios of the other two zones that gave a certain P.I.
The table below presents percentages but as discussed above, you have to use a decimal ratio or drop the x100 from the equation:
| P.I. | = 2 | ||
|---|---|---|---|
| Z1 | Z2 | Z3 | P.I. |
| 100% | 0.00% | 0.00% | |
| 99% | 0.50% | 0.50% | 2.00 |
| 98% | 0.99% | 1.01% | 2.00 |
| 97% | 1.48% | 1.52% | 2.00 |
| 96% | 1.96% | 2.04% | 2.00 |
| 95% | 2.44% | 2.56% | 2.00 |
| 94% | 2.91% | 3.09% | 2.00 |
| 93% | 3.37% | 3.63% | 2.00 |
| 92% | 3.83% | 4.17% | 2.00 |
| 91% | 4.29% | 4.71% | 2.00 |
| 90% | 4.74% | 5.26% | 2.00 |
| 89% | 5.18% | 5.82% | 2.00 |
| 88% | 5.62% | 6.38% | 2.00 |
| 87% | 6.05% | 6.95% | 2.00 |
| 86% | 6.47% | 7.53% | 2.00 |
| 85% | 6.89% | 8.11% | 2.00 |
| 84% | 7.30% | 8.70% | 2.00 |
| 83% | 7.71% | 9.29% | 2.00 |
| 82% | 8.11% | 9.89% | 2.00 |
| 81% | 8.50% | 10.50% | 2.00 |
| 80% | 8.89% | 11.11% | 2.00 |
| 79% | 9.27% | 11.73% | 2.00 |
| 78% | 9.64% | 12.36% | 2.00 |
| 77% | 10.01% | 12.99% | 2.00 |
| 76% | 10.36% | 13.64% | 2.00 |
| 75% | 10.71% | 14.29% | 2.00 |
| 74% | 11.06% | 14.94% | 2.00 |
| 73% | 11.39% | 15.61% | 2.00 |
| 72% | 11.72% | 16.28% | 2.00 |
| 71% | 12.04% | 16.96% | 2.00 |
| 70% | 12.35% | 17.65% | 2.00 |
| 69% | 12.66% | 18.34% | 2.00 |
| 68% | 12.95% | 19.05% | 2.00 |
| 67% | 13.24% | 19.76% | 2.00 |
| 66% | 13.52% | 20.48% | 2.00 |
| 65% | 13.79% | 21.21% | 2.00 |
| 64% | 14.05% | 21.95% | 2.00 |
| 63% | 14.30% | 22.70% | 2.00 |
| 62% | 14.54% | 23.46% | 2.00 |
| 61% | 14.78% | 24.22% | 2.00 |
| 60% | 15.00% | 25.00% | 2.00 |
| 59% | 15.21% | 25.79% | 2.00 |
| 58% | 15.42% | 26.58% | 2.00 |
| 57% | 15.61% | 27.39% | 2.00 |
| 56% | 15.79% | 28.21% | 2.00 |
| 55% | 15.97% | 29.03% | 2.00 |
| 54% | 16.13% | 29.87% | 2.00 |
| 53% | 16.28% | 30.72% | 2.00 |
| 52% | 16.42% | 31.58% | 2.00 |
| 51% | 16.55% | 32.45% | 2.00 |
| 50% | 16.67% | 33.33% | 2.00 |
| 49% | 16.77% | 34.23% | 2.00 |
| 48% | 16.86% | 35.14% | 2.00 |
| 47% | 16.95% | 36.05% | 2.00 |
| 46% | 17.01% | 36.99% | 2.00 |
| 45% | 17.07% | 37.93% | 2.00 |
| 44% | 17.11% | 38.89% | 2.00 |
| 43% | 17.14% | 39.86% | 2.00 |
| 42% | 17.15% | 40.85% | 2.00 |
| 41% | 17.16% | 41.84% | 2.00 |
| 40% | 17.14% | 42.86% | 2.00 |
| 39% | 17.12% | 43.88% | 2.00 |
| 38% | 17.07% | 44.93% | 2.00 |
| 37% | 17.01% | 45.99% | 2.00 |
| 36% | 16.94% | 47.06% | 2.00 |
| 35% | 16.85% | 48.15% | 2.00 |
| 34% | 16.75% | 49.25% | 2.00 |
| 33% | 16.62% | 50.38% | 2.00 |
| 32% | 16.48% | 51.52% | 2.00 |
| 31% | 16.33% | 52.67% | 2.00 |
| 30% | 16.15% | 53.85% | 2.00 |
| 29% | 15.96% | 55.04% | 2.00 |
| 28% | 15.75% | 56.25% | 2.00 |
| 27% | 15.52% | 57.48% | 2.00 |
| 26% | 15.27% | 58.73% | 2.00 |
| 25% | 15.00% | 60.00% | 2.00 |
| 24% | 14.71% | 61.29% | 2.00 |
| 23% | 14.40% | 62.60% | 2.00 |
| 22% | 14.07% | 63.93% | 2.00 |
| 21% | 13.71% | 65.29% | 2.00 |
| 20% | 13.33% | 66.67% | 2.00 |
| 19% | 12.93% | 68.07% | 2.00 |
| 18% | 12.51% | 69.49% | 2.00 |
| 17% | 12.06% | 70.94% | 2.00 |
| 16% | 11.59% | 72.41% | 2.00 |
| 15% | 11.09% | 73.91% | 2.00 |
| 14% | 10.56% | 75.44% | 2.00 |
| 13% | 10.01% | 76.99% | 2.00 |
| 12% | 9.43% | 78.57% | 2.00 |
| 11% | 8.82% | 80.18% | 2.00 |
| 10% | 8.18% | 81.82% | 2.00 |
| 9% | 7.51% | 83.49% | 2.00 |
| 8% | 6.81% | 85.19% | 2.00 |
| 7% | 6.08% | 86.92% | 2.00 |
| 6% | 5.32% | 88.68% | 2.00 |
| 5% | 4.52% | 90.48% | 2.00 |
| 4% | 3.69% | 92.31% | 2.00 |
| 3% | 2.83% | 94.17% | 2.00 |
| 2% | 1.92% | 96.08% | 2.00 |
| 1% | 0.98% | 98.02% | 2.00 |
| 0% | 0.00% | 100.00% | 2.00 |
Mike
@_Matthew It looks as though we are having a side conversation here!
I agree that what they are talking about is using ML rather than regression techniques. But what I am suggesting is this.
Suppose that the machine has learned from the existing kinds of data about us that we know TR has access to – age, gender, weight and training history. From this it proposes workouts for us and can learn from our performance on those workouts.
Now suppose that TR says, feed us your HRV [or whatever data] and then feed that into the machine. The machine has to learn whether the HR metric makes any difference to performance on the proposed workouts. As far as adaptive training is concerned, the question is simply whether the HR metric makes any difference to the proposed workout – ie, whether it improves the ability to accurately propose appropriate workouts.
But as far as science is concerned, the question is what the machine learns about the implications of HRV for performance. If the machine learns that HRV seems to make no difference, one hypothesis for the RCT people. If the machine learns that HRV seems to make a difference [perhaps for some types of people], then there is another hypothesis for the RCT people to go after.
That is, there are two interested “users” here – we, the people being trained; and the scientists, looking for hypotheses that have a good chance of being correct [or, for the pedants, of not being proven incorrect].
Really enjoyed the podcast and give two thumbs up to the TR team, Amber, and Nate.
I don’t see how you’re making this assumption. Yes optical HR sensors don’t do well during exercise at reporting accurate rr time but chest straps work fine:
https://www.hrv4training.com/blog/hardware-for-hrv-what-sensor-should-you-use
There’s a big thread about HRV which includes posts by someone involved in developing and marketing a HRV app. That uses either the phone’s camera or a chest strap.
I see what you are saying about hypothesis generation by gauging the contribution of a variable to model fit. I think it’s just really hard to tease out which variables matter more.
In stepwise regression variable selection, the order of entry of variables matters. If you call for forward or backward selection, the end model can differ. Imagine you added in HRV, then a variable which previously seemed insignificant becomes significant as if HRV was a negative confounder.
It’s better to have a priori hypotheses with a conceptual framework and then focus on one single variable of interest to put the hypothesis to the test.
To date, the form of big data hypothesis generation I’ve seen is something like screening drugs. You hit cell cultures, benign or malignant, with hundreds of different drugs to see if RNA expression changes in interesting ways. You might find something like a statin or an antihistamine affecting a expression of an oncogene (a gene that drives cancer), then you are left with wondering if that drug would really work to prevent or treat cancer. First of all, is that real or a chance finding. All tests have a rate of false positives. What dose matters? Over what time frame? When? There are big biomarker studies where you take blood or tissue from patients with X disease but at the end of the day you scratch your head if your findings are meaningful. In imaging, there are worries that AI overdiagnoses cancers.
AI, ML, big data - this type of stuff seems more helpful when the platform can create data rather than just look back on existing data. There’s a fascinating documentary on how the AlphaGo creators built an AI to play Starcraft 2. The computer plays so many games that it spends virtual time far greater than the total time of human existence. It then does some crazy things while pummeling professional players, like use an excessive number of mining probes that humans would think inefficient.
AI/ML is a very hot topic, but I’m waiting for concrete examples of it doing something to advance medicine before getting excited about it.
Ah, I see. I didn’t realise you were using two decimal places in the percentage; hence the number of variations you get for each PI value.
BTW, if zone 3 > zone 1 then PI is not valid based on info here:
The innateness of this thread has made me realise I like to simply go ride my bike sometimes(!) 
Yeah, that would make it HIIT, or whatever you want to call it.
Considering this is based on time in zone, its only a handful of ratios on the left hand side that are truelly polarized.
I’ll take that as a compliment. Around these parts calling someone a spanner is less than complimentary. Perhaps I should reconsider my handle…
Mike
@ambermalika Nate’s menstrual cycles at sea level and at elevation had me on the floor! You’re absolute right!!
Thank you @ambermalika and @Nate_Pearson. I have learnt more in the first 35 mins of this podcast than I have in the last 3 years of cycling combined. That was amazing!
I wish I had Amber as a teacher/lecturer at university when I was younger. I would of understood a whole lot more!
Isn’t one of TrainerRoad’s basic training goals to train ALL the energy systems? I’ve always understood this to mean that some workouts will focus on your aerobic endurance and some will focus on your ability to achieve and repeat VO2 Max and higher intensities. Sweet Spot trains both your aerobic and anaerobic capacity. I’ve always thought I need to do at last one or two intervals workouts/week above my FTP. If I try to do more than two HIITs per week I run the risk of overreaching and eventually overtraining. Only two of the five TIDs Amber identified include Zone 3, over FTP intervals.
Agreed.
I also have to confess to being bemused by the big data hype: it seems to go against everything I was trained in – you have to have a good idea before you go looking for / at data. But then, that was more than half a century ago.
But I do like the idea of adaptive training, which seems to offer the prospects of setting more personalised progressions than are possible in generic training plans. That is: “here is a rate of progress that it seems you can handle and when you get to this performance level, it will be time to move on to the next phase” rather than “do this for 6 weeks”.
Nice talking to you!
Thanks TR for coming up with reply to the whole Pol discussion. I am a big fan of TR and have been using it for the past 2 years and love the podcast to bits.
I love that TR is taking the necessary/proactive steps in advancing cycle training with Adaptive Training and now a Polarized plan.
BUT, as a former academia, and the small dabble i had in research, i was expecting a different approach from TR about this matter. Starting the podcast by dissecting the very few decent research on this matter wasnt one of it. What about all the studies that have been quoted in the podcasts in the past? Are we going to dissect each and every one of them back and re-look at recommendations made in the past podcasts?
A simple, TR plans are NOT threshold because of etc etc. would have been sufficient.
I respect TR more now because they are brave enough to venture into research of the matter. Looking forward see what the results of the huge randomised control trials on this matter shows.
All of TR’s other plans have been developed and tweaked over the years, and have been proven to yield improvements.
A polarized training plan, which is structured, initially would be aimed at presumably would only target one type of event/riding style. So I reckon TR won’t release e. g. a polarized crit plan alongside a polarized rolling road race plan and a polarized century plan. That means it makes a lot of sense to initially funnel only people to it who know what polarized is and want to try it.
I am not sure whether there exist long-term studies that for polarized that are not observational. With this I mean not studies across one or a few training blocks, but studies that investigate the efficacy of polarized structured training plans. I think the studies on the tri plans lasted 13 weeks if memory serves, so that’s about half a block and for a very specific application where in competition you spend most of the time at 70-85 % of FTP.
I think choosing a plan with too high a volume, lack of prep, sleep and inadequate nutrition are likely quite high on the list. That’s quite independent of training methodology. At equal time a polarized plan would be easier, because it packs less TSS.
I love that you live extreme ownership, @Nate_Pearson. Thank you for this.