Quantifying training

So you're not "going the distance", you're "getting the work done" or some alternative cliche? :) You'll earn your free swim cap for reaching 50 MJ instead of 50 mi? I like it, as a way of rewarding quality over junk yards. Y Whether you meant it to or not, I think the combination of your energy points together with the impulse-response model is, in fact, a decent method of modeling swim training. Sufficiently simple to actually use (for me), unlike others that may be impractical or too complex.

Whether you meant it to or not, I think the combination of your energy points together with the impulse-response model is, in fact, a decent method of modeling swim training. Sufficiently simple to actually use (for me), unlike others that may be impractical or too complex. I truly wonder though if it would not benefit from a weighting factor RPE based, so that it automatically gets this individual intensity factor common to all other scoring system being processed through Trimp-like rolling avg functions. Such a simple RPE based weighting factor would therefore factor in Economy. If economy improves, energy to swim a given distance in a given time goes down. Q's actual scoring system can not capture this. And if it doesn't get captured, both CTL and ATL will artificially get boosted, as you progress into your season. Anyway, my :2cents:

If economy improves, energy to swim a given distance in a given time goes down. Q's actual scoring system can not capture this. Yes, but capturing a variable that represents economy is difficult. How much drag did experience as you swam down the pool? If that could be answered a really good model could be built.

I don't think it's RPE that's missing. It's the connection between power and (the magic biomarker indicating a training effect). Most models use lactate ion concentration as this biomarker. The reason TRIMPS don't use HR directly, bu use exp(HR), is because this is what correlates with lactate. The reason TSS uses a rolling average of power^4 is because (1) lactate correlates with power^4, not power, and (2) it takes a while for lactate to build up or get flushed away, so the rolling average helps determine an average lactate concentration. Personally, I'm not at all worried about missing the rolling average. Energy points and rolling-average-energy-points would be (exactly) proportional to one another, so that distinction disappears in the normalization. It might be significant whether you use something that scales with power, or something that scales with power^4 (= pseudolactate). Call those "pain points" instead of energy points. But as far as I know, nobody really knows for sure that lactate makes a better "magic biomarker of training effect" than anything else. You raise an interesting point about economy changing over time. But I suspect that the half-life for changes in swimming economy (the holy grail for most of us) is enough longer than the half-life for fitness decay that I can assume that all of my base was obtained at roughly constant economy.

Why Use a Model? Training models try to capture quantitatively several training principles what we all know from experience: Any training I do today will make me swim faster in the long run (perhaps a few weeks or months from now) but will actually make me swim slower tomorrow. That's why we pile on the training in the months before a focus event, and taper for a few weeks ahead of time. Working hard is more valuable than being lazy. So if we already know this, what's the point of using a mathematical model to tell us the same thing? Mainly so that we can optimize our training. When, exactly, should I start my build phase: 5 weeks out? 8 weeks out? How long should it last? Am I better off swimming 1500 yards of intervals today, or 3000 yards of straight swimming? How long should I taper? Et cetera. Having a mathematical model allows us to answer questions about which of two training approaches might result in more optimal preparation. Plus, it's a great excuse for some of us to get our geek on. Impulse-Response Models The basic concept behind all of these models is that you keep track of fitness and fatigue separately. Every workout will increase both fitness and fatigue. At first, fatigue increases by much more than fitness in response to any training. For example, suppose I swim a workout today that increases my fitness by 100 "points" (more on this below). That same workout will increase my fatigue by 200 points. Because of this larger increase in fatigue, I would swim slower in a race tomorrow than if I hadn't trained today. But the effects of training (the "response" to this training "impulse") don't last forever. Both fitness and fatigue decay over time as we become untrained. Luckily, fatigue decays a lot faster than fitness does. Roughly speaking, you can think of fatigue as having a half-life of 10 days or so, while fitness lasts longer, with a half-life of about 30 days. The next thing to recognize is that all of this is cumulative. My fitness, today, consists of that 100 points from the workout I did this morning, plus half of the the points from the workout I did 30 days ago (because its effect has decayed), plus some smaller fraction of what I did 3 months ago, etc. Fitness is the sum of all of these past "training impulses", each of which began to fade as soon as it was finished. Likewise for fatigue, which starts larger but fades faster. The last piece of the puzzle is to define a performance score as the difference between fitness and fatigue. Think of fatigue as merely masking the fitness you have built up. It's in there, you just can't tap it because of the accumulated fatigue. That's why a 1-2 week taper is typical -- it lets the fatigue fade away without waiting so long that fitness decays too much. This difference between fitness and fatigue is what you try to optimize on race day. The math involved is actually relatively simple (although I'll save the details for another post). You don't actually have to keep track of every workout you have ever done. You just track a cumulative fitness and fatigue score. Just like counting yards, but you're counting "points" instead. Then every day those scores fade a little bit, and you may add onto them if you do some training that day. Quantifying Intensity So how do you determine how many "points" a workout is worth? That's where things become trickier. There are a lot of different approaches. The key is that these point systems need to recognize that swimming hard is worth more than swimming slowly. Heart Rate The first of these impulse-response models was Banister's TRIMP (training impulse) model. In that model, your heart rate is used to calculate points, where points rise exponentially with heart rate. For example, I can earn 6 points per minute if I'm exercising at close to my max HR, but I only earn 1 point per minute if I'm floating along at 120 beats per minute. The big advantage of using heart rate to calculate points is that it can be used across many sports. A TRIMP earned in the pool is the same as a TRIMP earned running, or cycling. This makes it particularly valuable for multisport athletes, like triathletes, or for those who want to include cross-training in their training. It's also a physiologically meaningful indicator of training intensity, and there is lots of data correlating heart rate to other physiological variables like lactate levels, etc. There are two big disadvantages for swimmers, however. First, it's tricky to track heart rate while swimming. Heart-rate monitors are difficult to wear in the pool (for men, anyway) and are too much hassle for most swimmers in any case. Counting your pulse after a set is inaccurate at best, and only tells you your heart rate after you've stopped swimming. Secondly, heart rate rises fairly slowly in response to exercise. That's no problem for a 1-hour run, but can be a real problem for a 30-second swim. Zones Because of the various practical problems with using heart rate, a simpler approach is to use a collection of intensity zones. There can be as many or as few zones as you like. For example, one approach awards 1 point per minute for easy pace efforts, 2 points per minute for sub-threshold efforts, and 3 points per minute for efforts above threshold. Alternatively you can use an RPE scale (rate of perceived exertion) from 1-10 or 1-20 and assign different point values for each level. These zone-based point systems have the disadvantage of being subjective. They are also inconvenient because the calculation can't be automated via a heart rate monitor or power meter. For swimmers, it may be inconvenient to assign effort levels and durations to different parts of a workout that consisted of many sets & reps on different intervals. Rick Sharp's Stress Score One example of a zone-based method is Rick Sharp's training stress score. With this approach, each minute of training gets 0 points if it was part of warmup or recovery 2 points if it was part of an aerobic endurance set (long distance pace, short rest) 6 points if it was part of an anaerobic power set (80-90% max HR, 100-500 yard reps, moderate rest) 8 points if it was part of a lactate threshold set (50-200 yard reps, hard effort, lots of rest) 4 points if it was part of a sprint set ('t require HR or other data collection, but can still be somewhat subjective if your set doesn't fit cleanly into one of the predefined categories. This stress score is built into the Hy-Tek's Personal Swim Manager software. It also has the advantage of being transferable relatively easily to other sports, since it is based only on duration and effort levels that translate well into other sports Power The most common method of assigning points to different exercise intensities is to use power measurements. For cycling, in particular, power meters have become common training tools that can log the power output at frequent intervals duirng a workout. These power logs can then be used to calculate a point score for a workout. As with heart rtae, points are a very nonlinear function of power in an attempt to reward higher intensity appropriately. In Coggan's very popular TSS (training stress score) point system, 100 points corresponds to a 1-hour time trial done at 100% effort, although you can earn those same 100 points by doing intervals at higher power or longer rides at sub-threshold power. Power-based point systems have a number of advantages. As a measure of intensity, power is completely non-subjective, and can be correlated with physiological variables such as lactate ion concentration. The availability of power meters for bicycles means that data collection is automated and fairly simple. There are a number of different software programs and websites that can perform not only the point calculations, but also the impulse-response modeling of the training plan, and even the construction of a training plan. In principle, power is fairly transferable to different sports. Generating 200 W while swimming would earn you the same number of points as generating 200 W while cycling. Unfortunately, it is difficult to measure power output directly while swimming, and there are no commercial devices that swimmers can use to collect power data. Speed There is no equivalent of a power meter for swimming (or running), but it is easy to measure swimming speeds or swimming times. Because there is a known relationship between velocity and power, times can be used to estimate power, which can then be converted to points of some sort. Skiba's SwimScore is one such method, that performs a weighted average power over a workout using swimming velocities. It is a complex calculation (with several ad hoc assumptions), but there is software commercially available to perform the calculations. In this system, 100 points corresponds to a 3000 m time trial swum at 100% effort. There is a marginal attempt to keep these points comparable to those from Coggan's system (although this is flawed for anyone who swims 3000 m considerably faster or slower than 1 h.) On this forum, qbrain has posted another point system that uses velocity as a proxy for power. The intensity used to calculate his "energy points" is merely velocity cubed, which is propotional to power. This calculation is fairly simple: each rep earns a certain number of points based on the interval time, with no weighted averages required. For example, you'd earn 100 points for every swimming 100 SCY in 1:27, while swimming 100 SCY in 1:02 would earn you 200 points. These points are not easily transferable other sports, however.

On this forum, qbrain has posted another point system that uses velocity as a proxy for power. The intensity used to calculate his "energy points" is merely velocity cubed, which is propotional to power. This calculation is fairly simple: each rep earns a certain number of points based on the interval time, with no weighted averages required. For example, you'd earn 100 points for every swimming 100 SCY in 1:27, while swimming 100 SCY in 1:02 would earn you 200 points. These points are not easily transferable other sports, however. It is not proportional to power, but is proportional to energy. Unless you consider it acceptable for a time variable to be in the multiple. After going through this exercise, it is obvious that I suck at explaining my ideas. The two people who cared enough to read it didn't follow my explanation, but the concept is simpler than TRIMPS, which neither of you had any problem with. So I will shut up now.

Actually, I understood and chose my words very carefully. Your points are intensity * time, as with all of the methods. Your intensity (v^3) is proportional to power. Your points themselves (v^3 * time) are proportional to energy. By somewhat clumsily saying that the "intensity used to calculate energy points" is proportional to power, I was trying to point out the valuable connection to power -- which Solar and others prefer -- while still describing the method correctly (I hope). I understood your explanation in the other thread very well, and actually like your energy points much better than anything else for swimming. I'm planning to start using them instead of the approach I currently use.

After going through this exercise, it is obvious that I suck at explaining my ideas. Hope you're not issuing this after having seen me failing to implement it the right way. I'm a slow mo learner. It is not proportional to power, but is proportional to energy. Isn't there a relation between power and energy anyway? So I will shut up now. Please don't. In order to fix this relative intensity factor (or lack of), I am wondering if an RPE based weighting factor could do the trick. @sjstuart : Your post is great. Thanks I understood your explanation in the other thread very well, and actually like your energy points much better than anything else for swimming. I'm planning to start using them instead of the approach I currently use. Me too. Not in place of but certainly parallel to.

Your points are intensity * time, as with all of the methods. Your intensity (v^3) is proportional to power. Your points themselves (v^3 * time) are proportional to energy. Obviously you understand very well, and I misunderstood you. The is nothing novel about "energy points", it is just a simplification of the other methods to a point that it is easy to implement while still being meaningful and I wanted something that could compete with GTD. I think you understand that as well. I am certainly not trying to come up with a better, more accurate model of effective swim training.

Yes, but capturing a variable that represents economy is difficult. How much drag did experience as you swam down the pool? If that could be answered a really good model could be built. Usually, RPE based weighting factor are used to achieve exactly this. Not as precise as HR some might say, but much more flexible though. The Excel file attached (sorry it's in Spanish) shows you a very simple example of this. I think it just goes duration*rpe. Therefore by doing so, you transfer the responsibility of fair guessing an Economy/Taxing indicator down to the athlete. I don't think it's RPE that's missing. It's the connection between power and (the magic biomarker indicating a training effect). Most models use lactate ion concentration as this biomarker. Yes and some models use RPE to act as this marker. But anyway, I buy your simplicity principle as is. No need to make all models the same after all. You raise an interesting point about economy changing over time. But I suspect that the half-life for changes in swimming economy (the holy grail for most of us) is enough longer than the half-life for fitness decay that I can assume that all of my base was obtained at roughly constant economy. Food for thoughts I admit.