Process for Developing Efficient Programs for Individuals?

[RobertFontaine]

There are a lot of competing philosophies of programming for strength that all seem to work. There's a complete range of differing use of Max Effort, Dynamic Effort, and Repeated effort. Some use a lot of exercises, others only 5 or 6 core lifts. Some see little use for anything but heavy singles others program a vast array of percentages.

I'm starting to suspect that they can't all be right for a specific individual or even most individuals. In many cases I suspect that these programs select athletes who flourish under them rather than the other way around. While this might work relatively well for a national program it is at least sub-optimal if you do not fit whatever mold that they have selected.

.... which of course leads me to the greater question.

How does one develop and sustain optimal (or at least really productive) strength programs for individual athletes rather than cookie cutters?

So far basic novice and intermediate templates have done me as well as they seem to do for most every one else. How does a competent strength coach adapt/develop programs for athletes to optimize their development. All the books I have read to date either prescribe programs or describe adaptation mechanisms and exercises. Everyone seems to be able to describe their systems at the macro level but when you ask what is the process for individual application it seems to get pretty darn vague

Note that I'm not talking about sport specific training (sst) but rather core strength training. I also recognize that there are a host of strengths to select from; limit, starting, explosive // isometric, concentric, eccentric // relative, functional, absolute. Many texts go on about the relative merits and application of each type.

For any 1 specific type of strength, i.e. limit strength. How does one develop and optimize programming for an individual athlete?

[Scrofula]

I've been thinking about this same question recently. Being a machine learning guy, my solution is to try make a simple model of the athlete's recovery ability and fatigue, and try to infer the parameters of the model from his or her training logs. You could then optimize over the space of possible training programs and choose the one that maximizes strength.

That's a bit hand-wavey, but I have some initial thoughts on how exactly to do this. Unlikely to happen in the immediate future, but I hope to have this tool done by the time I'm at the stage where I need individualized programming.

[RobertFontaine]

So what can one measure quantitatively?

Maximum reps at percentages of 1RM.
Rate of improvement as a percentage of bodyweight.
Perceived Exhertion.
Work Load.
Program.
Diet.
Waking Heart Rate.

--------------------------------

I can state the target strength as Press, Bench, Squat, Dead

The problem is there are so many levers in the toolbox and most of them interact. How to test the variables independently is a real hairpuller and when you add in the elements of time and training level it gets uglier.

It's pretty easy to answer the "Are you getting stronger?" question.

It's fragin hard to answer "Of two productive intermediate strength oriented programs which is the more effective?" for a specific lifter.

i.e. Is Wendler's 531 4 day triumvirate more effective than a Bill Starr 5x5 (similar philosophy) or even nastier Doug Hepburn's (Program A) for an intermediate lifter? advanced lifter? in developing limit strength.

As you say, by looking at a structured training log and getting a sense of a trainees ability to handle workload, max efforts, etc you should be able to weed out programs that are going to crush you like a bug (ala Bulgarian or possibly Sheiko) or provide you with inadequate intensity and/or workload or are simply designed with other objectives in mind.

You should also be able to interpolate/extrapolate modification to workload/dynamic/max effort/repeated effort but it gets muddy really fast as they all interrelate

[Scrofula]

Most statistical models are designed for a nice, simple world where everything is independent, but there's been quite a bit of work recently on learning models in more realistic settings where things can interact in complex ways. Of course, the more complex interactions you model, the more data you need to learn a model that can generalize well. The trick is coming up with a model complex enough to capture the things you need to, but simple enough to learn from the available data.

For instance, in this case, I'm planning to assume that the diet and external factors like sports remain constant (or at least don't change in any non-random way), and take them out of the model. This gives a simpler, less realistic model that's easier to learn.

The nice thing about this problem is that it's possible to evaluate the learned model quite easily -- the model makes testable predictions. For example, it could predict your 1RM squat. You could then test your squat, producing very informative training data for your model. You end up with an exploration/exploitation tradeoff, where you need to choose between (1) producing good training data for your model (making it more accurate, yielding better programming in the future) and (2) using the most effective programming according to the existing model (making you stronger, but yielding less informative training data).

[Tom Campitelli]

So far basic novice and intermediate templates have done me as well as they seem to do for most every one else. How does a competent strength coach adapt/develop programs for athletes to optimize their development.

These two sentences seem to contradict one another. On one hand you say the novice and intermediate templates worked well for both you and many others. Then you ask, how does a strength coach develop programs that work well? I'm confused.

[TrackJunkie]

Someday, the same shit stops working. He wants to know how to develop new shit? Or he wants to know how this shit got started?

[LudwigVan]

Isn't this what Practical Programming is about?

[RobertFontaine]

New shit, better shit, compare shit.

Current method is find a program, do the program, get average results till failure, find another program.

This seems to be a very common path and it at best leads to average results over a fairly long time. Excellent results would be a lot more interesting.

... books like SuperTraining, Programming and Organization of Training, Science and Practice of Strength Training, Special Strength Training all touch on effective program development in different ways. I'm still wading through and hoping Supertraining has more of what I'm looking for. Some of the translations are horrid.

[tertius]

Most statistical models are designed for a nice, simple world where everything is independent, but there's been quite a bit of work recently on learning models in more realistic settings where things can interact in complex ways. Of course, the more complex interactions you model, the more data you need to learn a model that can generalize well. The trick is coming up with a model complex enough to capture the things you need to, but simple enough to learn from the available data.

For instance, in this case, I'm planning to assume that the diet and external factors like sports remain constant (or at least don't change in any non-random way), and take them out of the model. This gives a simpler, less realistic model that's easier to learn.

The nice thing about this problem is that it's possible to evaluate the learned model quite easily -- the model makes testable predictions. For example, it could predict your 1RM squat. You could then test your squat, producing very informative training data for your model. You end up with an exploration/exploitation tradeoff, where you need to choose between (1) producing good training data for your model (making it more accurate, yielding better programming in the future) and (2) using the most effective programming according to the existing model (making you stronger, but yielding less informative training data).

Multivariate analysis and modeling is all well and good, but this presumes you have a workable pool of meaningful sample data, yes?

I understand you're looking at the training log, but for what, precisely?
If a lifter is reaching the end of their novice progression, is any of that data going to be useful for making predictions at what sort of intermediate program suits them? Or are we talking picking an intermediate program as a data collection tool, for a certain period of time, and then attempting to model response to stimuli from that program? Would you then generate a model for various programs, such that given a standard set of data, you could predict an individuals response to programs x,y, and z?

This then begs the question: Do all intermediate programs tax all components you are interested in making predictions for equally?
Or do you need a "data collection" program that allows you to create a useful sample set (what I would do, probably)? This seems doable, and could make a very interesting research project or thesis, but data collection will be problematic. Be exceptionally useful, if you could pull it off.

Failing a system of models and so forth, I don't see a way around trying a bunch of different systems of programming over the course of many months, and carefully analyzing the response of the individual to the stimuli of each program, and then selecting a programming system based on their desired goals. This might not be necessary for an advanced trainee, or someone who has spent a long time at an intermediate level (long enough to have a large data pool available to review for information), but for novices just moving into being an intermediate, I'm not sure you can make many assumptions. But what do I know? I'm just some schmuck on the internet.

[Kyle Schuant]

You're assuming that the particular routine really matters a lot. It does matter, but more important is consistent effort over time. It doesn't matter how good or bad the routine is if they don't do it. Most people do not apply consistent effort over time.

In fact, one measure of a routine's usefulness is how easy is it to be consistent with. 10-15 exercises with varying reps schemes changed every 6-8 weeks - as commonly prescribed in gyms - is very difficult to be consistent with; something like SS, much easier.

Ensure the effort is consistent, and even a mediocre routine gets great results. That's the biggest thing. You can say, "well we can take that as given for the purposes of the model," but then your model excludes 90-99% of all gym-goers.

Next to that, whether it's 5 reps or 3 or 11, 3 total sets or 30, 1 minute's rest between sets or 5, these things are really not that important.