Accelerometry as a Means of Quantifying Work, Workouts, and the Training Plan by Joel Stager, Ph.D. (2006)


Published


Introduction: Joel Stager is the professor of kinesiology and the Director of the Counsilman Center for Science of Swimming at Indiana University. He received his PhD in medical science from Indiana in 1980,and is an active member of the USA Swimming Sport Science Committee. He is the recipient of many grants for swimming research, including today’s talk, “ Accelerometry As A Means Of Quantifying Work, Workouts And The Training Plan”. From Indiana University please welcome, Dr. Joel Stager.

Dr. Stager: The first thing I need to do, is reinforce to you that despite the fact that I am a professor and direct this center it is a team effort. I have had a lot of fun in the last four or five years “recruiting” talented academic athletes to help us with what we are doing and as you can see, I have about a dozen students now that I work with who are engaged in some form of research that is partnered with swimming. Things do not always start out or end up where you plan them to be and this is an example of that sort of track or map. We didn’t start out looking at this particular question. We actually had a grant from United States Masters Swimming to explore the relationship between energy expenditure or how many calories are expended swimming. It turns out if you look in the literature, a lot of data that is out there is very, very old and it was all done on about a half a dozen swimmers. When I mean old, I am talking about in the early 1920’s. So if someone were to ask you how many calories am I going to expend if I swim a thousand yards, you almost can’t answer that question and that is actually what we were looking at. It turns out that from the CDC’s perspective and the World Health Organization, a lot of people do swim recreationally and so if they are plotting out caloric balances they cannot do so if they spend their daily recreational time in the water and this illustrates that point.

If you look at running, there is a fairly straight line relationship between the velocity of running and the energy expenditure, as you can see by looking at the Y axis up there in the corner on the left. It is pretty much a straight line and everybody falls within 5% or 10% of that curve. We don’t differ a lot in terms of efficiency when you are talking about running velocity. Now, you can split hairs if you are looking at high level performers which may be important in terms of economy running, but if we flip over to swimming and I have six or seven examples of energy expenditure while swimming at various velocities; each of those lines represents one individual’s curve. You can see already, we are not talking about 5% or 10%, we are talking about maybe 400% or 500% differences in instantaneous energy expenditure alright?

So, that actually tells you where we were coming from. We were trying to find out whether or not we could identify a device that would allow someone to actually quantify the number of calories they were expending if they went to the pool and swam a thousand, two thousand, or three thousand, whatever that may be. It just so happened, one day I convinced the swim coach at IU to allow me to put these devices on a number of her swimmers. When we got back from the pool and plotted the information out, this is what it looked like. The graph shows “N” athlete wearing an accelerometer throughout a morning or afternoon practice. Each one of these profiles represents output from one of our accelerometers and I will talk a little bit more about the accelerometer in a second, but you can walk right through each of these sets and identify what the athlete was doing. I am again referring to this as the serendipity of science; we never intended to look at this particular aspect, but it came to us that we could do a lot with this information. I will show you where this led us. You will see this graph again.

A new direction. I happened to go back in the literature to see if this has been a significant problem for coaches and the answer was obviously, it has. The question becomes, “how do I know how far or how intense or at what velocities my swimmer should actually be training at” and “how do we quantify training load?” The earliest quote that I could find was from 1894. They started asking this question and again, I just picked one from Doc Counsilman in 1968. Basically it is the same question, right? The hardest part of coaching is knowing how far they should go or when do we quit and how do you quantify that? So, here is the conundrum. I think we will accept the fact that if training is the stimulus for improvement, then the stimulus must be appropriate to enlist the intended and desired outcome, right? The point is, what you are all about is improving an athlete’s performance and the stimulus for that performance is the training bouts that you apply to the athlete. However, how do you know what training bouts to apply if you cannot characterize what it is you are applying? How do we decide what the appropriate stimulus is if we cannot quantify the qualities associated with it? Or even the magnitude of the stimulus that we are applying?

So that is really the new question. Again, measuring the response is easy. The response is measured with a stop watch, with a timing system. They either got faster or they didn’t, right? Quantifying and characterizing the nature of the stimulus is something completely different. So, for example, if I were to ask you, “what was your teams training volume this week?” what would your answer be? Almost everybody says, “we went 45,000 yards last week”, right? The units you use to communicate are yardage or meterage. In the US we use yardage, but the rest of the world uses meterage and the units of measure would be meters or distance, alright? But remember, the word I used there was volume.

What is training volume? You use that word all the time and I am going to tell you that volume is three-dimensional. It has a length and a height and a width. Yardage is one-dimensional. How far did you swim? That doesn’t really tell us everything that we need to know about the characterization of the stimulus. We have tried to use the word “training load” because that is really what you want to know. If you want to know three dimensions, you want to know the distance, the velocity, and intensity, right? Also, the load and the volume might be equivalent if we have all three of those qualities. Load addresses the physiological impact. What did I do in this particular set, as well as what was accumulated? What was the impact this week? What was it last week? How does this seasonal progression? How do I develop micro cycles for example or meso cycles?

Are there ways of doing this? Most of you have probably explored this and I want to say that this is the transition between science and the art of coaching. We understand the paradigm of heart rate, although how many actually use heart rate on a daily basis? How sensitive is heart rate? There are all sorts of situations involving heart rate. Perceived exertion – you do this all the time, so how do you feel? How did that set feel? You ask that question all the time. Some of you require that athletes put together training logs and then once in a while you go through the logs and read them to try to assess what is going on. Most coaches put together an annual plan, but there is a problem with that and then of course, we have been searching for appropriate biological markers for the last fifty years. Lactic acid, immune responses, sleeping patterns and so forth – to try and quantify the response to the stimulus that you are applying, right? – So that you can assess training load.

Problems? Well we have talked about some of them. Many of these techniques are invasive. It is not fun to finish a series of seven 200’s and have your finger pricked to take a blood sample. Some of them are subjective. How do you feel? “Well coach, I don’t feel real well.” “Why not?” “I just broke up with my girlfriend.” Well that doesn’t have much to do with what you are doing in the pool. Some of them are imprecise as we already know, and then I will argue that heart rate is pretty imprecise. There is a wide range of noise associated. There is limited validity; difficult to interpret; complex; expensive, and so forth. So we have this gismo and we are going to ask the question, Can quantifying movement be as simple as that? Can we use movement as a marker of training load? Can we partition these movements? Can we summate these movements over time and can that tell us anything?

I think about three years ago, I came across this device. This is an accelerometer. It is about the size of a postage stamp so it is not that big. You don’t have to strap it on a swimmer like a backpack. It is smaller than a stop watch and it is also waterproof. Alright, what does it do? Well, it measures acceleration. It measures acceleration and acceleration is a change in velocity, simple as that. How does it do this? That is a complicated explanation. Basically, it measures acceleration virtually in every dimension except for one and that is probably acceptable for us. What we have to do is align this so that it measures what we want it to measure most sensitively. You can also set what I will call its duty cycle, which measures the number of accelerations. For example, fifteen seconds, a half minute, or a minute. The accelerometers that we are using right now can actually be set down to a duty cycle that is two seconds long. As a result, we can do some pretty amazing stuff with this.

Now we generate some hypotheses. There must be some relationship between movement and training load, limb or body acceleration; positive or negative as related to training load. That is basically what we are testing. As an athlete moves their arm or leg a mass attached inside the acceleration creates a force against the ceramic blade that generates the voltage. Voltage corresponds to a change in acceleration because amplified filters sample the various lengths of time. Basically what we are going to do now is break this down into those elements that we talked about before which are theoretically important to you. I know a lot of coaches who have these very intricate training plans, but basically at the end of every week he will plot out the percentage of his practices that were in the E1 zone or perhaps the percentage of his practices that were in the A2 zone. And he will do this every week throughout the year. How do you really know they did that? That is a huge assumption, right? That is a very big assumption. We are going to be able to look at that, too. These are our components of our hypotheses that we can look at; arm and leg contributions to the training load. We can look at seasonal changes. We might be able to look at intensity of effort and we might be able to look at individual efforts.

Now, let’s see if we can do it. Well, this is good news. This is the first thing that we did. We took a group of swimmers and asked them to swim set distances: 800, 400, 200, 100, and 50. We just accumulated the accelerometer counts. Well again, for those of you that understand statistics, when you look at the R square over there it is almost to the point where you cannot believe that number. It is basically perfect. 1.0 is perfect, but that is a phenomenally high value so it looks like it is curvilinear, but it is not. It is only because we didn’t plot it on a log scale. We are going 50 to 100 to 200 to 400 to 800 so it looks like it is curvilinear, but that straight line is remarkable so this is really good news.

All we are asking now, is whether or not the accelerometer is sensitive enough to quantify the distance swum. The answer is yes. So, this falls off a little bit. Why does this fall off compared to the last one? The last one was simply freestyle and they were swimming a prescribed set. This group was involved in a workout. Can we still see, let’s say yardage? The number is still high, .71 for an R square is still high. It is not as good as it was, but, now we have drills taking place. We have different proportions of kicking sets, pulling sets, drills, freestyle and breaststroke and all those things are mixed in so actually that is a pretty good number. I can’t tell you how many different athletes there are, but we have different athletes, we have got different practice distances, and we have got different strokes, but it is still a reasonable number. Let’s see if we eliminate different people in this. Let’s just put it on one athlete and have him go through, in this case 23 different practice sessions. The number improves as we might expect. Now we are up around .8 which again is a high number. That is pretty good and we are just scratching the surface. We will summarize so far. For an individual swimmer, there is nearly a perfect relationship between distance and accelerometer output so we can almost self-calibrate. In other words, we can do a series of distances with an athlete, get an algorithm and basically say (virtually perfectly) we can predict how far they swam in a given bout. We can actually do that as well for practice. Even when we pull the whole team in to it, we can actually do that and all we are going to be doing is saying – well, how far did they swim? And you kind of go well – I know how far they swam because I told them how far they were going to swim – well maybe.

Let’s talk about velocity. It is not necessarily just how far you swam, but how fast you were swimming for that distance. It turns out that we get very close to the same relationship. What you are seeing is the velocity at which they swam these races is also very highly correlated to the accelerometer output. And we have it plotted exactly the same way. Now, I should probably step back and say there are all sorts of other questions. For example, “where do you put the accelerometer?” That is not a trivial question, so you have to test all possibilties. You could strap it around their head, neck, right hand, left hand, right leg, left leg, hip right side, or hip left side. We tested all those things simultaneously to see whether or not it made any difference. Basically, what we were looking for was the simplest possible result and it turned out to be the right arm and right leg. Again, we have combined arm and leg outputs, but it is right arm and right leg which did not really make any difference regardless of the swimmer you looked at.

You are going to see some really high numbers here so every time athlete X swam a 100, we took an accelerometer output. Let’s see how tight that is, regardless of their velocity. Well, very tight, again, .91 something. That is good news so now we know we can almost perfectly predict after the fact how far they swam, but we can also tell you how fast they swam for that distance Here is something else of interest. Look at men versus women. We had two men and two women who all swam a set of seven 200’s. it turns out it is the same line, right? Actually, as it should be. The women are simply swimming at slower velocities and the yellow part of that particular line is the overlap. So, it is the same relationship and again the correlation .94 basically says that this works equally well for men as it does for women. By just looking at limb movements we see that the women are not doing anything different than the guys. Now this is a little bit different. This is what happens if you put everybody on the team on the same chart? You start seeing the relationship decay a little bit because of those original curves I showed you. People do not sit on the same curve. They do things differently and yet, it is still a fairly high correlation there.

Alright, now we can start looking at some other aspects of this. Here is this standard set of seven 200’s and the question is, “how do swimmers swim faster?” You have several choices. You can increase your tempo, you can increase your distance per stroke, or you can increase your kicking effort. The question is whether or not these devices are sensitive enough to pick this up. What you see here is that at least in this range of velocity that we are looking at between 1.2 meters per second and 1.5. There is a small increase in the accelerometer output of the arms. The majority of the difference comes in an increase in kicking, which is something that we were sort of surprised at and as we looked at this further, it turns out that there may be different strategies. We don’t see this in the women as much as we see it in the men. The women tend to take the strategy of increasing their tempo. The men tend to take the strategy of increasing their kicking contributions.

As a result of a conversation with USA Swimming, we looked at identifying differences as the season progressed. We hadn’t thought about dividing it up into early season and post season. That is what this one shows. This is the same athlete at two points in the season and basically what it says is that at any given velocity, accelerometer count goes down as the athlete starts to rest. So what does that mean? Well, it is what we were talking about before. Apparently in this athlete what he is doing is increasing the distance per stroke; not necessarily increasing tempo. I believe the late season graph was three days after his taper meet. The point is, yes, this technology is sensitive enough to be able to see differences throughout the season. So, I am going to summarize again. We earlier established that there is a very tight relationship between distance and accelerometer output, accumulated output, however you want to do it. Now we are saying that there is also a very tight relationship between accelerometer counts and velocity.

Now, the units here are different. In one case it is just accelerometer counts per distance. Accumulated, however, is accelerometer counts per second. In order to estimate velocity, we have to convert it to an intensity value (if you will) so it is counts per second. Then we have these tight relationships with velocity. Relationship does not suffer much when you change strokes and/or if you add swimmers. unless you start talking about differences between males and females. I should probably also note that the group that I am talking to here is interested in competitive swimmers. If we start talking about recreational swimmers you can throw all this away. The differences amongst recreational swimmers are so great that this is almost worthless, but what happens is as swimmers become more experienced and I will use the word “efficient”, they become more alike rather than more different. Leg and arm work can be partitioned. Seasonal changes are evident. Men and women are similar, but not identical, at least in the freestyle lines for this set of seven 200’s.

Now, a lot of nice colors: red, white, and blue. It just so happens once again, if you put these on individuals and you send them into practice (after the fact), you can plot differences between swimmers. Now, this one (these are both female swimmers), is just the legs, but you can see there is a considerable difference. It looks to me like the blue swimmer, if we wanted to use the work intensity and if we want to talk about velocity, is putting a little more intensity in and a lot more velocity. Or, the other way around, a lot more velocity and a little more intensity than swimmer two. That would be our conclusion from this. You can look at the four 200’s up there were increasing efforts descending set and then there was a 200 all out and you can see the red swimmer either is amazingly efficient or didn’t put a whole lot into it. The next set was a set of four 100’s and then there was a 100 free all out and the same sort of conclusion. I would say the red swimmer was blowing bubbles and the blue swimmer was putting in some effort, but as a result of where we are, with the algorithms that we have developed, we can tell you what the velocity was for each of those swims which is pretty interesting. That one was just legs, this one is combined arms and legs on the same two swimmers during the same workout and it is basically the same situation. The blue swimmer (swimmer 1) maybe was a little bit more honest with the coaching staff than the red swimmer (swimmer 2) and we are now getting into “big brother is watching”. Iif you think about the things you can do with this, it is like, “well wait a minute here”.

Conclusions: (and then we will talk about it just a little bit more) Accelerometers can be used to quantify training velocity and training distance. Accelerometers can also be used to access these impressions. Accelerometers combine with markers of physiological analysis to quantify training volume, which is now three dimensions. So, how can you use this? Let’s suppose you are a college coach and your athletes are going home for Christmas vacation and then they are going to come back on a training trip and you have a series of eight workouts that they are supposed to perform during that ten days. Did they or didn’t they do them? Well, you can now answer that question because you can’t lie on this one. One of the swimmers says well what if I put this in the washing machine? Would that work? Could I do that? The answers are “no”, you can’t. We actually even took it in the lab and had these little things we call vortex spinners to see if we could fake somebody out and “no” you can’t do it. Therefore, it is true, you could use this to determine athlete compliance, did they do the workout, because like I say, you are going to get a profile. You can look at the profile. You can tell if they did the four 200’s or not. You can basically tell exactly how many yards they swam and at what velocity. The second one you can measure is their athlete effort. Basically you can measure this for every practice. You can store it on your computer in graphic form and you can basically show well here is the practice we did three weeks ago and here is the practice we did today, we have a problem, right? The last information to access with this is he can actually determine whether or not this training stimulus was what he had originally intended it to be. So I say, my friend that has these plots and he has every percentage known to man calculated out, he can calculate where they are in their zones. So, what is next? Where are we?

I could have entitled this “Back to the Future”. I don’t know anybody that can do this right now. We are probably not far enough along to be able to tell you what you need to do in order to go out and buy some of these things. There are other things that we need to establish. For example, right now what we are doing is every time we put one of these on an athlete we are having them wear a heart rate monitor. The goal, of course, is to look at the relationship between movement and heart rate to see whether or not our view of intensity is the same thing that coaches understand the heart rate Paradigm. Let’s break it into zones and let’s determine whether that is really reflective. We have the wherewithal in equipment to actually look at VO2 at any given set by analyzing after any given set. We can run seven 200’s with increasing intensity to look at the VO2 and the heart rate and the accelerometer outputs so this, where we are right now, looking at that relationship. Again, a coach could say well, okay, 38% of my practice last week was in the V2 zone, well maybe. Okay, we can compare accelerometer to paradigm currently used and understood by coaches i.e. training zones and that is where we are headed. We currently use accelerometers with a duty cycle of fifteen seconds. It turns out that we have one that has a duty cycle of two seconds. It is technically possible to program these with a two second duty cycle. We believe and we have some data to support this, that you can actually do race analysis with these accelerometers because the duty cycle is so fast we can tell essentially velocity every two seconds. We can see flip turns within a race and we can see when the swimmers are slowing down or speeding up because it still comes down to tempo, distance per stroke, and kicking effort (things like that). We can do this without the use of video tapes and all it takes is to take the accelerometer off and put it on a reader and hit a button; that is how hard it is.

Alright, thanks for the invitation, thanks for the opportunity, stay tuned. Next time, it is “Back to the Future” continued. There are probably some questions you have, but I don’t want to answer them. I can tell you ins and outs of this. One is that somebody is going to ask, well, how much are these? They are expensive. Unfortunately, they are not expensive to manufacture and that is not what you are paying for. You are not paying for the manufacturing, you are paying for the software to interpret the output, so somewhere I read these cost about $14.00 to manufacture. Now, the manufacturer may argue with me on that one, but what are they selling them for? What do you think $28.00? No, each one of these is $450.00. ($450!!) The point is, the technology to build these things is not that expensive; it is just that right now, it is me and twelve of my grad students playing with these things. They will come down in price. Again, it is not a manufacturing cost, it is the software deal that you are paying for their development. Absolutely, if someone would let us put them on the wrist and the ankle, but like I say, they are non-obtrusive. They are about the size of a postage stamp. The swimmers do not complain at all. Right now we do. We take readings every fifteen seconds and it is accumulated. We are trying to bring it down to two seconds and then you can actually look at the inside of a race. We think two seconds is probably fast enough to tell us what is going on in the 100 meter freestyle for example.

Q. How are you attaching the monitor now? Well again, that is just like everybody who has ever tried a heart rate monitor knows that in theory you can wear them. In theory, we just have a little Velcro strap like a wrist watch and put it on and what we do is we either put “cellophane”, or there is other stuff that you can just wrap around there to keep it from sliding. The problem we originally ran into is the orientation of this thing slips so it comes on the inside of the wrist that changes what you are measuring so we just put it on like a wrist watch and put a little piece of tape around it and it keeps it from spinning.

Q. how much do you do? I mean how long can you accumulate readings to get this two hour factor? He is asking how long one can accumulate the information. Well, it depends upon your program timing. If you set it up for every two seconds and your window collapses because you are collecting more data and it only has a finite storage space, you can go to a fifteen seconds every twenty four hours. If you go to once every minute or every five minutes you could go for a week, but you wouldn’t get the same amount of information. Again, it takes ten seconds to download the data on the computer, it is simple.

Q. No you can’t ask that question. That question is in your hands. I mean, in other words, I am going to get you the number, but you are going to have to interpret them from the standpoint of the training stimulus because ultimately you are going to have to mess around with it and decide was this effective or not? I mean that is the question you are always asking. You know, one of my good friends who is a coach says, “I didn’t really worry about it when my swimmers don’t swim well” he said “that happens a lot”. “What I worry about is when my swimmers swim really, really well and I can’t explain that and so of course I spend hours going back trying to figure out what I did that was effective,” but again it is just a flip side. You are trying to figure out why they didn’t swim well and why they did swim well so this is just another piece of information that seems to correlate with all these things that we have looked at.

Q. Way back there (Q) alright,well, do you know what? Does everybody know what dental dam is? It is like a big sheet of latex. Now imagine if you put a sheet of latex in your car, right about head level and imbedded your head into that sheet of latex. What happens when somebody steps on the gas? Your head goes back like that. It is reacting to the acceleration. What happens if they step on the brakes? You go like that. Well, what it is doing is it is deforming that sheet of latex and the latex is essentially 2-dimensional. It is so thin that it is essentially, well, that is essentially what is going on here. This little tiny wafer in there is, what do you call it?… “Piezoelectric” and when you deform it, it provides a current that you can measure so she is asking me, what is it that you are measuring? I can’t tell you that. All I can tell you is that this thing is called omni-directional, like I said, if we are back in the car with a sheet of latex the only plane that you don’t really move a lot, if you were doing this, would be this way. So that would be sort of the insensitive plane that this thing couldn’t see. Well it turns out the way we set it up, the insensitive plane is actually, strangely enough, the forward velocity plane. It is as if you put a rod through the swimmer and say well I can’t see that that direction, but guess what? Your hands are not moving in that direction, strangely enough, they are moving this way or this way or whatever and it turns out we can see that so I can’t tell you whether it is measuring the recovery stroke or it is measuring the in-sweep or it is probably measuring all those things, because it measures it wherever or whatever plane or whatever arrow of accelerometry since it can measure all that. (other than that one right there).

Q. Yes, it is definitely different across the four strokes. Now, we haven’t done that particular experiment. We haven’t had twenty athletes swim 800, 400, 200 butterfly. If you can get your athletes to do that, we will come over, well yeah, backstroke maybe, breaststroke maybe if you want them to walk again, but we haven’t done that. I mean there are a lot of things and that is a great example. One of the reasons that I enjoy talking to you guys is because you give me more ideas so that is what we need to do.

Q. Okay? I have gotten a lot of interest from other countries about this. For example we had a Chinese delegation at IU a couple of months ago and they all act like “oh yeah”, we have been there, we have done that, ahhh you guys are so out and then they saw the accelerometers and it was like, “okay” come over here, we are going to take pictures of this, so they were very, very, very interested. This was just something that they hadn’t thought about. The Australians have talked to me about this. The Belgians have talked to me about this and a lot of people now are. The Japanese are all over this stuff so there are a lot of people out there who see this as a possibility and it will become cheaper, eventually.

Sponsorship & Partnerships

Official Sponsors and Partners of the American Swimming Coaches Association

Join Our Mailing List

Subscribe and get the latest Swimming Coach news