Fast Talk Labs Advanced Performance Data Analysis Pathway Badge

The Data Revolution—How A.I. and Machine Learning Will Make You Faster

Tim Cusick joins us to talk all about data and the data revolution we are experiencing.

Mathieu van der Poel power data Tour of Flanders 2021

This episode is all about data. Our guest is the lead engineer for TrainingPeaks and their coach-focused software package WKO4, Tim Cusick. We’ll also hear from Armando Mastracci, developer of Xert Training software; Dean Golich, a head coach at Carmichael Training Systems; and finally, we’ll touch base with Joe Dombrowski, of the EF Education First-Drapac WorldTour team, to get his take on how pros are reacting to the data revolution.

Episode Transcript

Chris Case 00:00
Welcome to Fast Talk the velonews podcast and everything you need to know to ride like a pro. Hello, and welcome to another episode of Fast Talk. I’m Chris Case managing editor of velonews joined as always, by my distinguished co host, Coach Trevor Connor. Data. Not long ago, people looked at you funny if you had a two inch screen mounted to your handlebars, now, we ride with head units, the size of cell phones, Sensors connected to our limbs, and wearables that track every step and heartbeat, and no one bites an eyelash. A few episodes ago, we talked with Hunter Allen about the history of power and how he got to this point. Today we’re asking this question, where’s all this data going? And what do we gain by covering our bodies in sensors like something out of a Star Trek Borg episode? And yes, Trevor made me use that reference, I have no idea what it means. In this episode, we’ll talk about first the data revolution, there’s been exponential growth in the amount of information we’re now able to collect and analyze. That data is allowing us to look at our training in ways we never could before, but it also comes with some dangers. Next, we’ll discuss the rise of artificial intelligence and machine learning in training software. That’s a fancy way of saying that the software no longer just tells you how far you rode, and what your average power was. Increasingly, it is telling you what your items mean, where your form is at and gives clues to what you should be doing. Then we’ll address the three stages toward machine learning. First, there is the descriptive, how far you rode, how many hours you’ve trained this month, and so forth. The second where we’re at now is predictive, software crunches your data to predict what’s going to happen to your form. Finally, in a few years, we hope to move into the prescriptive stage where the software start telling us what we should do. Finally, since these changes are going to have a significant impact on both athletes and coaches will address what they each can expect. Should coaches start refreshing their resumes perhaps our guest is the lead engineer for training peaks, and their coach focus software package WKO4 Tim Cusick. He’s been coaching for 18 years, including some incredible athletes like Amber Neben, who is a multiple time national and world champion. He comes from the world of data analytics, which gives him a unique perspective on training science. He’s actually been working in artificial intelligence and machine learning since the 1990s. We’ll also hear from Armando Mastrachey, the developer of exert training software that grew out of Armando’s own experiences as an engineer and crunching large amounts of data to find trends. Dean Gulledge, a head coach at Carmichael Training Systems will share his thoughts on where the software is headed. And finally, we’ll touch base with Joe Dombrowski of the education first drapac World Tour team to get his take on how pros are reacting to the data revolution. With that, let’s make you fast, and hopefully Live long and prosper. So maybe maybe we’ll start at the data revolution, Tim, and maybe you could tell us about the breadth and depth and just the immense quantity of data that people have at their fingertips these days, compared to even a decade ago or two decades ago. You have

Tim Cusick on Data Revolution

Tim Cusick 03:45
to start at kind of the point we’re at today, when you start talking about data, right? The point we’re at today is you have a modern athlete, I like to call it what’s different with a modern athlete and with the athletes of the past is that modern athlete is now a data gathering machine. I mean, if you look at the explosion in the last really five to 10 years in wearables, data gathering devices, analysis programs, it really is this idea that there’s a data revolution going on the challenge of the data revolution with that I’ve seen in the years that I’ve been involved in it, I’ve had the luxury of looking at it as a coach as an endurance coach is, first we start with the data. So we didn’t start with a solution or a hypothesis. Devices actually matured before we know or knew what to do with the data. So the revolution really has been an athlete and a device driven revolution. We have a massive pool of data. That revolution now is evolving into maybe phase two of the revolution. And that phase two is what are we going to do with that data. Now that we have it and we’ve begun to learn with it and we’re putting it into a really cognitive approach to Improving training, reducing injuries, improving single athlete performance improving team athlete performance. But that’s the cutting edge of the revolution. So it went from data device to more and more athletes gathering data. And only now are you at the beginning where people are saying, what can we really, really deal with all of this data

Chris Case 05:20
yet, given that we have all of this data to work with now, and not everybody has all the tools necessary to deal with it and to interpret it into to manage it? What are some of the weaknesses of of the data revolution?

Tim Cusick 05:35
So when you look at the weaknesses of the data resolution as a whole, the problem is people who don’t know how to utilize all of the available data, and I shouldn’t say people, and maybe it is, you know, individuals or companies or whatever it is, you can really create what I call the data haze. Meaning if you selectively bang away at a big bunch of data, looking for a solution, eventually, you can actually find that solution if you change the database, do segmentation, don’t keep data integrity, you can validate that. And I think we’re already seeing that, to some degree, I think everybody should look at each one of these systems with a grain of salt. This is a very new field for sports as a whole. And for endurance sports, I think what you need to be weery of is that people are accurately using the data. Because if they’re not accurately using the data, you’re kind of creating a fad versus physiology. I’ve seen systems validating things, I refer to it like a fad diet, you know, if I just stop eating this, I’ll lose 40 pounds, the reality is, data gives more. And the amounts of data we have now do give more opportunity for mis interpretation of an mis utilization of that data.

Trevor Connor 06:47
So I was going to refer to the way I think this is cherry picking and it sounds like that’s what your talking about, when you have a lot of data, people are going to be able to pick the information, they want to show them what they want to see. So the classic example that I keep having to correct people on is somebody wants to figure out their FTP, and you know, their real FTP might be 250 watts, but they don’t like that they want it to be 280. So they they find the data that that confirms what they want to see.

Tim Cusick 07:19
I couldn’t agree more, I’ll tell you that, in the time for me, since we launched the power duration model and WK04 to sitting here today. I probably had that argument more than 1000 times. And it’s just true. And part of why we did the power duration curve. And what we’ve done in our side of analytics and thinking was we wanted to give somebody a good a well modeled a accurate process, to at least sit back and compare. And if you see big differences in what the data is telling you, of course, we leave that to the individual to interpret why. You know, and and maybe the Why is you just want it to be a big number or maybe the why you have a difference because you don’t understand something. But the power duration curve is an excellent example of how to use data to give you a tool to make sure you’re not introducing bias or cherry picking or using those small things to be as truth or you’re validating everything with as truth without some background.

Chris Case 08:14
Could you could you give an example of a bias in this instance?

Tim Cusick 08:20
Alright, but I’ll give you the honest answer. When we develop the power duration curve. I’m obviously talking about something in WK04 in our modeled science, there’s other models out there that do what we do. And I encourage people to look at different models, they do slightly different things to understand what each one does. And modeling has gone back really since the banister model. We’re not talking about something new. But I think we’re talking about something that’s pretty accurate. Obviously, I have that bias, but I do believe it. Now, for example is one of the things the power duration model has forced people to do is to rebook at threshold at FTP, you know exactly the example you’re just talking about. When we launched it, we’ve literally got hate mail, like my FTP is 20 watts higher than your stupid model, says and we would be like okay, so now walk me through your data, what’s been your peak, 20 minutes, 40 minutes, 60 minutes, you know, and what’s maybe your 60 minute normalized power, let’s look at different things that might give us insight. And time and time again. And I mean, literally more than 95% of the time, the data, the individual performance data really wasn’t even there to support what that individual believed their FTP was. So here’s where you have a data tool, right? A big something based on big data that gives insight into physiological performance. That is a very good tool that you can sort of vet your opinion, how fit you think you are, versus how fit the data says you are. The purpose of this type of data usage is it gives you at least something to look at to give you a good mathematical relationship of the performing human physiology what’s happening inside of me based on the data that I’m producing? And then it makes you think and question. And if you question and say, No, no, I’m right, that’s fine. And if you say question, you question yourself and say, No, no, this model is right. Actually, I would say that’s probably better yet to be honest with you based on the number of mistakes I’ve seen people make. But that’s a choice you make.

Armando on Exert

Trevor Connor 10:22
One of those other training software packages is Exert newcomer in the field developed by Armando Mastracchio, and Armando I apologize, my Italian is horrible. Armando is a software engineer and a recreational cyclist was fascinated by the data and decided to start looking for the trends, eventually developing a robust package, he shares his thoughts about this revolution means what he thinks riderso9 such as himself are looking for.

Armando Mastracci 10:46
What I think is amazing terms of what’s happening today is that we’re seeing a greater proliferation of athletes now training with power. Now, with the price of power meters starting to decline, as well as the interest in in winter cycling. So we’re seeing a growth of individuals Now, using trainers indoors, as well as you know, these virtual environments, like zwift, which are becoming extremely popular, and all of these are based upon power. So all these market forces are really bringing a lot more individuals into the market in terms of wanting to train with power and what’s, what’s different about these is that they’re not necessarily really competitive athletes, they’re not necessarily professional or, you know, looking to be the top racers in the country, a lot of these are very much recreational cyclists, right. So they’re just, they want to use power just to stay fit, right, they want to use power to kind of track their progress. That’s different, right? When you when you’re looking at what software needs to do, right? So software has always been characterized by Oh, and how do we use the tool so that we can better analyze all the data for these top athletes. So usually in the hands of either sports scientists, or really the enthusiests, who are really into examining the data, those were the original kind of targets for the use of power data and for interpreting it, but now that it has become a started the greater proliferation of power data and power meters, we’ve got to start to look at how do we provide information about their own fitness and their own progress within the software without having to dive into all of the analytics per say that are currently available. So this is kind of what we focus on within our software is how do we make the application and understanding of the power data something that’s more accessible to a broader range of people. So as an example, individuals who, you know, they’re not gonna do FTP tests on a regular basis, if they still want to track their progress, right? So so you know, we provided a tool that allows them to kind of track how well they’re improving, and how well they’re progressing. So this is really, really important, I think, for the greater, greater population. So just in the same way, we would say, okay, you never have to do an FTP test, right, we talk about, you know, we just grab your, your threshold, power your other fitness variables from your data, we grab that automatically. It doesn’t need to be a test, it might just be a ride, you just rode on Zwift, where you just hammered it out, or it could be a group ride, right? All these are our, what we say express your fitness. And they happen with greater frequency. And these expressions of fitness are really markers. It’s kind of like measuring, you know, doing an FTP test two or three times a week. Another one of the other key things that I like to add, though, is that and Ed kind of talked a little bit about this in your last podcast, which was kind of cool, which was that trainings got to be fun. Right. And a lot of people who are looking to train, they don’t necessarily want to follow the rigorous program that would be provided by let’s say a very static training plan, or from a coach per say, who’s going to be very specific about the training that that’s going to be prescribed and how it’s going to be followed. So they really want to see how they can use the power data to give them sort of a visual and understanding of how well they’re how well they’re training and what kind of training they can do to improve without the structure necessarily, that you would see within a professional athlete, for example, these athletes have other lives, they work they travel, they’ve got to have a system in place is going to accommodate their kind of variability that they have in their life. And so I think that’s another aspect that we’re going to have to start to see in software is something that’s going to be more accessible to a broader range of people something a little more adaptive, yet still provide direction and guidance and improvements and individualization of those improvements

Challenges in training data

Trevor Connor 14:46
will both Exert and Training Peaks have been able to take advantage of this explosion in training data, there are challenges. Let’s get back to our conversation with Tim about what these challenges are. So you had a couple caveats and I definitely want to talk to you about these One was that you, you need a lot of data. That’s part one, the other one that I always have a concern about is, is it based on your best numbers, or what you might call your more actual or day to day numbers. So I’m not explaining this well. But to give an example, you always have that danger of an athlete, just cherry picking a couple key workouts where they’re putting out great numbers, plugging that into the software, and they get a power duration curve, that’s just not realistic for them. And then if you’re getting recommendations based off of not realistic curve, they could also get into trouble. So how do you avoid that? I’m sure part of the answer that is you need a lot of data, right?

Tim Cusick 15:43
Well, ah, yeah, so let’s define a lot. So at a minimal, you need 90 days worth of performing data. Now, to look at some of the historical and go beyond that, you know, obviously, you’d want more than 90, but you can begin to do a lot of the optimization, and individualization with 90 days. Obviously, if you had years worth of data, you can go and take the insights further. But the data itself is very important. So what you need in your data is one, you need accurate data, one of the things that we’ve really focused on is the ability to look at the actual recorded data and make sure that you know that data is accurate. In the power duration curve is what we call it’s a smart curve. Meaning it runs its own checks, and you can actually there’s a report that shows you the plus minus the error, the sum of errors in your power duration curve, we can actually we want people to validate their own information if they’re not sure. But we’re looking at that at all the times, if we see an error that creates more than a 5% swing, it won’t even calculate. So you have to have good data to make it work. If you have bad data, like spiky data, or your power meters really wacky or whatever, it won’t work. So first thing is you need good data. Second thing is the power duration curve is based off all data, but it’s driven by maximal data, you don’t need a lot of it. And it’s unique. Just you know, the model itself is part of learning, right? It’s part of the machine learning, it’s ongoing as learning, you need a maximal effort at short, medium, and long. And that definition of short, medium long can slightly changed by your actual performance. It’s something we call unstructured testing, we do recommend that athletes using the power duration model to track their physiological changes, you need to once every four to six weeks, just like you would do in any kind of power base training program, you need to do three tests, a short, a medium and long. And there’s a way to identify and actually the model itself tells you where to test, you have to read it someday need to automate that that’ll go on the list. But you can easily see the short, medium long place. And then you test that. And if you just keep doing those three tests once every four to six weeks at maximal effort, then the model will be pretty darn accurate.

Trevor Connor 18:03
So what are typical links for the short, medium and long? Sounds like it’s personalized. But

Tim Cusick 18:09
typically, you see something, it’ll be between 10 and 45 seconds, it’ll be between two and eight minutes. And then it’ll be out somewhere between 14 and might be all the way out as far as 40 minutes. So it really depends what you were doing and what you’re focusing on in your training and stuff like that. So it’s usually in that range, three days before, right driven townline sign and you’ve gotten a really good, maximal five second, the model already sees that. So there isn’t a need to replicate that it’s looking for. normalized residuals is is what it utilizes as the mathematical term. So it’s looking for variations off the model, and it’s picking the negative variations and saying test in this area.

Trevor Connor 18:50
Now, is there any danger? Let’s say you did a race a week ago, you were peaked for it? You had just an absolutely banner performance and you put out some of the best numbers you’ve ever seen. Is there a danger of that giving you a curve that’s actually above your your typical day to day level?

Tim Cusick 19:07
Wow. Um, you’re probably gonna have to have most of your listeners put on their flame retardant suit right now. I’ll answer that

Chris Case 19:14
I’d like to hear this answer now.

Tim Cusick 19:15
The certain models, I’m not going to use model names because people argue about this too much. And it’s an over precision in my book, but you need to know the difference to understand what what we did at least and what I believe is the best answer. certain models utilize maximal data. So the curve if you can visualize it skims across the top of maximal data to predict other maximal data. Okay, that’s not what we did. So, I have to say that under say back you, if you have that, that that breakthrough moment where you have that kind of performance, the model will adjust to it, but in a in a way that you should be able to to replicate, our model is based not on that one off. That’s why if you look at it visually, if you look at a mean maximal power line against the power duration curve, the curve doesn’t bounce across the top the ultimate uppermost points, because we don’t want to make a model that predicts some maximal performance you might have, we want a model that’s reflective of the underlying physiology. And you’d actually, when you look at our model, the curve goes through your mean maximal power line, because we want it to be something that drives the training application, so that you’re doing things that are replicatable, not just a one off. The only caveat there is, if, let’s say, you’ve got to calibrate your power meter, and you just happen to be running 20 meter, 20 watts high. Yeah, you’re never gonna buff that out, right? That’s gonna happen occasionally, and you’re never be able to, but if it’s an honest data number, nope, it’ll make adjustments, because remember, it’s gonna tweak the underlying physiology not really predict the power number.

Trevor Connor 20:52
Which brings us to a really simple but really important point for all listeners, which is if you are analyzing this data, and you want to use these tools, you have to be careful about every time you go for a ride, calibrate your power meter, and you see a rider you go, that’s not real numbers, like I’m, I’m riding with my mouth closed in the same 450 watts, not uploading that file, because if you put it in enough bad files, you, I don’t care whose tool you’re using, it’s gonna give you bad data. I had this interesting conversation with a guy two years ago who was training at a studio that I work at. And he had his FTP way too high. And he was upset about these intervals, because they were five minute intervals, and he couldn’t complete one. And he’s like, well, I don’t get it, because my FTP is x. But But I just know, something’s up here. And I go, I don’t think that’s your FTP. “O Yes, it is. I’ve tested it” I went well, what it what do you define FTP as it goes, Well, it’s power you can hold for an hour. I’m like, Can you do a five minute interval at that power? No, then it’s not your FTP? Yes, it is. And I finally said, you know, where did you get this from? Well, he had done an FTP test two years earlier, had absolutely peak numbers. And it was saying, you know, that’s my FTP for life. And you had to kindly say, you know, it’s December, you did that in July, on a really good year, you might want to adjust this. And it’s that he didn’t want to face where he was at, he wanted to see the best he’s ever done.

Chris Case 22:25
That is a bit strange to me in a in a general sense, the story makes total sense. I think what’s strange to me is people’s desire to, to hold on to these things, as long as they do if your end goal is to improve, and you know that to improve, you have to train in certain zones and do certain things at certain times. But your numbers are all off, then you should be aware that you’re going to be training incorrectly at all times, almost.

Tim Cusick 22:54
So, Chris, and that is so right, what you’re both saying is so correct. People ask me all the time, because I have such a wide exposure and professional teams and riders and groups, I get this question or similar question all the time. They want to know kind of what’s what’s the biggest observation of the mistakes with data I see. And if I was to just move away from all the fancy talk about big data and modeling and all that other stuff, here’s the answer, right? Most people, and I bet you that number 60%. And maybe maybe it was more three years ago and a little less now have their threshold set too high. That means their training zones are too difficult to achieve, which results in two problems. One, they can’t deal with the fatigue resulting from that. So they’re too tired to adapt, it’s too much work. Or two and the one that I see most prominently that people don’t let up on is that they have a high number of failed workouts exactly like Trevor’s example, wow, this athlete, Scott, they’re, you know, they’re trying to do this five minute interval at a power rate that’s too high, they fail, they can’t complete one interval, or they can’t complete the set of intervals. And then they cool down and go home and just say that wasn’t wasn’t a good day for me. I just wasn’t feeling it. And then they come back tomorrow, and they repeat the same behavior. But they can’t be honest with themselves at times and say, You know what, maybe my threshold is set too high, maybe I’m working too hard. The message there is, that’s probably that performing athletes number one barrier to improvement. And to go forward, they need to go back. And if they don’t mentally let go of that point, they’re not going to go forward.

Trevor Connor 24:28
And and to throw a third issue in there. When you’re talking about FTP, which is this estimate of your maximal lactate steady state, that’s a true physiological break point in our bodies, where when you get above that intensity, certain systems, or certain pathways in our bodies kind of shut down, others pick up. I don’t want to go too deep into that. But basically think of it as when you’re training over that that mlss point. You’re really relying a lot on anaerobic metabolism, where when you’re training at or below you’re relying you’re much more working Your aerobic pathways. So if you have your FTP set too high, what you’re going to end up doing is really, really hitting your anaerobic and pathways developing those. And I see this all the time they become these anaerobic animals, but they can’t race very well, because Cycling is still an aerobic sport.

Tim Cusick 25:17
I think that’s an excellent observation. You know, and I think what you said there, that’s important for people to recognize one of the costs, right, so one of the things I preach in all coaching and training, everything’s got a cost benefit. And one of the costs of having your FTP too high, is you are doing more work anaerobically. And what’s so important you said is there is I call it an you know, there’s an action in the reaction, right, there’s a, there’s a response to all training, and some of that is sympathetic, and some is parasympathetic. If you’re doing that anaerobic work, and you’re doing it, you think you’re working because pain is game, right. But it comes down to the energy system that you’re training, you’re developing it to perform a specific function, racing event, time truck, whatever floats your boat, but you’re training that energy system to support your desire to improve. So the cost sometimes of doing that, it might only seem like your threshold set 10 or 15 watts too high. But that might have a significant change exactly, as your example is giving on the response to the training that you’re giving. And I think people in the name of having that number be a little higher, frequently make that mistake.

3 phases of analytics

Chris Case 26:24
That’s a great place to start delving into the three phases of analytics, maybe we could start with some definitions that people may not be familiar with. I know, we’ve already tossed out the term machine learning a couple times, but maybe you could give a couple definitions to set the stage.

Tim Cusick 26:41
So when you start talking about machine learning, and artificial intelligence, right, we always put those two together. And people think that they’re the same thing. They’re two separate things. So artificial intelligence, really, to be more technical. It’s a branch of computer science dealing with simulation of intelligent behavior and computers. That’s the technical definition, right? I’m reading that off a sheet. What that means is you’re taking a program, and you’re putting a whole bunch of data and programming and information in there. And you’re hoping it makes decisions like a human an act. So that’s when you start talking about artificial and AI, that’s what it is. Machine learning is a further extension of AI, where you design a program that learns, in this case, let’s call it learns like a human or maybe better, learns like a human would learn, where you design the program to learn. And then you plug it in to a bunch of data. And it learns. And those are two different things. But you’re on the cusp of that revolution. The AI revolution is, you know, it’s part of the data revolution. That’s today’s mode. Tomorrow’s mode is evolving that into machine learning, and and that’s happening now. The AI extending into machine learning and endurance, athletics, and in all athletics is happening more and more and more.

Trevor Connor 28:01
So basically, we just talked about this revolution and data that we have had this huge amount of data now. So really, what we’re talking about is creating the the tools that can take this data and give you some some real insights. And before this podcast, you said there was essentially three phases or three stages in this process. So did you want to take us through these stages? The first one you were saying was descriptive?

Tim Cusick 28:27
Yeah. So what you have, and these are the three stages that lead up to AI machine learning. So here’s how you get there. The first phase is descriptive analytics, basically, is your compiling and summarizing reports. And then a person looks at those and they’re looking for insights. And really, we’ve been doing this for a while, I could go back 25 years when I was writing a a training diary and a bunch of notes, and I’d write down how many miles I had ridden and how fast maybe I had ridden. And then I’d look for trends. oh, I rode 250 miles this week, I rode 1000 miles this month. And if I keep riding 1000 miles, Look, my speed goes up. Right? That’s a descriptive analytic, it can go all the way back to that words. And any online program that just tracks and compiles data does a really good job of that. The step that you’re seeing really being introduced to sports now is predictive analytics. Now, predictive analytics is where the program the computer does the work of looking at all that data, looking at all the trends in the underlying data. It’s using models and math and statistics and some programming and other things in there to make predictions. And really what it is, if athlete x does this, they’re going to get that. And that could be as simple as Oh, the prediction is if this athlete does more work at FTP, their FTP will go up. Right, that seems very simplistic in general, but it gets a lot more complex than that. But that is what a prodictive analytic is it’s simply saying, if the athlete does this, they’re going to get that. Why is that important today, and you’re seeing this being utilized more and more is, if I’m a performing athlete, and I’ve just finished up a good little microcycle of training, right? I’ve just been three, four weeks of really throwing it down and putting in my work. All I know, while I’m doing the work is I should be getting fitter. Now, predictive analytics can look at that same cycle and say, oh, in X amount of time, you should be this much faster, you should be this much stronger. And you should be specifically this type of stronger and faster. So predictive analytics are what’s hot right now. And you’re seeing this, you know, in a lot of different programs, not just what we do other people out there doing it,

Dean Golich on CTL

Trevor Connor 30:45
Dean Golich a head coach of Carmichael Training Systems as part of the original group, along with hunter Allen, Dr. Andy Coggan, and Alan Lim, who revolutionized training with power back in the early 2000s. He had some interesting thoughts on where we’re at with predictive software and the importance of the power meter tools themselves. You’ll also notice that he throws a lot of terms around like FRC and CTL. As this training software revolution continues, these terms will likely become commonplace, the way FTP and normalized power have become. For now, the one you’ll need to know for this interview is CTL, or chronic training load. It’s part of the performance management chart we’ve been talking about. CTL is often referred to simply as fitness, because it’s an estimate of your overall performance level.

Dean Golich 31:29
In the teeter totter example of accurate power meters that are sold for a good cause. And then software is probably a little bit more advanced right now for the actual power meters. And we’re we’re in the midst of predictive versus post analytics. So right now, in the post, analyzing what you have done, is probably pretty accurate. And the change amount over and say the last five years, is, here’s what a group does. And now you can individualize it with your own software to know what you actually do. So post analysis, but what we’re going to see next is probably a little bit more predictive. And with that software attempting already scientific principles, so whether you would Coggan version of FRC or say I guess skiba’s version of wattage prime, those are kind of physiological principles that are getting analyzed in that software to either validate them for you as an individual, and then predict what a certain training program would end up resulting in. So I think that’s what you’re going to see down the road is to be able to say, I’m gonna match my aerodynamics, I’m going to match my red blood cell profiles at altitude, I’m going to match what my ability to recover from anaerobic effort is. And I’ll be able to do that for me, not necessarily just the people that are in the study.

Trevor Connor 33:01
It’s essentially it’s it’s an increase in individualization.

Dean Golich 33:05
Yeah, but it’s predictive. So but I don’t think that predictive there yet. The post analysis is pretty good right now. But I think a lot of people are getting a little bit mixed up or their value valuing the post analysis is predictive. And they’re weighing that too heavily. So I think that’s what we’re in the midst of our next couple of years change here.

Trevor Connor 33:29
Yeah, I agree with you actually just read an interesting study. The title is a little deceptive, saying that it compared a training plan developed by British cycling to a training plan developed by a computer, and found that the computer’s plan produced greater gains. But when you actually dug into the study, they didn’t have athletes actually use either training plan, they just simply looked at its effects on the PMC and said, the computer generated training plan produced a higher CTL. So therefore, it was better. I found that really interesting. You said, because you can go well, that’s that’s nice. But that doesn’t mean that’s what’s actually going to happen with the real athlete.

Dean Golich 34:12
Yeah, that’s not good. And that’s exactly what I’m saying is the problem is a high CTL doesn’t necessarily, and that’s what we repeat over and over and over the best predictor of performance or measure of performance is performance itself. And Andy Coggan always kind of preaches that. And if you’ve been in the sport long enough, there’s so much variability, even if you’re looking at artificial intelligence and, and so on, and how you’re going to adapt it to take so many variables in you have to have an accurate measure. And software analytics will help you with that. But a prediction of those things is very difficult on those tools. those tools CTL ATL, foreign individuals vary, and you can post you know, ad hoc post, analyze them and get to a close number. But that’s good. I mean, we’ve made a lot of headway before, you know, you’d have an error in the power meter, you don’t have an error in the measurement, you have an error in the software. You would have an error in how you smooth the data. And so all that adds up and if you said at the elite level, you had a 5% error was a big difference between getting the metal and not getting the metal. So it wasn’t that accurate? Now, I’d say that, if you have the right measurements and some of the training that got you to CTL that you best performed under for you, then you’ll probably get a result within two or 3% of what you thought you were going to.

Prescriptive analytics

Trevor Connor 35:42
descriptive and predictive analytics are actually just the first of two phases. Let’s get back to our conversation with Tim, where he talks about the phase, it’s still five to 10 years away, but maybe the most exciting, prescriptive analytics,

Tim Cusick 35:54
it’s really going to get exciting in the next couple of years, even though I think that next couple of years can be five to 10 years before we could really see it mature, is prescriptive. Analytics. So prescriptive analytics reverses predictive, right? That means, hey, I’m an athlete, and I want to get this I need more 10 minute power, I need to be able to sprint better, I need to be able to survive six short, punchy climbs on this one crit. How do I do that? And then the software can say, oh, here’s the best training for you to accomplish what you want to accomplish. That’s when you get predictive. So the predictive analytic is the athlete, the coach, whatever, says, I need to create this for my athletes success based on the demand of an event, you know, what it’s going to take for them to win an event or, or you know, a race or whatever they’re in, and then the computer will come back based on historical performance and having data on that athlete and say, oh, here’s what you should do. That’s the most effective and efficient way to create that result. That’s the future in front of us that I personally believe we’re not there yet. There’s a lot of variables, there’s a lot of things that we don’t really control. We don’t have enough data collection capability yet to be truly prescriptive. But we’re on the cusp, it’s getting there. And if it ever gets there, it will be in the next five to 10 years.

Trevor Connor 37:14
What sounds exciting about it is so we talk quickly about the performance management chart, the performance management chart just gives an overall level it doesn’t say anything about Are you a good sprinter, you good time traveler, things like that. But it sounds like what you’re talking about what is this prescriptive stage is you can say I want to win a time traveler, I want to be a stage racer, and it can really hone in on on the the assets or the physiological changes, that you need to have the right assets for the particular target that you have. Is that what you’re saying?

Tim Cusick 37:47
Yes, one caveat, though. It doesn’t mean a sprinter can become a good time trailer. Right? Exactly. But it does mean if you so made that choice, the system can tell you, here’s your training strategy. And here’s the best type of workouts you should be doing. And even beyond that, and this is stuff we’re doing now is it can tell you the exact relationship in training between time and intensity, should I do a three minute interval or a six minute interval? Well, you can use the system, the WK04 power duration model kicks off right now optimized intervals. And that might say do four minutes and 48 seconds at 365 watts. Now, I’m using examples that are slightly over precise, because there’s a slight plus minus around that. But it gives the person a coach or self coaching athlete, a much more clear target. So what prescriptive analytics will do is say, oh, here’s your training strategy, here’s the type of workouts that you should do. And then utilize the same data, the same AI, the same machine learning to say, Oh and by the way, it’s not three minutes for that interval or five minutes for that other one, you should be doing them at four minutes and 15 seconds, and then you should be plus or minus 10. You know, 350 watts plus or minus 10. So it’s actually going to even change the way we think about zone and targeting training. It’s a prescription,

Chris Case 39:11
it’s like you go to the doctor, you tell them what you want to do. And the doctor in this case is the program that interprets all the data and spits back out a prescription that says do this, do this do this in a sense,

Tim Cusick 39:26
correct becuase what you’re doing now imagine the same example right? You get sick, and you’re self treating, and you don’t have a lot of data. So you do a bunch of you do 10 different things to get healthier. You take vitamin C, you sleep more, you drink more water, who knows you wrap garlic around your throat, you know, whatever, whatever think of it right? And then if you get better, you say great, I get better. I got better. It must have worked, but you don’t know what worked. And then in the future, you don’t know how to replicate it. So if you said I get sick again, you do the same 10 things right thinking that was the mix? Well, the reality is the prescription might be it only takes one or two of the things you did and if you could hone in on that prescription, then you have more time more ability, you can just get at the things that matter because the prescription is going to say, here’s what matters, you know, here’s what’s effective and creating what you need.

Chris Case 40:11
So that that would allow riders to scrap workouts that wouldn’t result in improvements and target more workouts that would do just what they’re looking for.

Tim Cusick 40:23
I bet you that will be the number one advantage when it really all comes to bear. It’s not so much. I mean, look, there’s a lot of smart coaches and smart athletes in the world. And they’ve learned over time, like, wow, if I do this, I’m doing that. And they know how to get good on forum, they know how to train in the basics, they know how to get on peak. But what the data will tell them in the first wave and you can do it now is you can find out the non productive things. If you think about improvement, and creating, you know, cycles of improvement. Eliminating non productive behavior is the first step. And that is a big deal of training. Because if you could replace non effective training time with effective training time, generally you get faster and stronger. It’s what happens.

Chris Case 41:01
I bet there are a lot of listeners out there that are salivating, just listening to this conversation and what it will do for their training their fitness, their performances. On the flip side, there’s probably some coaches out there thinking, do I have a job in the future if if this technology comes to full fruition, and it’s able to prescribe these workouts without a coach’s interpretation? So I’m wondering if we could jump into the value for both athletes and coaches of this next phase?

Tim Cusick 41:36
I think that’s a great question. Right. And for the, you know, I’ve done a lot of webinars and training and education programs, people know that I have an honesty, a blunt honesty issue, I guess. So let me answer it that way. Let’s start with coaches. I think this is going to be one of the most dynamic tools in improving the good coach performance. Now I’m choosing my words carefully here. Because I think coaches that invest in this learning right are going to benefit. And here’s why. So think about the athlete, what I said before, you have this modern athlete, right, and they’ve got two wearables on they got heart rate monitor, they got a power meter, they got a speed sensor, they have an accelerometer in their bike. They’re they’re motioning, how much. They’re measuring how much they’re rocking the bike and where their pedal stroke is, and all of this data, right? So if you’re a coach, and you have, let’s say, I don’t know, 10, 20 athletes in your system. Remember that athlete, that individual athlete goes home, they plug in their data recording device, and they sit down at their cockpit of computer screens, which now looks like a jet fighter, right? And they’re looking at all of that data. They’re looking at it every day, every day now. And if coaches don’t get better at compiling, and utilizing that data, that’s what things like AI are there for. That’s what the software programs are out there to do to do the analytics. So you’re not doing all the legwork to boil it down into cause and effect dose and response. Is this working? Is my athlete improving? And can I communicate that improvement to them, you’re going to struggle, because the athletes can sit at their cockpit every day, they’re going to go out and ride with all of those devices, they’re going to plug in that data recording device. And they’re going to look at all of that data, and you as the coach have to boil it down to something. That’s where investing and learning and understanding what is now all of these different types of analytics, I would call advanced analytics, good coaches are going to thrive because yes, you have a learning curve. But man is it going to teach you a lot, it’s going to give you insights, it’s going to help guide the decision making process, it’s actually going to make your job easier, once through the learning curve. The downside of that is if they’re not just a good coach, I don’t say bad coaches because it sounds softmoric. But there’s people out there coaches out there that are taking on clients, and they’re not investing in their own education. And they’re not learning about the underlying physiology, the underlying methodologies, and they’re just kind of throwing workouts out there and hoping to get paid. They’re gonna fail more because your point is exactly right. The athletes themselves, eventually their cockpits going to evolve and they’re looking at this the same data. And they’re going to say, Well, wait a minute, my, my coach is prescribing this, this, this and this and that doesn’t make sense. And I can see at a physiological level, my V02 max is going down. I don’t care what my my coach might be saying my modeled FTP or FRC, those are going down what’s going on here. So the coach is going to get a couple more questions. Good coaches thrive under this type because they learn how to use a superior tool in a superior fashion. Bad coaches that don’t invest themselves get exposed, it’s going to happen.

Trevor Connor 44:42
Yeah, we were talking about this earlier. And it seems like the the coach who should be a little bit afraid is the one who just likes to pump out kind of generic six week training blocks and handing them to the athlete and say just go do the training. Because that to a degree can be automated. And as you said the athletes can now really see The effects of their training. So the good coaches are the ones who are actually gonna be using these tools to really see what’s happening with their athlete and adjust the training accordingly. Is that what I’m hearing from you?

Tim Cusick 45:12
I think that’s exactly right, the bad coach gets replaced by a training plan. It really is that simple, the good coach, the only add I would add is the good coaches are going to see better and better results with their athletes, that the machines will quickly replace the bad coaches. And that’s going to be training plans or dynamic training plans. But that’s also coming down the pipe.

Trevor Connor 45:31
My personal opinions always say that the training plan is 10% of coaching, that’s the easy part. The hard part is the sleuthing and the individual interaction and helping an athlete when they get sick, or work is hitting them hard, or they have a lot of travel, it’s all the things that unfortunately, you can’t automate.

Tim Cusick 45:49
I agree, I couldn’t agree more. It’s so funny. I’m a data preacher. As people know, you know, it’s funny, it’s like, Andy, and I talk about this all the time, it’s the science in an art, and you can’t get the both working together are the ultimate solution. If somebody thinks it can be turned all the way into science, I personally believe for everything I’m talking about machine learning and AI, I don’t think it will ever truly get there. There’s too many human variables, I think some people might be threatened by it by your point. But I don’t think will truly get there, I think the art of coaching is being enhanced, the art is being given better tools to perform with, and you’re getting better and easier access to the science. So just utilize those and become a better coach.

Dashboard of Data

Trevor Connor 46:30
Let’s check back with Armando and his thoughts and this evergrowing dashboard of data that both coaches and athletes have at their disposal.

Armando Mastracci 46:38
I believe that the software is going to start to incorporate a broader range of information. So as we instrument more of our lives, right, that this information is now going to be incorporated as part of our overall training. So one of the one of the biggest challenges with our software, I think, for a lot of software to be able to really prescribe training is that there’s just so many variables involved. So you’re only really looking at it through one lens, right, so all you have is power data, you don’t have power data for indoor rising, outdoor riser, or when we only have power on one bike and on the other bike, it’s really difficult to get a real clear picture of what this athletes, what kind of training they’ve been doing. And then when you start to incorporate other types of training, whether they’re in the gym, you know, they may have other types of sports that they want to they want to train as well. Maybe they’re a triathlete doing multiple sports, in which case you need the instrumentation across these other sports as well. I think that’s where the really the tension is going to be is how do you provide a greater holistic view of training beyond just cycling, I think cycling is great it really created a way to kind of quantify things that you’re unable to quantify in other sports, because of the availability of the power meter. But we can apply the same principles across other sports and start to quantify and use the same ideas for different types.

Trevor Connor 47:59
Now, I do also see that incorporating recovery metrics, so sleep, absolutely, yeah, this is new. You know, obviously, people, you can think of the Fitbits which track how active you are through the day? Do you see those sorts of things being incorporated into the software?

Armando Mastracci 48:17
Absolutely. So you know, when I would talk about the instrumentation, this is what we’re talking about is are we going to be able to collect sleep data, are we going to be able to collect HRV data are we going to be able to collect other data related to how our various muscles are performing or being used, like y’know maybe motion, there’s a whole bunch of information that we can collect. And with the availability of this other information, and the ability with the computing power that we have now available to us, we can start to use that information to better identify what is best suited for a particular individual, given the types of exercise that they are performing the information that you’re gathering. And then the types of athletes they want to become all that information is now going to be much more valuable and useful. Rather than getting a very narrow view of an athlete, that now you’re getting a greater 360 view of that athlete.

Trevor Connor 49:04
Now what do you see the software doing because now thats gonna be a ton of data, obviously, they’re not going to want to be sitting there trying to analyze all that data themselves or interpret it. So what do you see happening with the software? In terms of using that data? What’s the software going to do with it?

Armando Mastracci 49:21
Well you know, there’s, when it comes to things like machine learning and neural networks, and capacity of software, it uses computing power. Certainly, there’s a lot of opportunity. We do have a lot of data. But the challenge with a lot of the machine learning processes is that how do we identify what we’re trying to optimize? That’s the biggest challenge, right? So as you’re reading this paper, recently by Tanya Churchill, who’s based at University of Canberra, so part of her PhD thesis she did his sort of machine learning for training, implemented what’s called an artificial hybrid, artificial neural network to analyze all this training data for cyclists, right, which is, which is really good. If you’re ever interested in understanding kind of how machine learning will be able to be applied, what the challenges are, she really gets into a lot of great detail. So I’d recommend anybody who’s interested in this to read her report, you know, there, there are a couple of really big challenges, right? The big challenge is, is what are you trying to maximize? what’s the what’s the goal? What’s the objective, right? These are really difficult to identify. So that’s one of the challenges is to identify within the data itself, what are the goals of that particular training, and those can be a challenge. And then the other challenges is that, you know, the the data is very individualistic. Everyone comes into training with some level, some context. And u as a coach would know that, when you first have an athlete, you need to understand what the ethic learn the athlete first before you can really tell them what to do. Right, right. So part of this whole process is gathering data first, about an individual. So all this data needs to be collected first, before we can start to use it and make meaningful information out of it. So if you’re just going to show up, and you have all these new devices attached to you, you say, Okay, I’m going to get the best training, well, you’re not there yet, right, you’re gonna have to collect the data for a period of time, before we can really prescribe the training. So it’s everything is very individual. So you can’t really solve that problem by throwing more data at it. Because each individual is different. And they’re starting to see where they’re starting the process is very different.

Benefits of software to athletes

Trevor Connor 51:34
Now that we’ve talked about the impact on coaches, let’s ask Tim, what the benefits are of the software to athletes.

Tim Cusick 51:40
Bias and bias control. So much of my coach learning has come from failure, meaning so many times you’ve just been like, man, I’m working so hard with this athlete, I want to get everything right, we’re doing this training, and I’m going to implement this new idea and I’m working on that, right? and the athlete doesn’t achieve their goal and it’s heartbreaking. And you’re like, Oh, I just that’s so hard for me and you fail and you fail and you fail, and then eventually start learning from all that you get smarter and better and smarter. Now at a personal level, right? You think you could take all that learning and all those mistakes and the success you’ve had with your athletes and apply it? It’s really hard because we’re more biased when we coach ourselves than we even can admit. And the reality is we introduce bias in our planning and in our diagnosis, meaning, you know, we look at ourselves as an athlete and say, What do we really need to be successful? In our prescription? Like, what amount of work Am I really going to do and have the time for, we introduce a lot of bias and self coaching, where machine learning and AI and these type of insights will do, you should look at each phase of your self coaching relationship with yourself, I guess, and remove the bias. It’s not about what you think, like, I think I’m a better sprinter than I am, or I think I’m a better you know, I have better longer power than I really have. And you use the modeling and the data and the things that are coming out a to remove that bias as best you can. And that’s not just diagnostically but it’s workout to workout did I really do the work could I have gone harder was I going too hard? This type of analytics that we’re seeing, you know, in WK04 and by utilization of power duration curves, which give clear ideas of of what I could have been able to do. It makes me be more honest about should I have failed? Why did I fail? Did I have the right prescription? So it does remove all of that bias from the individual coaching athlete and makes them better in that way and then to it gives guiding insight, you do get those same insights where man I could do an interval at you know, I should be doing four minutes and 15 second intervals at 350 watts. That’s there also. But what you lose is the athlete is what happens if you fail halfway through that? Do you have a coach who’s well experienced and can talk to and work through. So it does help the athlete that way. It does remove bias. It does give direction and insight. But it doesn’t become the wisdom and inter exchange that you would have with your coaching. You’ll never recapture that art.

Joe Dombrowski on latest software

Chris Case 54:07
As we’ve already pointed out, amateur racers seem to be adopting the data revolution quicker than the pros. I caught up with Joe Dombrowski of the education first drapac World Tour team to get his thoughts on the latest software and why pros have been slower to adopt them.

Joe Dombrowski 54:22
Yeah, I actually was working with a guy last year on a machine learning software that was predictive in terms of basically how long a rider could sustain whatever effort they were at, in real time, based on data accumulated over hours and hours of riding, of you know heart rate and power. It seems like one of those things that’s popping up more and more now even like I’ve noticed, I use the Garmin 1030 in training the new one. Yep. And when you finish your ride, you click Save ride. It has all these metrics that pops up like your new V02 Max or your, you know, this was how much anaerobic stress you had in this ride. This is how much recovery you need. from what I gather, it seems like it learns you over rides and races and basically using your data to sort of model or try to model with suggestions for maybe how long you need to recover or, you know, how hard your ride was that sort of thing?

Chris Case 55:36
Is that something that you use at all? Or do you sort of see it as a it’s in its infancy? Maybe someday I’ll use when it gets a lot better?

Joe Dombrowski 55:47
I would not say it’s something I use a it’s, it’s interesting. I could see it being something that potentially is useful in the future. I think at this point in time, there’s so many inputs, when it comes to training and racing, and machine learning as it applies to data for us cyclists. In my experience, it seems like most of it is power and heart rate dependent. And I don’t know if there’s much else that goes into that, to get an accurate read on a lot of that stuff. It seems like the picture would need to be a bit more holistic, like, even something like how are you fueling? Are you trying to lose weight? Are you restricting carbohydrates? I guess to me, and this may change, but at this point in time, it sort of seems like well, if it doesn’t know that, how can it accurately? Do you know what I’m saying?

Chris Case 56:49
Yeah, exactly. doesn’t have all the information to make judgments necessarily.

Joe Dombrowski 56:55
Right. Right. So yeah I guess I’ve seen more of that. It’s, it’s interesting, and it’ll be interesting to see where it goes. But you know, do I make use of it? No, not really. And I would say, in general, with that sort of thing. I would say most World Tour riders are not super keen on that, I guess, I would say most riders are happy to have a coach basically tell them what to do. And they go out, they train hard, they rest hard, they focus on the basics. And really, they have killer instinct. And that’s what makes them great riders is you know, they’re highly motivated and work really hard. I guess, from my point of view, like the the tech side of it is, is interesting, but and maybe I’ll eat my words in 10 years. But like, at this point, I don’t really see it offering that much to me, it’s sort of like, last year when I was working a bit on this, this machine learning project in terms of like sort of predictive, how long you can sustain this given pace, it was interesting, but I think it would probably be most applicable to amateur riders. Because I would say in general, professionals have such a good feeling for kind of where they are in an effort. You know, it’s almost akin to like, for example, I’ve been riding this new saddle, and you have to adjust the saddle height at the seat post a bit because maybe like the you know, the stack height of the saddle slightly different to get it to the same position. This was the case this year, well, I changed saddles this year. And, just going by feel not really measuring it at all, like a ride, you know, 30 minutes and then put it where, where it feels right. And then when the mechanic came to take my measurements with the new saddle on and I was like, Okay, I’m happy with this. Now, it was on the exact same millimeter in terms of saddle height, and setback of the old saddle. So I guess that’s kind of an example of where like, you know, when you do that much riding, I think you have a really good feel for like, where you’re at, I suppose, in my mind some of this machine learning technology as it’s coming along. It could be potentially quite useful to really most of the general public, who you know, don’t spend all their working hours on a bicycle.

Trevor Connor 59:44
Now, you also said in your notes, one of the benefits to the athlete is individualization, which is one of the aspects I love Even though you were talking about the power duration curve. Everybody has a different shaped curve and that and looking at the shape of your curve can tell you a lot about, both the type of rider you are but also your strengths and weaknesses and what you should potentially be working on.

Tim Cusick 1:00:05
Yeah. Great catch. So really, when you think about what’s been the biggest improvement of the revolution, right, where we’re standing here today, my answer would be the individualization of training. That’s the umbrella that all of this improved results is living under. So 10 years ago, you had a system that was based off a number, FTP, and then you would work in training zones and levels actually is what they were originally. And you would target these levels and we would assume that everybody generally fit within the bell curve of the way the levels were written. Where today, we don’t really think like that anymore. The reality is, we’re utilizing the performing data to define the individual athletes physiology, once we understand your physiology down to muscle fiber type and VO2 Max and metrics like P max and, and your FRC and FTP and stamina and time to exhaustion. You know, we could do all that, we know your underlying physiology. So if we know the underlying physiology, the prescriptions of workouts of training strategies begin to train. So for example, let’s say I ride with the local guy here who’s a really good time triallist. And I’m, I’m more of a pursuit, sprinter style rider, I’m punchy and sprint II and I don’t like to ride my bike for more than 20 minutes hard ever. Just, you know, that just, you know it ride my bike more. So the reality is where this person is really good at time trial, we have the same threshold. But we train using eye levels and optimized intervals. And we trained together, I don’t know, one to two days a week. So when we train, we’re doing intervals together, if we’re doing longer intervals, he’s usually a little he needs to go a little longer than me I your point about optimized, and he’s typically going a little bit harder. If we’re out there doing short intervals, I might be doing 100 watts more, and my target is 100 watts more in the old system based off levels, FTP and levels. A lot of coaches would prescribe both those athletes at the same for that two minute interval, where once we understand the underlying physiology and the computer, the program is giving you some insights into the best response that you’re looking for. You can say, Wow, athlete A needs to be doing that at 500 watts and athlete B needs to be doing that at 400 watts and tweak the times that individualization whether you’re a coach or an athlete, you get the advantage of that. That’s one of the key changes. And one of the key reasons why we modeled physiology not predicted power. So we can really target the individual, right.

Trevor Connor 1:02:40
And this goes back to that whole point that FTP does not give a complete picture of an athlete. And I think of the so Chris, and I just did that climbing article with Sepp Kuss, and you look at me, I’m a time trial style rider. So my VO2 max power is very close to my threshold power, where somebody like Sepp, who’s a climber, we measured his threshold at about 326. But he was able to go out and do a 23 minute time trial at 387. So very, very different type of rider and you just can’t capture that in the threshold and the FTP.

Tim Cusick 1:03:19
Absolutely. And you know, and you see it in coaching all the time. I know you run into it, and now you can quantify it too. That was such a great article you guys did. And I really hope people read it like five times. Because no, not just to help you guys with some magazines. But the reality is, you know, you read an article and you read all the words, right? You put all those words, it’s like looking at all the trees in the forest. But I hope people step back and look at the forest, every rider right now listening to this podcast, could go out and learn that same stuff about themselves. And they could learn it and apply it and become a better climber and find ways to maximize it, you had some great insights into how you’re each climbing in there, I thought it was awesome. You could look at that same analytic, that was great how you guys did the kind of heat map approach that was awesome. that’s available to anyone right now you could totally learn all that and an individual basis just by data, you don’t need to go to a lab, you don’t need to invest in, you know, big programs and stuff like that. The power meter, you know, the data devices that you’re collecting data with can give you access to all of that all you need is some software and some learning to unlock it.

Trevor Connor 1:04:23
So I’m gonna say I hope everybody reads it five times as well. Again, not because I’m trying to sell magazines because it really hurt

Tim Cusick 1:04:34
Great stuff. You guys, that was an awesome article.

Chris Case 1:04:36
Thank you.

Trevor Connor 1:04:36
Thank you for saying that.

Chris Case 1:04:38
So Tim, to wrap up the all of the fascinating stuff that we’ve talked about today. Maybe you could bring it back and give us and the listeners out there. Some simple take homes from all this information.

Tim Cusick 1:04:51
You know, I think the simple take home it might surprise you is I think people should be encouraged by listening to this type of discussion to go out see what’s out there and learn and find out, you know how to begin to utilize this data, meaning everybody listening knows there’s more data than ever, right? Because you’re wearing the devices, you’re collecting the data, you’re plugging into the cockpit, where I think it really is such a fun time in our sport, it’s an interesting time in our sport, my number one take home is, it’s a great way to improve your training, whether as a coach working with athletes, or a self coached athlete or to somebody who wants to learn, it’d be easy to say, do it this way. And here’s my opinion and hear your take home should be trained harder. My take home is smart coaches, smart athletes, people who invest in their own learning always end up with better results, you have all this great data. And you have all this great analytic tools of which our software is not the only one, tre fix WKO4 is the product I represent. It’s in my title, so I can’t duck it. For the same side, I try other stuff, I’m always looking at other things, I really encourage people to go out there right now and look, learn. And I know it’s a little bit of learning curve, push through the learning curve of what’s happening with data. And you might find those significant improvements in training that you’ve been looking for, I think it was surprising if you put in the time to go out and experiment with some of them.

Trevor Connor 1:06:12
That’s great. So my take home, I just have have one. But you touched on this at the beginning of the podcast. And it’s such an important point to me. And it’s I say to my athletes all the time, you always have to be willing to face yourself. So to explain that. This is called a training tool, not a racing tool. So it’s not about peak numbers. Peak Performance racing is about peak performance. Training is about training, right. Meaning doing your intervals in the at the the right intensity, doing the right length of interval doing the right accumulated volume of interval. And if you are trying to fool yourself and look for peak numbers, you aren’t going to train right. So go into when you’re using these tools go in with an open mind and just say, what is this tool telling me about myself, instead of going into it saying, here’s what I think about myself, and I’m going to fudge this data until I get it to tell me what I want to see. Just let the tool tell you where you’re at, look at its recommendations for how to train and then go and try to use that to train right.

Tim Cusick 1:07:21
I think that was a great answer. I should have stole that. Sorry. No, it’s it really. Don’t try to sell validate learn. I think that’s a great point.

Chris Case 1:07:33
Hey, Tim, one other question for you. What does WKO stand for?

Tim Cusick 1:07:39
That’s super secret. If I told you we’d have to kill you. WKO stands for workout. So don’t overthink it. We just named this goes back to Kevin and gear and hunter in the gang way, way back in the day. They just needed. You know, like every file has an extension. And they’re like, what extension should we use? And they said workout and we said okay WKO.

Chris Case 1:07:58
There you go. That was another episode of fast talk. As always, we love your feedback. Email us at Webb letters at Subscribe to fast talk on iTunes, Stitcher, SoundCloud and Google Play. Be sure to leave us a rating and a comment. While you’re there. Check out our sister podcast the velonews podcast, which covers news about the week in cycling. Become a fan of fast talk on slash velonews and on slash velonews. Fast talk is a joint production between velonews and Connor coaching. The thoughts and opinions expressed on fast talk are those of the individual for Trevor Connor, Tim Cusik Armando Staci, Dean goldrich and Joe Dombrowski. I’m Chris case. Thanks for listening.