Models, Measures, Performance and Passion--or, How to get Susie to Harvard

When competing in endurance events--against others, against the clock, against ourselves--it's sometimes hard to know what trajectory to take. Aim too high and you'll crash and burn. Aim too low and be filled with the worst kind of shame: knowing you could have done better. Icarus knew the former, and I'm sure all of us have silently suffered the latter. Quite simply, there's a fine line between challenging yourself and fooling yourself. As competitors, it's our job to find this line and get as close to it as we can. As athletes, it's our job over time to push this line forward. In fact, it's the battle between being realistic and pushing that line that forms our motivation and inspiration. This battle can be highly toxic and it can be extremely powerful.

Different people take different approaches to finding the line. On one end of the spectrum are the total agnostics. Agnostic translates literally from the Greek to "without knowledge." These are the people that just do it: let's go out there, bang heads, and see what happens. On the other end are the cognitive determinists: based on hitting certain points in training, we can know exactly what to expect on race day--one determines the other, completely. Most of us find our place somewhere in between.

I admit to erring on the side of agnosticism, if only slightly.
Consequently this post is written in response to a certain cognitive determinist. This guy is one smart dude, a fantastic coach, and a person for whom I have great respect and admiration. Still, I take issue with his determinism. This is why.
______________

Numbers are objective. Numbers speak reality, not magic.

Yes.
The problem is that to make predictions we have to interpret these numbers using models and, as any good modeler will tell you, "all models are wrong, but some are useful" (George Box--statistician). Add to this the problem that the numbers we use for performance are typically measures of something and all measurements are subject to error and our uncertainty increases. But forget that latter part for now; I'm more interested in models and model error as it has to do with how we predict and perceive our performances and athletic potential.

We can use objective information (or information that is almost free of error and therefore really close to being objective) to predict performance, but how we use it (our model) is always subjective. The very choice that we make to select or create a model is subjective.
In short, we must interpret numbers--they do not speak to us without interpretation--and interpretation is fallible.

Having established what I think can reasonably be treated as objective (numbers, aka model inputs) and what has to be considered inherently subjective (model outputs, predictors of performance and potential), I want to get at the latter, or how we interpret our models.

Numbers and models can lure you into a false sense of determinism; that is, if your numbers and inputs are objective, then the prediction you spit out must also be objective. If so, then we could easily predict beforehand, based on our training, what we will do in the coming race.

But you know by now that I don't believe this.

Even if the numbers were perfect, a model is just a simplified way of describing how things work and therefore spits out predictions that are imperfect. Just as there are more things in heaven and earth than our philosophy can hold, Horatio, no model can perfectly contain all that goes into determining performance. Even if the person who made the model is wicked smart.

So what does determine performance? Or, more practically, what are the most important influences on performance? I know where you think I'm going--that heart is most important thing-- And you're somewhat right, but not really.

I believe in educated predictions. I really do. It's just that our knowing has its limitations, and this is my first beef. Our predictions can be based on the facts, and yet still be wrong. This is includes knowing how a given athlete will perform on race day when solely using numbers as performance potential.

It also includes knowing how an athlete will react to a numerical analysis of her performance potential, and that's my second beef. I believe the emotional/mental effect these numerical/modeled predictors has on an athlete is variable and affects performance, positively or negatively, depending on the personality of the athlete. I would argue further that female athletes become vulnerable to psychological defeat when a coach makes sure she understands her limitations based on the evidence that has been presented by her current performance indicators. Said female athletes become discouraged, maybe even despondent. They began espousing how they are not innately athletically gifted, and how they feel that perhaps they should just throw in the towel. Their hard work cannot translate into the dreams they had for themselves, at least not in the short term--all is lost--they are not worthy--coach doesn't believe they can make onto the podium, or to Kona, or wherever they aim to go.

This is not the intent of the coach, of course. It's the coach's job to help the athlete find the line--to make sure she is both challenged and not fooling herself. The coach wants to make sure the athlete is being realistic about her potential performance. But the thing is, the cognitive deterministic coach doesn't actually know an athlete's potential, within a reasonable band of expectation. He can make an educated guess at it using the models he's created, but as we've determined, those models can be flawed. Additionally, he hasn't taken into consideration so many different variables of athletic performance: the conditions on race day, the competition present, how the athlete reacts to competition, whether the athlete is able to digest enough in terms of fuel, the way an athlete reacts to the pressure and pain that racing presents, and finally, and importantly, the feelings the athlete has about herself and her performance potential come race morning.

These last two variables, an athlete's reaction to pressure and pain and her believe about herself and her potential on race morning can be coached, but it can be a challenge to do so. Athletes must believe they can endure pressure and pain. I'm very interested in how coaches/we can train our brains to interpret extreme pain differently so that we can race through it. However, that's not what I want to get at here. What I want to get at is that other variable: how an athlete feels about herself and performance potential.

One thing I learned in my 15 years of teaching is this: tell a child she's not going to make it to Harvard--her IQ (a fantastically subjective model, I will add) just doesn't support it--and she won't. Oh, every once in a while a child (usually with amazing parents) will say FUCK YOU and prove you wrong, but usually this isn't the case. When a kid works hard, despite what her IQ or intellectual testing may indicate, she needs to be told she can go anywhere, do anything. She's smart, she's hot, she's got the world in her hands. She needs to be told this ESPECIALLY when her testing reveals that her performance needs some major improvement if she wants that Harvard degree. She is already worried that she sucks. Part of a coach's job is to coach her out of this belief--which will only hold her back. She does not need to be told her goal to get to Harvard is unrealistic. She needs to be told that she is worthy, she is smart, and that she should hold fast to that dream. As a bonus, once she believes this, going to Harvard doesn't really matter.

So this is it:

Girls/women beat themselves almost every moment of every day. Most of us totally fear we suck--some of us downright are convinced of it. We are ready for you to say we can't make it. If you confirm the fear that our goal is not achievable, we just may believe you and we just may limit ourselves to the numerical expectations you put forth. This is not solely an argument against model accuracy. The greater enemy is the self-fulfilling nature of deterministic thinking to which you don't want us to fall prey. Instead, if you hold your tongue about those numbers and don't allow subjective objectivism to get in the away, we may just move beyond the doubt in our minds and move to another level in our performances--a level you never would have predicted. If you remind us we rock, and have the potential to rock more, if you voice your pride and your belief in us, while simultaneously focusing on getting us to do the day to day work that will get us the slot and asking us to stay tough, continue to work hard to improve on our performance indicators, and keep the faith, then we will achieve our dream. And if we don't? Who the fuck cares at that point? We held fast to the dream and at least we had a chance to fly.
----------------
Last summer I competed in two 1/2 Ironman races. My performance indicators were the same for both events. I weighed the same, the hours I had put in were the same, my speed potential and durability as measured by my workouts and racing were the same. The two courses (Timberman and Clearwater) were very different, of course. At Timberman, my performance indicators put me at a 5:25. At Clearwater my performance indicators put me at a 5:06.

The model I used (some of you are familiar with it ;) made it appear that my training for each race was nearly the same. Yet it neglected some key factors. For the second race I had more intensity and speed work and I had a different psychology going in. But the model didn't capture these variables--it didn't cover all the important aspects of the physical training and it didn't include what I believe to be crucial difference between the two races--the psychological preparation. But how could it? One model can't measure everything. One model can't be solely accurate. One model can't solely capture the sum total of an athlete, her training, her mindset, and therefore her performance potential.

I completed Timberman in 5:21.
I completed Clearwater in 4:55.

In both performances I out-raced my indicators, but with Clearwater the difference was eleven minutes as opposed to four. Why?

I believe it's in the words my coach spoke to me in the days leading up to Clearwater and the day before the race. She didn't analyze my numbers. She knew them, but we didn't discuss them. She knew that what was holding me back was my fear--especially blowing up on the run if I went too hard on the bike, or telling myself I went too hard on the bike and so I couldn't run. She told me to race without fear--to just go. She believed I could do it. This would be the race of my life. I needed to continually tell myself while competing that I was having the race of my life.
Low, when I got on that bike and I began to back off, thinking about the run to come, I picked my pace up again. No fear. My bike split was a 2:34. When I got onto the run and my pace started to slow I pushed the thought that I had worked too hard on the bike out--and I ran a 1:44. It was my psychology that got me those 11 minutes. My. psychology.

To sum this marathon of a post up:
"The purpose of computation is insight, not numbers." (R.W. Hamming, mathematician and computer scientist). How insightful is it to rely solely on numbers, take passion for granted, and neglect psychology in assessing what we will do and how far will we go? Maybe as important, how fun is it?
The answer is, of course, subjective.