What I learned this year - 2018 edition.

Alan Couzens, M.Sc.(Sports Science)

Dec 12, 2018

2018 was a good year. After a couple of roller coaster seasons with the crew, it was nice to return to a little more stability this year. Of course, in the high-performance triathlon game, stability is a relative term :-)

Even in the absence of major crises, any year in the real world of endurance sport is bound to have some highs and lows. Highs: Tim Reed (pictured above) had a ~20min course PR (8:19) & his best finish to date in the Ironman World Championships. He also nabbed 3 wins across the season, in the process (including Sunshine Coast pictured above). Not an easy feat to accomplish given the immense depth of the pro field these days! There were also some really strong age-group performances, with a number of AG podiums through the year, capped off with Tamara Green and Owen Martin also having their best finishes in Kona with top 20 placings (in the World!) for each of their respective age-groups.

Maybe an even bigger high for this year though, compared to last, was the absence of major injuries among the team! There were a few, of course: A few crashes, a few persistent soft-tissue irritants that proved, well... irritating. Inaki De La Parra had a bad crash that derailed our plans for Ironman Hamburg when his fitness was at his peak for the season and Magali Tisseyre picked up an S.I. issue that led to a tough DNF (after a stellar race prep) for Ironman France but, overall, looked at in totality, a much better year than last, that resulted in a lot more athletes making it to the finish line of their major races and allowing their hard-earned fitness and smart tactics to shine.

Selfishly, these sort of years are a lot more exciting for me because I get to accumulate more true data-points and I get to learn more! When an athlete crashes out of a race, it’s easy enough to hypothesize and prognosticate on ‘what could have been’ but, bottom line, when an athlete doesn’t make it to the finish line, you lose an outcome. So, in that sense, it was a very good year. Most athletes made it to the finish line of their A Races and so we got to learn a lot.

By the numbers…

We had 31 completed A Race Ironmans among the team (at a mean finish time – male, female pros and age-groupers of 9:34. This was a little slower than my (overly optimistic :-) average prediction of 9:20 so my mean error was 14 minutes. This is an average, of course, some were very close to the predicted time, while a couple were approaching an hour off of my predicted time! At the time, this kind of discrepancy between what I thought would happen and what did happen carried an emotional punch - a horrible, almost unacceptable error that could leave one going 'all Taleb' on the issue & questioning if the sport is even controllable/predictable at all! But then, when looked at in the larger context of the other results, numerical sanity prevails and, given the factors that are less controllable in something as long and environmentally susceptible as Ironman, I’m pretty OK with 14 minutes. Never 100% satisfied, but OK, at least for now :-)

Actually, that’s not true. I was OK a week ago but now, we’re in the process of planning for a new season and so that quest to get better is in full swing. The questions are at the forefront: When I was off, where was I off? What was I missing? What factors came into play that my model wasn’t accounting for? Can I account for these factors, in the training plan, to lead to greater impact over race outcomes next year?

So, that’s the subject of this post – what did I discover was missing last year that I aim to rectify as we head into 2019?

Of the races that fell more than one SD from my prediction, i.e. the races that I was horribly wrrr…

....where my predictions were way off the mark, there were 3 primary factors at play…

  • Individuality/elite physiology
  • Heat/Environment
  • Pacing

1. Individuality/elite physiology

As I work with more and more elite athletes, one thing becomes apparent – they’re different. Not just different from us, but different from one another. I wrote about that in this post on how I program differently for ‘volume vs intensity responders’. In that post, I came to the conclusion that a general load model really won’t do for athletes at that level – where it’s not as much about CTL as it is about the interplay between different types of training.

Additionally, as I outlined in another (surprisingly controversial) post, I found that the TSS/Banister model really breaks down at those upper echelons and over-predicts performance for elites and high level age-groupers. I found this confirmed in this year’s analysis – where a good amount of my overly optimistic errors were often as over-prediction among the higher level athletes at the high levels of load. If you use the Banister model for an elite population, you’re going to be wrrrrong a lot of the time. I’ve found a big improvement in this as I’ve started to make the move to better models – namely Neural Networks as my predictive model of choice. I’m finding that, for the average athlete, a simple volume-intensity Neural Networks can predict performance with ~half the (root-mean-square) error of the Banister model, but significantly more so for elites.

Another area that I feel I fell short in this year wrt individuality comes in the risk management side of the equation. After being burned with a lot of injuries/illnesses in the team I was perhaps too cautious (and general) in risk modeling. Individual injury/illness models are much tougher than individual performance models but they are an important piece of the puzzle, esp for those looking to achieve their full potential in the sport.

Needless to say, I will be continuing to work with & develop Neural Networks both in predictive performance modeling and predictive injury/illness modeling in 2019 and beyond. As data builds, hardware performance increases and price drops, I’m excited to apply Deep Learning techniques to these sports performance tasks in the coming years.

2. Heat/Environment

Last year I proposed a very simple model based on the data I’ve observed regarding the impact of temperature on race performance in Kona, coming to the (unsurprising) conclusion that heat & humidity have a huge impact on performance. While bordering on common sense, I find that I am still not sufficiently emphasizing and quantifying these effects. We do benchmark sessions in home environments and ‘good’ key workouts can turn out to me as much a curse as a blessing, as they give the athlete a false confidence of what to expect in the literal heat of battle. At the risk of stating the obvious – athletes can expect to be quite a bit slower in hot race! Less watts on the bike, and especially slower pace on the run! Of course, the best solution in determining just how much slower is to do these key workouts where you will be competing. Though, this is obviously not always practical and so, in addition to specific physiological preparation (heat acclimation sessions etc) we need some sort of good, realistic estimation of the likely environmental impact on performance. Heck, we need a good environmental model! I will be working more on this over the coming year. A big issue with not having a realistic appraisal of output in different conditions is that it compounds the error when it also factors into the following factor that is a big player in ‘poor’ performances.

3. Pacing

Over-ambitious pacing is a double-whammy when it comes to not getting the race that you deserve from the fitness you take in. From a pure physics perspective, athletes who put out big efforts in high drag portions of the event are penalized on the finish clock by throwing watts/speed into the wind/water. More of the team started using run power meters this year and I wrote about how we can use run power to help us better allocate our whole-day energy resources here. So, that's one factor - sub-optimal bike/run distribution.

Beyond that, physiologically uneven pacing within the individual segments of the bike & run is exponentially costly. Glycogen use doesn’t increase linearly with effort increases, but at an exponential rate. A 2x increase in power for instance, will amount to an ~4x increase in glycogen use! So, the penalty of any surges/periods above the mean pace of the race is more costly than most athletes are aware. When coupled with heat, this is an even greater factor as body temp increases exponentially with increases in intensity and is very tough to bring back down. Couple these factors together and you end up with athletes who over-pace portions of the race are hugely penalized and end up performing well under what their fitness would indicate.

I think the solution to this is 2 fold – part me, part you. Better pacing models for the different courses/different conditions but also, and more importantly, more humility from the athlete when it comes to pacing – if your HR is higher than predicted for power – listen to it, don’t write it off. On the other side, if power is way over cap in order to ‘bridge to the wheel’ take a pause and assess whether that wheel is really worth it!

This aspect of the race is becoming more and more of a concern as athletes view the bike portion of the Ironman as a bike race, subject to the same concerns of ‘holding the wheel’ as your local crit. At the risk of stating the obvious, it is not. A wheel 12 meters up the road is not worth the same as a wheel 12 mm up the road. Is that wheel 12 meters up the road worth zero? No. Is that wheel 12m up the road worth blowing your run up for? No. I even added a 'legal draft' option to my power calculator to help better quantify the relative benefit of holding a legal draft. If you play around with the various scenarios and see how little the true benefit really is in most situations (at least compared to 'real'drafting), you'll see the message is clear – use what is around you legally & to full effect but above all else – race your own race, i.e. within the fitness limits that you bring to the event!


A few other random thoughts that came to mind in the year just passed…

* Pro’s make their living by racing (not training).

A couple of decades ago, I remember Phil Maffetone pushing strongly for a young up-and-comer named Tim Deboom to hold off turning pro (despite the fact that he was winning everything at the time as an age-grouper). I didn’t necessarily get that at the time but now I do. “Turning Pro” carries its own pressures, namely the pressure to race frequently (giving sponsors exposure etc). This pressure can make it tough to create enough space within the year to put in the training to get better! And, I think it is one of the reason that true “breakthrough performances” are quite rare among the Pro’s. It’s very tough to get better when most of your year is spent racing. Needless to say, I concur with Phil, if you are a strong amateur with pro ambitions, in a currently stable position, stay there and become as good as you possibly can before ‘hitting the circuit’.

* Training Camps are incredibly valuable (to both the athletes and the coach!)

As I outline above, performance models are really helpful in identifying unforeseen factors that lead to an athlete underperforming relative to fitness. However, the flipside of this occurs when an athlete’s performance numbers ‘take off’ and are unexpectedly better than the model would predict. Again, I was reminded this year that Training Camps (especially but not exclusively altitude camps) consistently lead to this situation where an athlete is getting a lot more out of the training weeks than the pure training load would suggest. I’m not 100% sure on the reasons for this but I suspect (in addition to altitude where applicable), it is a combination of – a simplicity of life/minimization of other stressors – i.e. a pure focus on eating, sleeping, training coupled with the social energy that comes from ‘sharing the load’ with others with a common goal that is very powerful. Providing the athlete is able to stay healthy and absorb the camp, a significant performance boost (above and beyond the load accrued) is inevitable.

Apart from the performance benefit, I’m also seeing regular Training Camps as a great alternative for ‘mature’ athletes (and 'mature' coaches :-) to the live-in squad scenario. Ironman is a 30-40 yo’s game and generally by the time an athlete hits their 30’s they are growing into the rest of their life as well. Whether relationships, other business opportunities, family or just more of an independent life, that decade typically marks a turning point from the ‘dorm life’ of the 20’s. And while it is hard to deny the power of the athlete’s version of ‘dorm life’ i.e. residential athletic programs, that type of living has a finite life span (equally so for coaches who want to avoid burnout and be in the game for a long while!) I’m becoming aware that regular camps offer a great middle ground for this and, honestly, may even be better in a lot of ways to full time ‘live in’ situations as the athlete can periodically remove themselves from the stress of the competitive environment, to retain life balance & recover. Needless to say, I will be aiming for many more #MADcrew excursions in 2019 and beyond!


I’ve been doing these “what I learned” posts for a while now and maybe the biggest thing that I can learn from looking back on my “what I learned” posts is my increasing tendency towards objective measures as the mark of what's important. Our brains are very good at creating explanations and finding connections, sometimes where there are none. Mine especially! :-) I’ve long given up the quest for ‘figuring out the puzzle’ 100%. Training a host of different types of athletes for an event as complex as Ironman triathlon is an ever-evolving game - the puzzle pieces are constantly shifting, which can be both wonderous and maddening at the same time! So, in the interests of my own sanity, my quest is refined: Every year improve just a little bit and be a little less wrrrrrong. :-)

Train smart,



Don't miss a post! Sign up for my mailing list to get notified of all new content....


Have no fear - I won't spam you or sell your info.