Perverse Incentives

The Current Version of my Weekly Training Log

Goodhart's law states that "When a measure becomes a target, it ceases to be a good measure." A business owner might want to know how productive her nail factory is. To measure that, she determines how many nails the factory produces in a day. To please their boss, the managers of the factory increase their nail count by producing only very small nails. Unfortunately, these nails are useless. Making the measure into the goal ends up defeating the purpose of the factory.

Quantitating progress is an activity not limited to nail factories. It is something that is part of many different human activities. For example, cyclists do all the time, me very much included as evidenced by the figure at the top of this post. We have a variety of measures for assessing our progress: miles ridden, VO2max, maximum power, and all the rest. What is so insidious is how seductive these measures can be, even when we know we should use them with care. 

For the purposes of this post I'd like to consider the case where I am not training for some specific event but rather riding for its health benefits. How best should I accomplish this? Is it even possible? In a recent post I reviewed a scientific publication that argued that there are no health benefits to exercise. That is a bit of an overstatement of what the authors actually wrote and in any case I would not reverse all my beliefs about the health benefits of exercise based on any single publication. That said, I do think this paper is a good reminder that we don't have nearly as much firm knowledge about the benefits of exercise as we wish we did. Given that necessary skepticism, what should I do? For one thing, I think I should act based on the weight of the evidence, the consensus of the medical community, rather than some philosophical standard of absolute proof.

Based on a very large number of studies carried out over several decades, the consensus of the medical community has been that individuals should participate in at least 150 minutes of moderate intensity aerobic exercise or 75 minutes of vigorous intensity aerobic exercise a week, and that ideally, twice that. There is also a growing consensus that, although 300 minutes of exercise is optimal, most of the health benefits come from the first 150 minutes. According to this, I should worry much more about doing some cycling consistently than hitting that arbitrary 300 minute goal. And yet, I am who I am, and I always want to know "how I am doing" (whatever that means.)  I collect all kinds of training statistics (e.g. measures.) And, not only do I want to know how I am doing, I am happier if I am doing better, so I try to make these statistics look good. That is how my measures become targets.

The above process led my recent obsession with mi/yr (miles ridden during the past year) as recorded in the third column from the right of my Weekly Training Log shown at the top of this post. How did this particular number end up in my Weekly Training Log? It was not in the last version of that log that I posted about back in 2021. It comes from a recommendation from Coach John Hughes, the coach that I follow; specifically, his "riding for health" training plans that I blogged about in 2022. He suggested four different training plans, depending on the athlete's cycling history, and one measure of that history was miles ridden per year. Which plan Hughes recommends is based on a point system, and an athlete gets one point if they ride less than 3.000 miles per year, two points for 3,000 to 3,500 miles a year, and three points if they ride more than 3,500 miles per year. In that 2022 post, I pointed out problems with that measure, but it got stuck in my brain nonetheless and thus I included it in my training log. While I was training for group metric century rides, I didn't pay all that much attention to that measure, but now that I am no longer participating in those rides, I am looking for some kind of goal. That goal became miles per year. Because the ups and downs of my health over the past several years, I have gone above and below that 3,000 mile goal. In 2023, I feel below that goal as a result of problems with my back, and in 2024 my training for metric century rides and my birthday ride brought me back above it. As I searched for meaning after the birthday ride, not allowing myself to fall back below that 3,000 miles per year goal became an obsession.

The miles per year measure could be tracked on a daily basis, but a training plan doesn't have much meaning day by day. I would argue that the smallest unit of a training plan that makes any sense is a week, and so I only calculate miles per year once a week. A weird property of this measure is that it depends on the riding I am doing this week, but it also depends on the riding I did a year and a week ago. Each week I get credit for the new miles I ride that week, but I also lose credit for those miles which are now more than a year old. As a result, I found it easy to stay above 3,000 miles per week at first, but as I started approaching the time of year when, a year ago, I was training first for the Art of Survival Metric Century and then for my birthday ride, the miles that would "fall off the back" each week got larger and larger and so it got harder and harder to stay above the magic number of 3.000 miles.

So at long last, I can explain why my goal of staying above 3,000 miles per year was a Perverse Incentive. Now that I am riding for health it makes no sense to overtrain today to make up for training that I failed to do six months or a year ago. That's one problem. The other is that, as is well known, miles is a flawed measure for how much cycling an athlete has done. It is flawed because 20 miles in the mountains represents much more exercise than 20 miles on flat roads. So, as I was struggling to stay above 3,000 miles per year, it became tempting to favor flat rides over hilly rides. To do so would be perverse.

Let me now turn to a less extreme example, my goal of riding for 300 minutes per week. Because this is such a well established goal of the medical community, it is much less problematic than the Hughes goal of 3,000 miles per year. And yet, it still needs to be used carefully. There are many factors I have to consider as I plan my weekly ride schedule: how I feel, what other tasks I need to accomplish, etc. Imagine I come up with a schedule that fits how tired a feel, involves rides that increase my enthusiasm, and give me time to accomplish some important tasks, but adds up to 298 minutes of riding. Now imagine I could change my schedule to make it add up to 307 minutes of riding, but only by giving up some of these benefits. Obviously the 298 minute schedule is the correct choice, but because it doesn't actually meet the arbitrary goal of 300 minutes, I confess I am often tempted by the 307 minute schedule.

Realistically, how much of a problem do these perverse incentive create? In my experience, not as much as might be feared. Let's consider the example of the Hughes goal of 3,000 miles a year. As perverse as this incentive might have been in theory, I think it was helpful in practice, and therein lies a tale. That tale starts with a question: What is a person to do to prevent a measure from becoming a goal, or at least, to minimize the harmful effects of that? My true goal was to bicycle enough to stay healthy. Therefore, what I had to avoid was to allow my attempt to stay above 3,000 miles per year to push me into unhealthy or demotivating behavior. But if I could use that measure as a motivator towards my true goal (health), I might "beat the devil" so to speak, to violate Goodhart's Law without suffering the consequences. The key reality that made this possible is that the worst thing I could have done for my health was to get discouraged and to stop cycling entirely. For health, it probably doesn't matter if I ride hills or flats, just so long as my heart rate remains above what the medical community calls "light" effort and stays within "moderate" or "vigorous" range, something that I find easy to do even on a flat ride. (A counter-example where this is an issue is my rides on my trainer.) As it happens, deciding between a flat ride and a hilly ride is not so important from a health perspective. What is terribly important is not skipping a ride altogether, something I find very tempting when I am discouraged. And in that case, looking at my progress as measured by miles per year encourages me to do something, a flat ride, a hilly ride (so long as it is not an easy trainer ride) both of which benefit my true goal of better health.

Whither my Alpine Ride?

 

Rides on one of my four Alpine routes have been of tremendous importance to me since my move to California in 2017. In my second post after moving to California, I identified a 23 mile long ride near my new home, a ride I called The Alpine Ride. Within a month, I had identified an extension I could ride when I felt like going longer, the 34 mile long Alpine-Cañada ride. When I moved from San Carlos to nearby Emerald Hills, I was able to continue on two almost identical routes that differed only in being a mile shorter. I accumulated hundreds of rides on these four routes. As a result of this accumulation of data, I was inspired to use my speed over these routes as a measure of my cycling ability (a metric known as Form) and did a statistical analysis to show that was justified. Recently, however, I have stopped riding those routes. Why did I abandon them? Is my abandonment permanent? Does it matter that I have abandoned them? What does it even mean to say that I have abandoned them?

In response to that final question, the bar graph above shows the number of Alpine rides (rides over one of the four above routes) I have ridden during each month since I move to California in 2017. There is a lot of month to month variation, including the occasional month with no Alpine rides, but overlying that is a larger trend. Starting in August of 2022, there is a fairly steady decline in the number of my Alpine rides until February of 2023 when there were none. It is not until September that there are some Alpine rides again, then another four more months with no Alpine rides, then an eight month return to those rides, and finally and most recently, I haven't ridden on an Alpine route since October of 2024, nine months and counting.

Why did I stop riding on the Alpine routes? My reasons fall into two general categories. Some months I didn't ride my Alpine rides because I wasn't riding at all, or was only doing very easy rides on my trainer. This is not so much an abandonment of the Alpine route as it is an abandonment of cycling, and for the purposes of this post, I am less interested in those months. Other months I was riding, but on routes other than the Alpine routes, and it is those months that I want to focus on here.

At the same time the number of my rides on Alpine routes started going down, in August of 2022, the number of my rides on another, similar route, the Cañada route, started going up. As I noted in an earlier post, the Cañada route is prettier and has less traffic than the Alpine routes which is why I switched.

So why does this matter? I found a new route that I liked better than the Alpine routes and I switched. That's all good, right? Not entirely. The current version of the Alpine route is 22 miles long and the current version of the Alpine-Cañada route is 33 miles long as compared to the Cañada route which is only 17 miles long. It is relatively easy to adjust my schedule so that my total weekly mileage is about the same as it used to be. What is more difficult is to find longer rides, especially rides as long as the 33 mile Alpine-Cañada route. Two 17 mile long rides on Monday and Tuesday do not have the same fitness benefits as one 33 mile ride on Tuesday. And in fact that is the explanation for the reappearance of Alpine Rides first in September of 2023 and then again in February through October of 2024. My fitness goals at both those times required the longer Alpine rides, especially the Alpine-Cañada rides.

The length advantages of the Alpine routes are especially important when I am preparing for group metric centuries, rides like the Art of Survival, Golden Hills, or Ride the Rogue. In May of 2024, I struggled to keep up during the Art of Survival, and as a result, I have recently decided  it is time for me to give up group metric centuries. Without the stimulus of preparing for these rides, the need for me to ride the Alpine routes is largely gone.

There is another advantage of the Alpine routes and that is the large amount of data I have about my speed over those routes. If I am riding one or another of the Alpine routes regularly, there will be ride to ride variation in my speed that are not meaningful, but looking at the pattern of my speed over multiple rides and comparing that to what I was doing in the past gives me a sense if I am getting faster, slower, or staying the same. Of course I can do the same thing with my Cañada rides, but there is less total data and it does not go as far back in time so is less useful.

Given the advantages of doing at least some Alpine rides, might I ride these routes again in the future? Absolutely, but it is also possible I might not. Might the fact that I haven't done an Alpine ride for nine months suggest that it is more likely I might not? Maybe, but that brings me to one more factor that has reduced my enthusiasm for the Alpine routes. The least attractive part of these routes is the five to six miles near the start which are on Alameda de las Pulgas. Although most of this stretch has a decent bike lane, the traffic is somewhat heavy and there is a lot of cross traffic due strip malls that make this stretch uncomfortable and dangerous. Recently, this stretch has been the site of construction by multiple government entities which have made these problems significantly worse. That's the bad news. The good news is that this construction won't go on forever, so I could imagine a time in the future when the construction has been completed, I grow nostalgic for these routes, and restart riding them, at least occasionally.

The Cedar-Elm Loop Go-To Ride

 

Three years ago I posted about something I call a Go-To ride. You can read that post to get the whole story but briefly it is a route that I tend to ride over and over again. In the picture above, I am showing the routes of two Go-To rides, the Tamarack Sprint on the left side of the picture and the Cedar-Elm Loop running diagonally from upper left to lower right corners of the picture. Both of these routes are loops and a ride consists of multiple laps around that loop, typically nine laps for the Tamarack Sprint and eight laps for the Cedar-Elm Loop. The Tamarack sprint is one mile around (with the part I sprint being 0.23 miles) and the Cedar-Elm loop is three miles around. I posted about the Tamarack Sprint a while back and I am describing the Cedar-Elm loop in this post for the first time.

What do these two routes have in common?

1. They are located a few blocks from each other.
2. They both involve riding the same loop multiple times.
3. This:

The Zombie with his 1963 Bianchi Specialissima at Eroica California in 2019

...not me, the guy holding the bike, but the bike itself, my 1963 Bianchi Specialissima. I raced this bike as a member of the Berkeley Wheelmen during the 1970 season, rode it from Boston to Montreal in 1972, and, accompanied by my wife Agi, completed a week-long Inn to Inn bike tour on this bike in 1979. When I restarted cycling 30 years later, it was initially on this bike. Although my Specialissima has been largely replaced by more modern bikes for my everyday riding, I still love this bike and look for opportunities to ride it.

What is it about this bike that links it to these two routes? Sadly, it is that it is now impractical for many kinds of rides. There are a number of things about it that make it impractical, but the one most relevant to this post is that it has "sew-up" tires, tires that are glued to the rims, tires which I cannot change on the road. That means that if I get a flat tire, I have to walk my bike home or to where I can get a ride home. When I first got my eBike, my Orbea Gain, I had a similar problem. Because of the carbon rims and tubeless-ready tires, I did not know how to change its tires and so I developed my Emerald Hills Go-To ride as a ride where I could walk home from any point on the ride. Unfortunately, this route would not work for my Specialissima because of another aspect of its impracticality, its lack of low gears. Because the Emerald Hills ride is in my neighborhood, it is extremely hilly and there is not a snowball's chance in Hell that I could complete it on my Specialissima. That is why I drive to my old neighborhood in San Carlos to ride my Specialissima. 

Before developing the Cedar-Elm loop, I rode my Specialissima on two routes in San Carlos, the Tamarack Sprint and the Neighborhood Go-To Rides. (Both of these rides are described in the same post from 2019.) What motivated me to develop the Cedar Elm Loop is that I have recently been finding that my body is responding well to long Zone 2 rides. (Zone 2 rides are easy, low intensity rides.) Although I have gotten better at completing hilly rides while staying within Zone 2, that requires low gears or eAssist, both of which my Specialissima lack. Both the Tamarack Sprint and the Neighborhood routes are too hilly to complete as Zone 2 rides on my Specialissima. Therefore,  to accomplish the two goals of enjoying my delightful Specialissima and completing longer rides in Zone 2 required finding a flatter route that stayed within walking distance of where I parked my car and the Cedar Elm loop was that route. Eight laps around that loop gives me a ride that is 24 miles long and takes me two hours, a respectably long ride. Although I have not yet done so, I could generate an even long ride by simply increasing the number of laps.

Doesn't riding around and around the same roads over and over again get boring? After all, I have said that I find it intolerably boring to ride my trainer for rides longer than 30 minutes. In fact, I find riding laps outdoors much less boring than riding on my trainer indoors. As proof of that, back in Houston, a large fraction of my rides were on the Rice Track, a third of a mile course which meant that even a short ride could involve 35 laps. I wouldn't want all of my rides to be on the Cedar-Elm Loop, but for one ride a week or so, the delight at riding my Specialissima makes up for a less than exciting route. Besides, the city of San Carlos is rather pretty, further reducing the potential for boredom.

Does Exercise Extend Lifespan?

---

For some time now I have wanted to write a blog post describing the limitations of scientific studies that associate health benefits to exercise. I published a gee-wiz post back in 2019 describing a study that suggested that subjects in the top 2.3% of fitness were a stunning five time less likely to die in a given year than those in the bottom 25%. (For purposes of comparison, the impact of smoking on annual death rate is a much smaller: 1.4-fold.) In that post, I did my best to describe the limitations of that study and alternative explanations for its results but I felt like there was more to say, and so I had been collecting facts and ideas to be used in a future blog post. What finally got me to write this post was this scientific study: "Does exercise prevent major non-communicable diseases and premature mortality? A critical review based on results from randomized controlled trials" by Marcel Ballin and Peter Nordström (09 July 2021) https://onlinelibrary.wiley.com/doi/10.1111/joim.13353. The short answer the authors give to the question in their title is 'No', a shocking answer indeed. As my son said when I chatted with him about this post, "That's absurd, it flies in the face of decades of research and thousands of other scientific publications." He has a point.

How can Ballin and Nordström justify such a radical conclusion, that exercise does not extend lifespan? To explain that requires some background. The first thing to point out is that this publication does not contain original data, rather it is what is called a meta-analysis. Instead of doing experiments, the authors combined the results of multiple published studies. An example of that is shown in the figure at the top of this post. This figure displays results from twelve published studies on the effect of exercise on either cardiovascular disease (e.g. number of heart attacks) and all cause mortality (i.e. how long the subjects of the studies lived.) Let's focus just on all cause mortality for which the authors looked at ten studies. In eight of those studies, the subjects who exercised lived longer, on average, than those who did not. In one study, the subjects who did not exercise lived longer than those who did. In one study, there was no difference between the two groups. However, every study has uncertainty, and in nine of the studies, due to that uncertainty, it was not possible to be 95% certain that there was any difference between the two groups. Further, when all ten studies were combined, the combined probability indicated that from these ten studies considered together, it was not possible to be 95% certain that exercise had any effect on all cause mortality. Thus, the conclusion of the authors of this meta-analysis is that there is no statistically significant difference between the longevity of those subjects who exercised and those subjects who did not.

But what about all the other studies that conclude that exercise does decrease all cause mortality? How did none of them end up in this meta-analysis? To answer that question I have to explain the difference between interventional studies and observational studies. The gold standard for scientific studies are randomized interventional studies. If you wanted to know if exercise extends lifespan, you would enroll a bunch of people in your study, randomize them into two groups (e.g. have them flip a coin, heads means exercise, tails means no exercise), have one group follow the exercise program you think will lead to a long life, have the other group not do that, and then keep track of them for decades until all of them have died. You then compare the average lifespans of the two groups. There are a lot of obvious problems with this approach. What exercise program do you test? There is almost an infinite number of possibilities. What does it mean not to follow that program? Supposing a member of the control group enjoys walking around their neighborhood. Are they forbidden to do that? How would you ever get anyone to agree to participate in such an invasive study? How would you ever get the review board at your university agree to such an immoral study? How would you ever guarantee continuous funding over the long period of this study? And what is the general public going to do during the decades of waiting before the last subject dies?

There are standard workarounds for many of these obvious problems but none of these workarounds are perfect or even close to perfect. As a result, an alternative approach to answering these questions has arisen, the observational study. The very beginnings of this field can be traced back to one such observational study done in the mid-1900s by Jeremy Morris and Ralph Paffenberger. They noticed that bus conductors, who walked around the bus during their work day, lived longer than the drivers of those buses who sat all the time. Today, such studies are usually based on a questionnaire. A list of participants is gathered, they are asked to fill out a questionnaire about how much they exercise, and then they are followed to see how long they live. This shares some of the same problems as the hypothetical interventional study I described above but eliminates many of the most serious problems and thus the vast majority of studies linking exercise to longevity are observational. This is important because observational studies have one major disadvantage compared to interventional studies; they cannot prove that one thing caused another, only that the two are linked. In the case of exercise and longevity, an observational study can show that exercise is linked to longevity, but not that the exercise caused the longevity. But what other explanation can there be? Ballin and Nordström would suggest that maybe both exercise and longevity are caused by a common factor, inherited good health for example. The idea is that a person who is born healthy will feel good and therefore want to exercise. They will also live a long time, not because they exercised but because of their inherited good health. If this contrarian opinion seems implausible to you, you can take comfort from the fact that the bulk of the medical community agrees with you. That is why my doctor recommends that I exercise and yours probably does too. That said, this disadvantage of observational studies is why Ballin and Nordström limited their meta-analysis to interventional studies. Because these studies are randomized, they are immune from this alternative explanation. If everybody's intuition is correct about the observational studies, such studies should be confirmed by the interventional studies, and according to Ballin and Nordström, they are not.

So is that it? Was all my Cycling for Health a waste of time? Not necessarily. Perhaps it is the interventional studies that are misleading rather than the observational studies. Those who would criticize the observational approach would use a catch-phrase popular in the scientific community "Correlation does not prove causation", that is, just because subjects who exercise live longer does not prove that the exercise caused the longevity. However, those who would criticize the interventional studies have their own catch-phrase: "Absence of proof is not proof of absence." Just because the interventional study failed to provide proof that exercise causes longevity doesn't prove that exercise does not cause longevity. Perhaps the wrong exercise program was tested or perhaps too few subjects were included to get the statistical significance needed for such a proof. To their credit, Ballin and Nordstöm noted these concerns and my guess is that they would be among the first to argue that the jury is still out on this important question.

So how does this publication affect the cycling I am doing? Not at all, as it happens. To me, someone with over 30 years experience as a working scientist, I am used to the eb and flow of scientific research. I believe that it is very difficult to study the effect of exercise on longevity for the reasons mentioned above and for many other reasons as well and so I am not at all surprised that definitive evidence for the value of exercise is still lacking. That being the case, I simply have to go with my gut and gamble on the answer that seems most plausible to me, and that answer is that exercise is good for my health. That said, I applaud Ballin and Nordstöm and wish them and all their fellow exercise the best of luck in the many years of exercise studies that will be necessary to provide a more convincing answer to this question. Meanwhile, I will continue biking.

Door to Door or Drive to Ride?

My Bianchi Volpe strapped to the back of my car at the start of my exploration of the San Francisco Bay trail in July of 2022. In the background is the San Francisco Bay, and to the left, the Dumbarton Bridge. The ten year old bike rack shown in this picture has since gone to the bike shop in the sky and has been replaced by a more modern rack, one that can accommodate all my bikes, something which the old one could not do.

---

There are many good reasons to prioritize riding "door to door", that is, to start and end my rides from home rather than attaching my bike to my car and then driving to the start of a ride ("drive to ride"). One reason is some combination of social responsibility and esthetics; a bike ride is supposed to be an environmentally friendly activity, a feature at least somewhat corrupted by driving to the start of the ride. Another is that a door to door ride is more time efficient. The time to load up the bike, drive to the destination, unload the bike, reload the bike after the ride, and unload it at home, all add significantly to the time it takes for the ride. When time is tight this might be the difference between fitting in a ride or skipping it.

I am far from the most fanatical of door to door cyclists. When I was an active randonneur, I used to read about those randonneurs who would bike significant distances from their home to the start of a brevet and then back to their home when it was over, making an already challenging ride even more difficult. There never has been a time when I didn't drive the start of a group ride. However, for routine training, I usually ride door to door.

Why would I ever want to do a training ride that was not door to door? I would want to in order to have access to rides that are not practical if riddent door to door. The value of such rides became much greater the day I moved from a house in a relatively flat neighborhood in San Carlos to a house in nearby Emerald Hills which was anything but flat. If I wanted a flat ride, and I did, then I would need to drive to it.

Since I restarted cycling back in 2008, the extent to which I have been willing to drive to the start of a ride has varied depending on my circumstances. At first, the only rides I drove to were group rides but in 2010 I discovered an uniquely valuable route that was not practically accessible from home, the Terry Hershey/George Bush park ride. What made this ride worth the drive was that it featured attractive scenery, was long, and was completely car-free. Thus, it lent itself to the riding needed to train for long group rides. This became important in 2012 when I started randonneuring, a version of cycling whose goal was to complete long rides known as brevets. By riding back and forth over the Terry Hershey/George Bush route a few times I was able to accomplish the 90 mile long training ride I needed to prepare for a 200 km (126 mile) brevet. When, in 2014, I decided to stop randonneuring, I decreased my riding on the Terry Hershey/George Bush route and my cycling returned to a more door to door pattern. This door to door pattern continued when I moved to California in 2017. However, I probably should have gone back to a more balanced mix of door to door and drive to ride when I moved into a hilly neighborhood in 2020, but the advantages of door to door riding and inertia delayed that rebalancing. However, by 2022, my pattern had started to shift.

The first thing, post-move, that caused me to ride something other than a door to door ride was my continuing efforts to ride my Hetchins. That Hetchins does not have very low gears so is unrideable door to door. As a way to work around that limitation, in February of 2022, I threw my Hetchins onto my bike rack and drove to the Bay Trail for a reasonably long yet flat test ride. Because that ride was successful, a month later I drove down to another segment of the Bay Trail with the Hetchins on the back of my car to ride with the local Classic and Vintage bike club. As it happened, that ride traversed some parts of the Bay Trail that I had never before ridden which made me want to explore the Bay Trail more on my own. One barrier to exploring the Bay Trail during a door to door ride is the steep climb that would come at the end of the ride to get me back home but another is the issue of time. If I started from home, by the time I got to the part of the Bay Trail I wanted to explore, I would be out of time and would need to turn around and return home leaving no time for exploration. Using the drive to ride strategy was a solution to both these problems. By taking this approach I was able to undertake a series of explorations of the Bay Trail that made my 75th Birthday Ride possible two years later.

What was good for my Hetchins was good for another antique bike I own, a 1963 Bianchi Specialissima. As is the case with my Hetchins, the very narrow gear range on my Specialissima makes it unrideable in my neighborhood, so I threw it on the back of my car, made the very short drive down to my old neighborhood in San Carlos, and the Specialissima rode again. Being willing to drive to ride has made it possible for me to enjoy this somewhat impractical but deeply nostalgic and totally delightful bicycle.

In addition to hills, another change that has impacted my cycling is that I have gotten older. These two changes exacerbate each other and together have caused me to continue to incorporate more drive to ride rides into my schedule, not just to accomodate an antique bike, but to accomodate my aging body. These days, it is more often than not that a week of riding will include at least one ride that is drive to ride, an approach which I very much hope will allow me to continue cycling for years to come.

Putting Banister To Bed

 

Graph of the Fatigue, Fitness, and Form the Banister model predicted for me based on my training over the the last few years.

---

I have blogged a lot about the Banister model. At the bottom of this post you will find a list of my earlier posts on this topic. The input to the Banister Model is daily training Load, where Load is a combination of how long I ride (in minutes) and how hard I ride (which I calculate from my heart rate.) The output of the Banister Model is predicted performance a.k.a. Form. (Think of this as perFORMance.) How well does that prediction match my actual performance? In my most recent post on this topic I said "I have been using the Banister model for more than two years now and it certainly has not been perfect at predicting the impact of my training on my cycling ability, but I have the impression that it does give me hints that are helpful in combination with what my body tells me in optimizing my training." That evaluation is both subjective and qualitative. The management consultant Peter Drucker is famous for his saying "If you cannot measure it, you cannot improve it." In this post I am going to discuss my efforts to quantitatively compare my performance to that predicted by the Banister model.

I have all kinds of subjective impressions about my performance. For example, last May I rode the 60 mile long Art of Survival with my friend Roger and felt that my performance was poor because, despite the fact that I relied heavily on the eAssist of my eBike and despite the fact that he reported that he had not had the opportunity to train enough before the event, I found it very difficult to keep up with him. Obviously that is not a metric I can use to determine if the Banister Model's prediction of my performance is accurate. What could I use? Can I look at the scientific literature concerning the Banister Model for guidance on a metric? I have read a few papers on this topic and the kinds of metrics they use tend to be things like maximum power output during a five minute time trial. (See, for example, this paper.) What I am trying to optimize is average speed during a five hour group ride which is a very different thing. The former is a measure of fast twitch muscle strength whereas the latter relies much more on slow twitch muscle fibers. It would be like trying to determine how strong your legs are by measuring your arms. I am unenthusiastic about using such short term metrics.

I could use a longer time trial to assess my Form. The problem with that is, in my old age, a long time trial is tiring to the extent that the very training I am trying to optimize is disrupted. It is similar in concept to crash testing a car if you only have one car. After the test, you might know the car was safe, but unfortunately it is now destroyed so that knowledge is of no value. So what could I measure that would be relevant to my training goals and would also not disrupt my training? If you look at my recent posts, it's obvious that I have been hoping to use the average speed at which I ride my training rides for that purpose. For many reasons, that is a very imprecise and subjective measure but back in 2021 I did a statistical analysis that my speed on selected subsets of these rides might be useful nonetheless. For this post, I used my speed on a subset of rides all of which were on my Cañada route, where the level of effort I attempted during ride were all the same, and the bicycle I rode was the same. This is how the speed on those rides compared to the performance predicted by the Banister model:

As is shown in this graph, the Banister has essentially no ability to predict my speed on a Cañada ride. R², shown at the top of the graph, is a measure of predictive power that varies between 0 and 1, where a value of 1 means perfect prediction and a value of 0 means no prediction. The value of 0.001 measured here is about as close to 0 as one is likely ever to see.

Why is the Banister model not working for me? There are many plausible explanations, but before I get into them, I think it will be useful to say a little more about how the Banister model works. As noted above, the Banister model takes as its input the Load of each of my bike rides. It then uses that to calculate two hidden variables, Fitness and Fatigue. I call these variables hidden because there is no way to measure Fitness and Fatigue on their own, instead they are used to calculate Form (performance) which can be measured. To calculate Fitness and Fatigue, Banister uses two constants called k1 and k2. It multiplies Load by k1 and adds it to Fitness and multiplies it by k2 and adds it to Fatigue. In modern usage, k1 and k2 are often customized for each athlete but out of the box, the Banister model sets k1 = 1 and k2 = 2 and those are the values I am using. If this was all there was to the model, Fitness and Fatigue would just keep increasing forever. However, because we know that Fitness and Fatigue decrease over time, every day those accumulated values of Fitness and Fatigue are decreased exponentially. Specifically:

FatigueToday = FatigueYesterday x e^-1/Ta

FitnessToday  = FitnessYesterday x e^-1/Tb

Ta and Tb are two more constants whose value can be customized to the athlete, but out of the box, have values of 15 and 45, respectively. I use these out of the box values, and as a result, my calculated Fatigue decreases by half after 10 days and my Fitness decreases by half after 31 days.

Finally Fitness and Fatigue are used to calculate the output of the model which can be measured as follows:

Form = Fitness - Fatigue

...where Form (perFORMance) is related to how fast I can complete a ride. And that is what is not working for me, the Form I calculate has nothing to do with how quickly I complete a ride.

There are reports that Banister model and the related, widely used Training Peaks software package do work for many athletes so this is a problem with me, not with Banister. Given that, I can now turn to the question of why.

1) Maybe the Banister model is not working for me because my training is too constant.

The idea here is that I try to plan my training so that I ride my Cañada rides on days when I am fairly well rested. Thus, if I never vary my Load and resulting Fatigue much, I cannot expect my ride speed to vary much either; what looks like a lack of correlation (low R²) might be a lack of variation in my level of Form caused by a lack of variation in my Fatigue on the days that I measured my performance. It is not that the Banister model cannot predict my ride speed, rather it is that I never gave the Banister Model anything to work with.

Consistent with this idea, if you look at the graph at the top of this post, it looks as if Form doesn't vary much; it is always uniformly high. This is a bit misleading; Form on that graph uses a different Y axis than does Fitness or Fatigue. (I did that to make it more visible.) In fact, over the set of suitable Cañada rides, Form varies 7-fold from a low of 114 to a high of 765. On the other hand, if I look at all days, not just the days I chose to ride a Cañada ride, the variation is 43-fold, from 21 to 894. Perhaps if I had attempted a Cañada ride on a day where my Form was 21, I might have seen a more dramatic effect on my speed.

2) Maybe the Banister model is not working for me because of confounding factors.

If Point 1 were the whole problem, we would see that ride speed doesn't vary much. In fact, it varies between 9 and 12 miles per hour, a range that feels like a big difference to me. If changes in the Banister-predicted Form, Fitness, and Fatigue don't explain this, what does? As I have noted before, I often feel tired after a stressful day, even if that stress is mental rather than physical, a day when my babysitting responsibilities are particularly heavy, for example. Depending on how large the impact of non-cycling stress is, this could obscure the effects of my training.

3) Maybe the Banister model is not working for me because training speed is a bad metric.

In all my reading, I have never come across anyone who has attempted to use training speed as a metric in the way I have. Maybe there is a reason. Coaches recommend time trials as a metric to measure Form. One aspect of a time trial is that, because the athlete is riding as fast as possible, some confounding factors are eliminated. When I go for a training ride, I might be slow because of the Fatigue from an earlier ride or I might be slow just because I am not in the mood to go fast.

4) Maybe the Banister model is not working for me because I have not customized its constants.

Most current writing about the Banister model recommends not using the out of the out of the box constants but rather fitting these constants to the data from each individual athlete. I am working on software to allow me to do that, but to date, have not completed that work and don't know if or when I will. It is possible I will complete this software, fit the constants, and all of a sudden the Banister model will start working for me. If that happens, I will post about it.

5) Maybe (gasp!) the Banister model is wrong. (Not likely, but hear me out.)

There is a lot about the Banister model that doesn't make intuitive sense to me. It is my understanding that most coaches and exercise scientists believe that it is important that an athlete not do a workout (at least a hard workout) until that athlete has recovered from the previous workout. What "recovered" means varies, but at a minimum, it is when their Form has returned to what it was before the workout. According to the Banister model using the out of the box constants, that takes 16 days. A more stringent version of this rule is that "recovered" means until an athlete's form has peaked. According to the Banister model, this would take 39 days. Both of these numbers seem totally bonkers to me. Similarly, I am puzzled that following any reasonable training program means that, according to the Banister model, Fatigue always remains high. As I was laying awake thinking about this one night before bed, I asked myself what seemed reasonable to me. It seemed to me that a linear model of recovery might better fit my experience than the exponential model of Banister, that my Fatigue would be reduced by a fixed amount each day until it was gone entirely. But by how much each day? My experience has been that it takes me two days to recover from a Cañada ride, so I set up a model where my Fatigue decreased by that much. When I fit the same set of Cañada rides I used above to test the Banister model, I again found no correlation between my level of Fatigue and my speed on a ride. However, I remembered my post from last January where I found that the correlation between heart rate and ride speed improved dramatically when I looked at rides over the course of a few months rather than over the course of a few years, so I tried again, using a shorter time frame. In addition, to improve my chances, I used a time frame during which my predicted Fatigue changed substantially. Here is what I found:

An R² value of 0.229 is very unimpressive and not very useful, but it is a lot higher than 0.001. But that is an unfair comparison, my Linear Fatigue only had an R² value of 0.009 when evaluated over all the Cañada rides. What if I looked at the Banister model over the same reduced time frame? It did improve,  with an R² value of 0.079 for Banister Fatigue and 0.082 for Banister Form but not to a degree to be useful.

Conclusion:

For some unknown reason or reasons, the Banister Model is not able to predict my performance. If that changes, I will revisit the subject, but until then, I am through blogging about the Banister model; it's time for me to put the Banister Model to bed.

---

Banister Posts:

• "Modeling Fitness, Fatigue, & Form" posted December of 2021
• "My Fitness, Fatigue, and Form" posted May of 2023
• "Cycling Career Saved" posted July of 2023
• "Update on the Banister Model" posted October of 2024

Winter Break

In the blog post I published at the beginning of December, I promised "I will do my best to maintain as much fitness as I can given the weather and will refocus come February 2025." I was probably a bit overconfident in picking February as the precise time when the weather would improve, it is March and the weather is still sketchy, but other than that I feel like I am more or less doing what I had promised, and I will talk more about that below. However, I want to underline what I assure you was a deliberate pessimistic tone to that promise; '... do my best ... given the weather ...'. What's that about?

The Weather

California has a very strong wet season/dry season pattern. During the last seven years, the wet season has run from some time in November or December through some time in February or March. If you look at my cycling log, the impact of this seasonal pattern is unmistakable. My goal for 2024/2025 was to reduce the impact of the rainy season on my fitness level.

Weather impacts my cycling both directly and indirectly. It impacts my cycling directly because, for safety reasons, I don't ride outdoors when the roads are wet. It impacts my cycling indirectly because I know that there will be days that rain prevents me from riding my scheduled ride making it unrealistic to try to work towards any goal that requires a fixed schedule. During the rainy season, my goals have to be both more modest and more flexible, I just try to do the best I can to somehow do enough cycling to maintain my health. This year, the schedule I came up is as follows:

• When the weather cooperates I ride my Cañada route, 17 miles, 90 to 100 minutes, three times a week and recovery rides on my trainer, 30 easy minutes, two times a week, a schedule I referred to hereafter as my Outdoor Schedule.
• If it rains once or twice during the week, sometimes I can move my rides around so I can still stick to the schedule above.
• When I can't just work around the rain, when it rains every day, for example, I ride on my indoor trainer, and therein lies a story.

My Best

I am a huge Clint Eastwood fan, and one of my favorite of his lines is "A man's got to know his limitations." As is the case for many people, a lack of willpower is responsible for much of my suboptimal behavior. I am of the belief, a belief I think most psychologists would support, that simply trying to conjure willpower out of thin air is doomed to failure. Rather, one must treat willpower as a limited resource and deploy it to best effect. I find riding on my trainer boring at best and miserable at worst. The harder I ride on the trainer, the less I can tolerate it. Thus, in the interests of preserving willpower, this year I worked to find routines I could do on my trainer that required a minimum of willpower while generating a maximum of fitness.

I originally set up my trainer for recovery rides, short rides at low intensity (Zone 1*) which have as their purpose facilitating recovery from an earlier, hard ride rather than generating fitness on their own. I was inspired to do this when I moved into the well-named Emerald Hills; those hills made low intensity rides on the road impossible. I find the trainer extremely boring so I originally assumed that recovery rides was all it was good for. This year, I decided to reconsider if I could also use that trainer as a way of dealing with bad weather. The key was not letting Best be the enemy Good. While it is true that given my limited tolerance for boredom meant that my trainer could never be a complete replacement for rides on the road, perhaps there were rides I could do on the trainer that would be better than an uncalled for recovery ride. 

I had previously found one possibility, my Gillen Interval Ride, six sprints in Zone 7. During the last few months, I have found another. While doing the riding that lead to my Counting Talk Test post, I convinced myself that a ride on my trainer in a 71" gear at 70 RPM was a legit Zone 2 ride while at the same time not being that much more unpleasant than a recovery ride at 55 RPM. Thus, on a day where I felt that a recovery ride wouldn't do, I could simply up my RPMs. To be honest, this is still not a great workout but it is much better than nothing and better than a recovery ride. Best of all, this ride is quite sustainable, I can ride it six days a week. If I am feeling especially motivated, I found I could increase the length of that ride from 30 minutes to 60 minutes, doubling the benefit. However, I confess that by the end of those 60 minutes my willpower is exhausted such that I would find it difficult to do more than one of these in a week. So, a sustainable schedule is six Zone 2 trainer rides a week, five of them 30 minutes long, one of them 60 minutes long. Hereafter, I will refer to this as my Indoor Zone 2 Schedule. By way of comparison, the Gillen Intervals take me about 45 minutes to complete and again, I can only manage one of these a week. So my final of the three schedules discussed in this post is five 30 minute trainer rides in Zone 2 and one Gillen Intervals ride, hereafter referred to as my Gillen Schedule.

Sometimes, depending on weather, I do a mix of my outdoor schedule and one one of these trainer schedules. Finally, I occasionally break out of my rut and do something altogether different, but for the remainder of this post I will focus on the three basic schedules described above, my Outdoor Schedule, my Indoor Zone 2 Schedule, and my Gillen Schedule.

Before moving on, I would like to share a thought: I am very lucky to have found outdoor cycling, which I can do most of the year, as a form of exercise I enjoy. Even on a day that I am not in the mood for a ride, a 90 minute outdoor ride is more pleasant and easier to complete than an easier (lower intensity) 30 minute ride on my trainer, and when the weather is nice and I am in a mood to ride, a 400 minute ride can be quite fun.

How Did I Do?

My goal was to do the best I could, given the weather, to cycle enough to maintain my health. Back in 2022, I blogged about cycling for health, describing routines recommended by Coach Hughes, the coach I follow, and by the Medical Community. I am not going to repeat that post, so if you are interested in the details, look there, but briefly, both recommend a Minimal Schedule and an Optimal Schedule. (Coach Hughes also recommends a Super-Optimal schedule which is well beyond my reach so I will speak of it no further.) 

My Outdoor Schedule exceeds the requirements of the Medical Community's Optimum Schedule and Coach Hughes Minimal Schedule. It does not meet Coach Hughes Optimal Schedule but it comes close, it checks all the boxes and has about 80% of the recommended minutes.

My Indoor Zone 2 Schedule exceeds the Medical Community's Minimal Schedule and has about 70% of the minutes of the Medical Community's Optimal Schedule. Coach Hughes is very prescriptive in terms of recommending a specific number of minutes at specific Intensities, but if I can be allowed to count minutes in Zone 2 as minutes in Zone 1, then this schedule meets Coach Hughes minimal schedule except for the minutes in Zone 3. In my opinion, this counts as quite close. However, it comes nowhere near Coach Hughes Optimal Schedule. Clearly, this schedule is way better than nothing and is way better than what I have done in prior years but is clearly less desirable than my Outdoor Schedule.

My Gillen Schedule is the trickiest to compare in that neither the Medical nor Coach Hughes recommendations anticipate the High Intensity Interval Training (HIIT) that is central to this schedule and so I have to estimate some equivalences to make that comparison. For comparison to the Medical Schedule, I am going to use the equivalence claimed in the original Gillen et al. paper that one minute of HIIT is equivalent to 45 minutes of Moderate Intensity exercise. Using that equivalence, my Gillen Schedule easily exceeds the Medical Community's Minimal Schedule and has about 80% of the minutes of the Medical Community's Optimal Schedule making it a bit better than my Indoor Zone 2 schedule. My Gillen Schedule comes nowhere near Coach Hughes Optimal Schedule. Compared to his Minimal Schedule, the total minutes, given reasonable equivalences, is pretty close. I would say it comes about as close as my Indoor Zone 2 schedule but interestingly, deviating in the opposite direction. My Indoor Zone 2 Schedule is missing High Intensity and my Gillen Schedule is missing Low Intensity. This might suggest some mix of these two schedules could be my best compromise for maintaining my health.

In summary, doing the above analysis has left me pretty satisfied that I did a reasonable job this winter of working around both the weather and my distaste for riding on my trainer with schedules that maintain my health. How does this year compare to last year?

Compared to this same time last year, the amount of riding I have been doing has gone up but my speed has gone down. I compared December 2023/January 2024 to December 2024/January 2025. Last year I averaged 208 minutes per week of riding as compared to 291 minutes a week this year. To compare speed, I considered rides on the Cañada route where my intention was a Pace ride (Zones 2 and 3) and where I completed the ride on my Bianchi Volpe (no eAssist.) Last year, I rode 11 rides that qualified and completed them at an average speed of 10.7 miles per hour. This year, I rode 10 rides that qualified and completed them at an average speed of 10.1 miles per hour. There are a lot of ways to look at this data to try to explain away this decrease, and to be honest, I don't think my speed has really decreased by 0.6 miles per hour in just one year, but I think there is no doubt my speed has been decreasing over the last several years, and that decrease does not come from a reduction in training.  This is illustrated in the next two graphs.

The first graph illustrates the drop in my ride speed since my move to California in 2017. This is one of several analyses I have done which show my speed falling over this time period. I have posted about this before. In this analysis, I collected the highest speed on the Alpine route for each month since my move and plotted that as a function of time:

The blue points connected by the jagged line are the data. The red line is the best fit to that data. The R² value of 0.34 suggests that about 34% of the variation is my monthly maximum speed can be explained by its decrease over time and that about 66% is something else. The probability that there has been no decrease in my maximum speed over time is very low. Having said that, what has caused that correlation is open to many explanations. One obvious explanation is that for whatever reason (e.g. increasing boredom) how much I have trained has decreased over time. The next graph addresses that possibility:

Here I have plotted the total number minutes I have cycled each week since my move to California. Again, the data is in blue, the best fit line is in red. Although my training has varied a lot week to week, on average, there is no correlation of that with time. Of course, there are many ways my training could have and has changed that would not be reflected by time on the bike and I have looked for and will continue to look for changes that might be relevant, but this does suggest that my decrease in speed is not due to a change to my training. The next most obvious explanation is that my decrease in speed results from aging. Eight years is a significant length of time at my age so that certainly is a reasonable suggestion.

Whatever the cause, this decrease in my speed has been very impactful. I feel like I can no longer keep up with my friends Roger and Dave even with the help of my Orbea Gain eBike. Not being able to ride with them eliminates what had been a great source of inspiration.  In the past, the need to get ready for rides with Roger and Dave gave me the inspiration to ramp back up come spring, but absent that, what do I have to look forward to this year? Last year, I compensated for that a bit with my Birthday Ride. However, as I noted at the time, that seems like that was a one time thing rather than an ongoing source of inspiration. Coming up with a new source of inspiration, with new ways to have fun on my bicycle, is perhaps my most important cycling-related goal right now.

---

* How hard I ride, e.g. how fast, can be expressed in terms of Training Zones, Zone 1 (very easy) through Zone 7 (very hard.)